End-to-End Strategy Execution

Business architecture experts define synthetic stakeholders as AI agents and advocate for formal governance to manage AI deployment risks.

52 min Listen to the original Episode page
Business ArchitectureAI AgentsWorkforce AutomationProcess AutomationCorporate GovernanceRisk Management

Summary

William Ulrich and Speaker 1 define 'synthetic stakeholders' as AI agents acting as autonomous software systems within business architecture frameworks. They argue that traditional deterministic systems and flow-based process models are too rigid for AI, necessitating dynamic rules-based routing and state machines. The speakers warn against unmanaged AI deployment, citing rising AI-related layoffs and the risks of 'shadow systems' like spreadsheets. They emphasize that business units, not IT, should manage AI agents, with strict security privileges and formal documentation to prevent chaotic 'AI agent wars'.

Key takeaways

  1. 1
    A Fortune piece published on October 22, 2026, claimed AI could displace one in 11 workers, while another study indicated AI-related layoffs quadrupled from 2024 through late 2026.

    Unidentified speaker (speaker_1) · 8:58

  2. 2
    Shadow systems, particularly spreadsheets, are described as rampant in organizations, with tens of thousands in small-to-mid-sized firms and hundreds of thousands in larger ones, posing significant risk if they break or data changes.

    Unidentified speaker (speaker_1) · 14:00

Episode sections

  1. 0:04 – 1:36

    Introduction and Definition of Synthetic Stakeholders

    William Ulrich introduces the topic of synthetic stakeholders as AI agents replacing or supplementing human roles. He defines an AI agent as an autonomous software system using AI models to pursue goals and execute workflows on behalf of humans, noting that 'synthetic stakeholder' is a business architecture term for this role.

  2. 1:36 – 4:36

    Business Architecture Context: Value Streams and Capabilities

    Ulrich recaps that value streams are enabled by capabilities, which deliver outcomes to customers, partners, and human stakeholders. He states that capabilities are powered by people and technology, and shifting towards more technology results in synthetic stakeholders. He excludes business processes from this empowerment list, citing a previous podcast.

  3. 4:36 – 8:50

    Stakeholder Mapping and Human Resources

    The speaker explains that the 'Stakeholder Map' is one of the 10 domains in business architecture. He categorizes stakeholders into Customer/Partner (seeking value) and Human Resources (employees, contractors, volunteers). He emphasizes that contractors are human resources regardless of contract type and map to specific value stream stages.

  4. 8:50 – 11:40

    Automation Trends and AI-Related Layoffs

    Ulrich cites a Fortune piece from October 22, 2026, claiming AI could displace one in 11 workers. He also mentions a study showing AI-related layoffs quadrupled from 2024 through late 2026, with over 60 CEOs announcing layoffs for efficiency. He warns against random layoffs without a formal framework.

  5. 11:40 – 18:50

    Risks of Unmanaged AI Deployment and Shadow Systems

    The speaker argues that IT should not manage AI agents; the business should. He warns against 'AI wars' where agents compete for supremacy. He criticizes the proliferation of redundant business processes and 'shadow systems' like spreadsheets, suggesting leadership should phase out overpaid contractors maintaining these systems rather than cutting key architects.

  6. 18:50 – 24:58

    Value Streams and Operational Scope

    Ulrich states that a company typically has 30-35 value streams, rarely exceeding 40, determined by business model complexity rather than size. He introduces 'Execute Operation' as a multi-purpose, background value stream present in all organizations, involving work cycles and assets.

  7. 24:58 – 28:27

    Stakeholder Personas and Decomposition

    He lists typical stakeholders (e.g., risk managers, accountants) and notes that job titles are insufficient for defining roles. He provides examples of persona descriptions for risk managers and accountants, suggesting that agentic AI teams should decompose these roles further and apply constraints.

  8. 28:27 – 32:12

    Stakeholder-to-Business Unit and Capability Mappings

    Ulrich identifies two critical mappings often missed: Stakeholder-to-Business Unit (enabling global AI agent reuse) and Stakeholder-to-Capability (where capability behaviors differ by value stream stage). He uses the example of customer preference management varying between marketing and agreement-oriented value streams.

  9. 32:12 – 33:33

    Dynamic Rules-Based Routing and Value Streams

    Speaker 1 discusses mapping stakeholders to capabilities and introduces dynamic rules-based routing as a design discipline constrained by value stream stage boundaries. This approach uses state machines to manage nonlinear work items across multiple stakeholder queues simultaneously.

  10. 33:33 – 34:08

    Take-a-Trip Value Stream Example

    The speaker illustrates dynamic rules-based routing using a 'take-a-trip' value stream, where a passenger remains in a stage until specific states (like landing or exiting a vehicle) are met. This example highlights how state checks allow navigation within and across value stream stages.

  11. 34:08 – 37:16

    Business Architecture Guild Resources

    Speaker 1 recommends joining the Business Architecture Guild to access reference models, such as the transportation model, which contain detailed value stream examples. These resources help organizations understand how to apply these architectural concepts.

  12. 37:16 – 38:22

    AI Agents vs. Deterministic Systems

    Citing an IBM article by Alex Draley, the speaker argues that deterministic systems are brittle and struggle with ambiguity, whereas AI agents offer adaptability. The speaker contrasts this with flow-based business process models, which he claims are too rigid for AI deployment.

  13. 38:22 – 41:00

    Critique of Business Process Models

    Speaker 1 references his October 21st podcast episode to criticize flow-based business process models as the 'antithesis' of viable AI foundations. He asserts that these models rely on mythical happy paths and cannot handle the edge cases inherent in real-world operations.

  14. 41:00 – 52:01

    Security and Privilege Limits for Agents

    Referencing a Meritalk article by Gail Emery, the speaker advocates for AI agents to have minimal privileges rather than inheriting employee access rights. He links this to business architecture reference models, suggesting that security capabilities should be automated and tied to business objects.

Claims worth verifying

AI could displace one in 11 workers, according to a Fortune piece published on October 22, 2026.

Unidentified speaker (UNKNOWN) ·8:50

How to verify: Search Fortune archives for articles published on or around October 22, 2026, discussing AI displacement statistics. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

AI-related layoffs have quadrupled since 2024 through late 2026.

Unidentified speaker (speaker_1) ·9:18

How to verify: Identify the specific study cited by the speaker and locate its publication or data set to verify the layoff trend figures. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

60 plus CEOs have announced that layoffs due to AI achieved efficiencies.

Unidentified speaker (speaker_1) ·9:18

How to verify: Search news databases and press releases for announcements by CEOs in the 2024-2026 period citing AI-driven layoffs for efficiency. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

A typical company has 30 to 35 total value streams, rarely exceeding 40.

Unidentified speaker (UNKNOWN) ·19:24

How to verify: Review business architecture literature or case studies referenced by the speaker to verify typical value stream counts for complex organizations. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

The Stakeholder Map is one of the 10 domains in business architecture.

Unidentified speaker (speaker_1) ·4:51

How to verify: Consult the Business Architecture Guild's BIZBOK Guide or official documentation for the list of business architecture domains. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

Business Architecture Guild provides reference models that include stakeholder descriptions and value stream cross-mappings.

Unidentified speaker (UNKNOWN) ·27:33

How to verify: Access the Business Architecture Guild's website and download their industry reference models to check for stakeholder mappings. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

A manufacturing reference model from the Business Architecture Guild includes stakeholders like inventory managers, part managers, or material managers.

Unidentified speaker (speaker_1) ·28:09

How to verify: Download the Manufacturing Reference Model from the Business Architecture Guild and inspect the stakeholder list. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

Alex Draley wrote an article titled 'Flexible by Design, Reliable by Proof: Building and Evaluating AI agents that work in the real world' on IBM's website.

Unidentified speaker (speaker_1) ·37:16

How to verify: Search IBM's website or corporate blog for the article by Alex Draley with the specified title. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

The speaker referenced a podcast episode from October 21st discussing the shortcomings of flow-based business process models.

Unidentified speaker (UNKNOWN) ·38:22

How to verify: Check the podcast archive for 'End-to-End Strategy Execution' or 'Strategy Execution Podcast' for an episode released on or around October 21st (year not specified, but context implies recent). For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

Gail Emery wrote an article titled 'Agencies Urge AI Agent Access Limits and Weigh Human Oversight' on Meritalk.com.

Unidentified speaker (speaker_1) ·41:00

How to verify: Search Meritalk.com for the article by Gail Emery with the specified title. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

Kate Kellogg is a professor of work and organization studies at MIT Sloan.

Unidentified speaker (speaker_1) ·43:52

How to verify: Check the MIT Sloan School of Management faculty directory for Kate Kellogg. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

William Ulrich is the co-author of 'End-to-End Strategy Execution from Inception through Solution Deployment'.

Unidentified speaker (speaker_1) ·36:10

How to verify: Check Amazon.com or publisher listings for the book 'End-to-End Strategy Execution from Inception through Solution Deployment' to confirm authorship. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify the Fortune article from October 22, 2026, regarding the 'one in 11 workers' displacement claim and the source of the AI-related layoff quadrupling statistic.
  • Review the Business Architecture Guild reference models, specifically the transportation model, to understand the proposed value stream examples.
  • Check IBM's article by Alex Draley on the limitations of deterministic systems versus AI adaptability.
  • Read Meritalk's article by Gail Emery on AI agent privilege limits and security capabilities.
  • Research MIT Sloan professor Kate Kellogg's work on 'agent supervision ambiguity' and its implications for management.
  • Audit current organizational 'shadow systems' (e.g., spreadsheets) to assess data integrity risks and redundancy.
  • Evaluate current IT vs. Business ownership models for AI agent deployment to identify potential governance gaps.
Full transcriptexpand
0:04
Unidentified speaker (speaker_0): Welcome to the Voice America Business Channel show that reveals the real reason strategy execution keeps failing. William Ulrich, one of the authors of End to End Strategy Execution, is here to clear the fog and guide you toward meaningful results. Now, here's William.
0:24
Unidentified speaker (speaker_1): Welcome to the Strategy Execution Podcast. on VoiceAmericamedia.com. I'm your host, William Ulrich. Today's podcast is titled Synthetic Stakeholders in the Age of AI. This topic is a follow-up to last week's podcast, which was titled Dynamic Value Stream Realization.
0:42
Unidentified speaker (UNKNOWN): At
0:42
Unidentified speaker (speaker_1): the back end of that episode, I discussed one element of dynamic value stream realization being the transposition between human stakeholders.
0:51
Unidentified speaker (speaker_1): and synthetic stakeholders. In general, a synthetic stakeholder is a business term for an AI agent, which can be which is more widely used in terminology.
1:03
Unidentified speaker (UNKNOWN): An
1:03
Unidentified speaker (speaker_1): agent, just so we're on the same page, an AI agent is an autonomous software system that uses an artificial intelligence model, usually a large language model, to pursue a goal. plan and execute workflows, make decisions and execute actions on behalf of a human.
1:21
Unidentified speaker (UNKNOWN): So
1:21
Unidentified speaker (speaker_1): when I say synthetic stakeholder, you can consider these somewhat interchangeable. But I think the term synthetic was more appropriate as we started looking at business and business architecture, but the role will be played by an AI agent.
1:36
Unidentified speaker (UNKNOWN): So
1:37
Unidentified speaker (speaker_1): for listeners that missed that episode, let me just do a quick recap. Formally defined business architecture value streams are enabled by capabilities. For readers of my book, listeners of my podcast, you'll be aware of that. If you need more information on business architecture capabilities, value streams, take a look at my end-to-end
1:56
Unidentified speaker (speaker_1): strategy execution book, and I'll refer to that a couple times today. And the appendix has a nice little overview. of business architecture.
2:05
Unidentified speaker (UNKNOWN): So
2:05
Unidentified speaker (speaker_1): these capabilities deliver outcomes that collectively deliver value, and they deliver value to customers, those are our main priority, partners, business partners, and human stakeholders. Capabilities are in turn powered by people and technology or some combination thereof. Hence, we
2:25
Unidentified speaker (speaker_1): can say, well, if it's less people and more technology, we start getting into the realm of the synthetic stakeholder. I purposely excluded business processes from my capability empowerment list for reasons that I explained in my October 21st podcast, which was titled Moving Beyond Business
2:45
Unidentified speaker (speaker_1): Process Models.
2:47
Unidentified speaker (UNKNOWN): So
2:47
Unidentified speaker (speaker_1): in reality, the focal point's gonna going to be on people and technology. I'll get into the
2:53
Unidentified speaker (speaker_1): lack of the role of a business process in AI as I go through my discussion today, but I'll be more specific as to and cite some different industry literature and some other areas, why business processes as the flow-based approach to business processes are not the ideal.
3:13
Unidentified speaker (UNKNOWN): So
3:13
Unidentified speaker (speaker_1): we'll get into more of that later.
3:15
Unidentified speaker (speaker_1): So uh combination of people and technology coupled with high integrity information. I always want to mention that. You guys are probably sick of be hearing me talk about good quality data. There will be a lot more on that in this podcast series as well.
3:29
Unidentified speaker (UNKNOWN): Uh
3:30
Unidentified speaker (speaker_1): those things allow an organization to thrive and become an industry leader, or in cases where the data's data is bad, the technology lags, the people aren't doing their job.
3:39
Unidentified speaker (speaker_1): then what happens is you lag behind the competition, lose customers, fail to deliver on your mission, and in a worst case scenario, you end up closing your doors. So for our discussion today, consider a stakeholder, an individual, or an organization that's either seeking value from the organization or contributing value to the end state
3:59
Unidentified speaker (speaker_1): being sought. So, a uh, quick example, I have a customer, they come in, uh, they are going to file a claim with an insurance company, right? And or they maybe they want to obtain a product. But in most cases, a stakeholder or multiple stakeholders are going to be involved. In addition to the customer themselves,
4:19
Unidentified speaker (speaker_1): maybe you have partners working with you, uh, you have other workers such as claims agents and so on.
4:25
Unidentified speaker (UNKNOWN): So
4:26
Unidentified speaker (speaker_1): getting into a little more on that, we'll talk specifically about what some of those roles are and how they break down into stakeholders, and they can break down to more detail and so forth.
4:36
Unidentified speaker (UNKNOWN): But
4:36
Unidentified speaker (speaker_1): I want to get into some formal terminology that business architecture uses. The human aspect of powering value streams is captured in the stakeholder map. That is one of the stakeholders, one of the 10 domains in business architecture.
4:51
Unidentified speaker (speaker_1): For those new to business architecture and think that a business architecture is essentially a capability map, it is not. It specifically is not. Capability is one of 10 domains and the only one of the four core domains. Customer and partner categorized stakeholders represent personas seeking value from the
5:11
Unidentified speaker (speaker_1): organization.
5:12
Unidentified speaker (UNKNOWN): So
5:13
Unidentified speaker (speaker_1): those are both important to the customer and the partner experience. We looked at customer experience recently in an earlier podcast this season. We can also look at partner experience. We can also look at the human resource experience. So we can expand that thinking. The human resources, and I gave a couple examples earlier, but the
5:32
Unidentified speaker (speaker_1): human resources, which can include employees, contractors, volunteers, and other workers. under the business ecosystem's domain of control, are the human aspect of engagement with the value stream. So we have different stakeholders playing different roles to help move things along. We document that and are documented
5:52
Unidentified speaker (speaker_1): in the business architecture.
5:54
Unidentified speaker (UNKNOWN): And
5:54
Unidentified speaker (speaker_1): yes, a contractor working for a partner in some capacity for the organization is a human resource. It doesn't matter that they don't have.
6:03
Unidentified speaker (speaker_1): an employee contract, employer-employee contract. They're still a human resource. And I have worked at some organizations where you couldn't really tell the difference between the employees and the contractors.
6:14
Unidentified speaker (UNKNOWN): So
6:14
Unidentified speaker (speaker_1): the human resource is again playing a role to further value delivery in one or more value streams. Stakeholders are associated with, or what we say in business architecture, map to each value stream stage where they play a given role.
6:27
Unidentified speaker (UNKNOWN): So
6:27
Unidentified speaker (speaker_1): under one or more scenarios, you're going to see a claims agent or some other kind of worker helping a customer who's making a claim.
6:38
Unidentified speaker (UNKNOWN): So
6:38
Unidentified speaker (speaker_1): let's take a look a little closer at this example. So a claims agent, right, for an insurance company may be involved in multiple stages of a settled claim value stream, along with an insurance agent. maybe a repair specialist if it's a vehicle.
6:53
Unidentified speaker (UNKNOWN): If
6:54
Unidentified speaker (speaker_1): it's a healthcare situation, you'll have healthcare professionals, a claims officer, an investigator, and even other personnel. This list would also include the customer who filed the claim.
7:06
Unidentified speaker (UNKNOWN): A
7:06
Unidentified speaker (speaker_1): triggering stakeholder, in this case a customer, naturally participates in any value stream where they're seeking value. Of course, you're going to have to the customer involved. So the customer in this settle claim value stream is going to be interacting with all of these other stakeholders as well as technologies. And again, we'll get back to more of the automation aspect of
7:26
Unidentified speaker (speaker_1): this, but we we like to understand that context of that engagement.
7:30
Unidentified speaker (UNKNOWN): Now,
7:31
Unidentified speaker (speaker_1): if one or more stakeholders gets automated, that's not going to necessarily change anything other than maybe we're hoping it's going to get done more quickly.
7:38
Unidentified speaker (UNKNOWN): So
7:39
Unidentified speaker (speaker_1): we normally think of again as these stakeholders as people, but more and more we're seeing stakeholders being replaced by automation,
7:46
Unidentified speaker (UNKNOWN): which
7:46
Unidentified speaker (speaker_1): of course predates the AI explosion. AI is just a smarter way of using technology to play the role of a stakeholder.
7:55
Unidentified speaker (UNKNOWN): So
7:55
Unidentified speaker (speaker_1): with AI expanding, we're seeing that technology taking over more and more stakeholder roles, and we're seeing that at an increasing rate.
8:02
Unidentified speaker (UNKNOWN): Now,
8:03
Unidentified speaker (speaker_1): it all depends on what stats you believe, and we'll talk about some of those and whether or not these hold true, or you have competing studies and so forth. And I'm not here to really judge that one way or the other. What I'm talking about is if it's happening, here's what you should do as a business leader to think about those things. So, as
8:23
Unidentified speaker (speaker_1): an aside, uh you can get more specifics on all of these business arguments. architecture terms like capabilities and value streams and stakeholders, uh, information concepts and so forth, as well as the IT interactions with those in the appendix of our strategy execution book, end-to-end strategy execution.
8:41
Unidentified speaker (UNKNOWN): Now
8:41
Unidentified speaker (speaker_1): I'm focusing this podcast episode on the topic of AI filling stakeholder roles based on the real-world real world backdrop that it's happening, right? It's happening all around us.
8:50
Unidentified speaker (UNKNOWN): We
8:50
Unidentified speaker (speaker_1): know that. this because companies are saying that they are replacing people with AI.
8:54
Unidentified speaker (UNKNOWN): Now,
8:55
Unidentified speaker (speaker_1): you may or may not believe them, and that's your decision.
8:58
Unidentified speaker (UNKNOWN): But
8:58
Unidentified speaker (speaker_1): according to a fortune piece published in October 22nd, 2026, very recently, AI could displace one in 11 workers. Although companies may wait for recessions, Sessions, there was the next recession to lay people off. So everybody should be aware of that in case that's coming. coming. Another study showed that AI-related
9:18
Unidentified speaker (speaker_1): layoffs have quadrupled since 2024 through late 2026. In fact, 60 plus CEOs have announced that layoffs due to AI achieved efficiencies. Again, take it all with a grain of salt, but the trending line in the studies and the postings and
9:38
Unidentified speaker (speaker_1): so forth. are AI is coming in, and in some cases, people are stepping away.
9:45
Unidentified speaker (UNKNOWN): So
9:45
Unidentified speaker (speaker_1): the stories go on. So, however, you feel about the veracity and ethical nature of these and other AI stats relating to citing human worker replacement, it raises an important question for leadership. If we take these statistics at face value, then we have to ask what role is AI playing when it does replace human resources with an
10:05
Unidentified speaker (speaker_1): automated equivalent? And we have to figure out how to plan today about that and get around that. And I'm going to talk about that today. That's what the focal point of the podcast is. Does leadership understand where and why these layoffs are or should be occurring? Is there a grand plan behind using a formal framework
10:25
Unidentified speaker (speaker_1): for understanding how to maximize? the value of an AI agent across multiple business scenarios? Or are we headed down the same path organizations pursued with business processes and other technologies? One of inconsistency, redundancy, fragmentation, and essentially ineffective governance. And if
10:44
Unidentified speaker (speaker_1): leadership pursues the same meandering path with AI, will they prolifer be proliferating AI agent replication?
10:52
Unidentified speaker (speaker_1): potentially opening the door to internal AI agent battles for position and supremacy? It's not that far-fetched when you hear about what's going on in the news. So if you're creating these agents, you have to know what you're creating. You have to document them, they have to be under your control. And you certainly shouldn't be creating AI wars
11:11
Unidentified speaker (speaker_1): internally in your own organization while they claim dominance.
11:16
Unidentified speaker (UNKNOWN): So
11:16
Unidentified speaker (speaker_1): you've got to put the safeguards in, and we'll talk about a lot of that today.
11:20
Unidentified speaker (UNKNOWN): So
11:20
Unidentified speaker (speaker_1): maybe leadership needs to consider being more systemic in their approach by leveraging a formally architected uh approach to a framework, basically, for laying these things out, organizing them with a business-driven context, not a technology-driven context. And we say that all the time.
11:40
Unidentified speaker (speaker_1): But again, people will hand this off to IT. In fact, one of the pieces I was reading, just as an aside when I was doing research for this podcast, was IT needs to manage the agents. That is wrong, right? The agents need to be managed by the business as a whole. Now you can figure out who's gonna do that. They have an agent manager and so forth. But you
12:00
Unidentified speaker (speaker_1): don't let IT manage your human resource points. pool. I don't think they should be managing your agent pool. IT is the implementation arm. They're not the ones in charge. And you have to put that in context. And I think that's really important.
12:15
Unidentified speaker (UNKNOWN): So
12:15
Unidentified speaker (speaker_1): if leadership and management teams continue to randomly lay off people with little or no understanding of the implications,
12:22
Unidentified speaker (UNKNOWN): then
12:23
Unidentified speaker (speaker_1): you'll end up in the same situation I discussed in my October 14th podcast.
12:27
Unidentified speaker (speaker_1): Organization Design in the Age of AI, where essentially workers, specifically the ones can play a key role in helping leadership navigate the very transformation challenges that many organizations are facing, will be let go, summarily dismissed, even as the individuals who are proliferating the issues that leadership is seeking to overcome
12:47
Unidentified speaker (speaker_1): remain entrenched. So are you cutting the wrong?
12:51
Unidentified speaker (speaker_1): people?
12:52
Unidentified speaker (UNKNOWN): So,
12:52
Unidentified speaker (speaker_1): who falls into this latter category of entrenched players that maybe are making things worse, not better? Well, let's start with the proliferators of redundant, convoluted business processes. How many people have business process teams that are listening here? And I went through all this on my business
13:12
Unidentified speaker (speaker_1): process discussion podcast a few weeks ago.
13:15
Unidentified speaker (UNKNOWN): But
13:16
Unidentified speaker (speaker_1): how many people have business process teams that are growing larger, spending more money, delivering more processes that are essentially doing the same thing that have already been modeled and already been automated? Only they're doing it in redundant, replicated ways. So is that making any sense, especially as we're trying to
13:35
Unidentified speaker (speaker_1): move into a much more sophisticated architecture and deployment? approach?
13:40
Unidentified speaker (UNKNOWN): It
13:40
Unidentified speaker (speaker_1): also includes individuals seeking to expand versus curtail the expansive use of spreadsheets and shadow systems. Shadow systems are the ones beyond the formal architecture, consider that IT, their formal line of sight. Spreadsheets are rampant in most organizations. They can be running
14:00
Unidentified speaker (speaker_1): in even a small to mid-sized organization. they could be in the tens of thousands. In larger organizations, there's every evidence to believe they're in the hundreds of thousands. That is terrifying because you have spreadsheets running your company. And if any one of those break in the daisy chain, or if you change certain
14:20
Unidentified speaker (speaker_1): data coming in, you're going gonna to run into problems.
14:23
Unidentified speaker (UNKNOWN): So
14:23
Unidentified speaker (speaker_1): instead of laying off the very workers that can help you navigate organizations. transformation, maybe leadership should consider a phase drawdown strategy for the hundreds of overpaid, underage contractors that your system's integrator is using to hold you hostage to the tens of thousands of spreadsheets they maintain.
14:39
Unidentified speaker (UNKNOWN): If
14:39
Unidentified speaker (speaker_1): you're not sure what's going on in the trenches, take a quiet tour of your operations center and then take a close look at where the agent where the money's being spent, especially where integrators are involved. The concept is of course a striking alternative to leadership asking the few people who are left in the organization who actually know what's going on to step down
14:59
Unidentified speaker (speaker_1): and step away. Who are those individuals is the question, right? Those are your formal architects. Now, you may not like the way they deliver work or the work that they've delivered, and that's fine, you should fix
15:11
Unidentified speaker (UNKNOWN): that.
15:12
Unidentified speaker (speaker_1): But certainly you should have a business architecture. team, you should have a data architecture team, you should have solution architects with skills that are hooked into the other architecture teams. And I always point to enterprise architecture as sort of the umbrella for that. There's other elements of architecture, technology architecture, but if you look at a lot of enterprise architecture teams
15:32
Unidentified speaker (speaker_1): are spending a lot of time on technology. But take a close look at who these people are and what they're doing.
15:39
Unidentified speaker (UNKNOWN): And
15:39
Unidentified speaker (speaker_1): I was just talking to somebody a few weeks ago that said their business architecture team and their data architecture teams were broken up and disbanded with about two-thirds of the players laid off and the rest of them scattered to the four winds. So I guess they couldn't make any more trouble. You need those architects. You need the architecture. And
15:59
Unidentified speaker (speaker_1): if you don't, then you're not listening to what I'm saying on this particular podcast today. So we're going to explore the idea of business stakeholders as AI agents as we go through the podcast today. We're going to have probably be looking at hybrid situations where the same stakeholder will have both uh
16:18
Unidentified speaker (speaker_1): automation in terms of an AI agent as well as a human, maybe backing them up or maybe playing the same role in a different context. So we will look at ecosystem-wide stakeholder assessments and look at AI agent deployment on an ecosystem basis. Remember, I said we don't want AI wars. We don't want agent wars
16:38
Unidentified speaker (speaker_1): where they're teaming up and fighting against each other because they think that they're supposed to be playing the same role. You have to think about that as you're starting to deploy these things and put some controls and bombings around them. We'll get into more on value streams, more on scenarios, uh, more on stakeholder roles. And again, we're going to dig into uh a little
16:58
Unidentified speaker (speaker_1): more on all of this as we go forward, and then we're going to look at what some of the recommendations are from an industry perspective and how we can align with some of those recommendations internally. And again, I just want to throw in the word architecturally and formally. But before we do that, we're going to take a quick break. You're listening to the Strategy Execution podcast
17:18
Unidentified speaker (speaker_1): with William Ulrich, co-author of End-to-End Strategy Execution from Inception through Solution Deployment. You can get our book on Amazon.com or other outlets wherever you get your books. And we'll be right back after a short break.
17:41
Unidentified speaker (UNKNOWN): Do
17:41
Unidentified speaker (speaker_0): your strategic objectives fail to deploy? Wondering why technology budgets keep growing but deliver less? Are your strategy deployment efforts splintering into dozens of disjointed, disconnected projects? Does your organization lack a clear vision for maximizing AI and related technology investments long-term? long term? If this sounds
18:01
Unidentified speaker (speaker_0): familiar, order a copy of William Ulrich and Charles Bowman's new book, End-to-End Strategy Execution from Inception Through Solution Deployment, and get the answers you're looking for today.
18:14
Unidentified speaker (UNKNOWN): When
18:14
Unidentified speaker (speaker_2): it comes to business, you'll find the experts here. Voice America Business Network.
18:25
Unidentified speaker (UNKNOWN): We're
18:26
Unidentified speaker (speaker_0): back on the Voice America Business Channel, where William Ulrich is breaking down the hidden barriers that keep organizations stuck in repeated failures. Let's continue.
18:37
Unidentified speaker (speaker_1): Welcome back to the Strategy Execution Podcast. I'm your host, William Ulrich. You can check out our podcast at voiceamericamedia.com or subscribe to it on Apple Podcasts, Spotify, or wherever you get your podcasts.
18:50
Unidentified speaker (UNKNOWN): So
18:50
Unidentified speaker (speaker_1): let me uh bring us back to where we left off at the break. I want to start by
18:55
Unidentified speaker (UNKNOWN): in
18:56
Unidentified speaker (speaker_1): this next segment systematically breaking down which stakeholders are playing key roles across the universe of value streams that contribute to value seeking customers
19:05
Unidentified speaker (UNKNOWN): and
19:05
Unidentified speaker (speaker_1): expand that analysis to value seeking partners and in-house stakeholders Folders because we do have to accommodate them. So let me just from a value stream perspective let me just take a drop drop back a second and take a look at what we might be running into in a real real world situation. Okay.
19:24
Unidentified speaker (UNKNOWN): So
19:24
Unidentified speaker (speaker_1): as I as I've noticed in past podcasts,
19:28
Unidentified speaker (speaker_1): a given company, based on the complexity and multifaceted nature of its business model, might have thirty 30 to thirty-five 35 total value streams, rarely exceeding 40. And again, it's the complexity of the business model, not the size of the company. A gigantic insurance company that has no financial
19:48
Unidentified speaker (speaker_1): elements to it, financial management elements, maybe just has one or two kinds of insurance, might be a very simple business model and not have a whole lot of value streams, but they may be very large.
20:01
Unidentified speaker (UNKNOWN): The
20:01
Unidentified speaker (speaker_1): flip side of it is a mid-the-mid-the-large company that has both retail and manufacturing and maybe transportation all blended in there will have a larger set of a larger business architecture in terms of value streams, capabilities, and other things.
20:19
Unidentified speaker (UNKNOWN): So
20:19
Unidentified speaker (speaker_1): again, not the size of the company, but the complexity of the business model.
20:23
Unidentified speaker (UNKNOWN): So
20:24
Unidentified speaker (speaker_1): value streams provide a targeted environment for examining the nature and role of stakeholders acting across the business ecosystem. The same stakeholders that can potentially become AI agents. They become our candidates. Value streams represent the entire ecosystem-wide perspective of work that's going on across the organization for
20:43
Unidentified speaker (speaker_1): value delivery, end-to-end To end value delivery. It's bigger than the legal bounds of the organization because it incorporates customers. customer and partner interactions and roles,
20:54
Unidentified speaker (UNKNOWN): but
20:54
Unidentified speaker (speaker_1): it also includes all the internal human resource-triggered value streams as well.
20:59
Unidentified speaker (UNKNOWN): So
20:59
Unidentified speaker (speaker_1): say this in another way: everything happening within and across an organization should ideally be represented by its value streams. That means that instances of internally and externally triggered value streams are in constant motion. This may include a multitude of customer. customers concurrently moving through instances of product acquisitions, service requests,
21:19
Unidentified speaker (speaker_1): or similar value streams.
21:21
Unidentified speaker (UNKNOWN): The
21:21
Unidentified speaker (speaker_1): numbers scale up based on your ability to support them from both a human and technological ability. Obviously, with we're hoping at least with AI, we can improve that and process more and more quickly. That's what people are looking for, and they're they were already moving on some of that.
21:39
Unidentified speaker (UNKNOWN): A
21:39
Unidentified speaker (speaker_1): value stream is conceptual, it's not an implemented thing, although I did talk about dynamic value stream realization and how we can look at that as a more automated thing. And that's in the last podcast I did prior to this one. Either way, if it can't scale to the necessary degree that an organization requires, that will trigger capability investments
21:59
Unidentified speaker (speaker_1): in people and technology, with work management design being an element of.
22:03
Unidentified speaker (UNKNOWN): that.
22:03
Unidentified speaker (speaker_1): One example for those who doubt that the entirety of work across an ecosystem is involved in some type of value stream activity, should just take a look at a not very well-explored value stream that most organizations have, in fact, all do, called Execute Operation.
22:19
Unidentified speaker (UNKNOWN): It's
22:19
Unidentified speaker (speaker_1): a multi-purpose value stream. It has many instances running in the background in at any given organization.
22:25
Unidentified speaker (UNKNOWN): The
22:25
Unidentified speaker (speaker_1): value stream is more pronounced from many. manufacturing companies who would be running, for example, a shift at a factory or maybe a transportation center that's running on 12-hour shifts. But the operation concept applies to every organization. Even an insurance company has work cycles. So they would have an operation or multiple operations
22:45
Unidentified speaker (speaker_1): running in the background, right?
22:47
Unidentified speaker (UNKNOWN): So
22:47
Unidentified speaker (speaker_1): it tends to run nonstop over periods of time, like data centers. There's other other kinds of examples out there. The operation is a formal thing. It's a business object. It is recognized by cap and processed and managed by capabilities. It is something that is stored as data in your data architecture.
23:05
Unidentified speaker (UNKNOWN): And
23:05
Unidentified speaker (speaker_1): it connects to other things like workers and equipment, assets, right? So operation is a very formal thing. Execute operations. operation is getting it going and turning it off, right? And ending it. And we have lots and lots of those going on.
23:21
Unidentified speaker (UNKNOWN): So
23:22
Unidentified speaker (speaker_1): in general, the work being performed across an organization is engaged in pursuit of one or another value propositions that include all the customer triggered as well as all the internally triggered and partner-triggered value streams.
23:34
Unidentified speaker (UNKNOWN): That
23:34
Unidentified speaker (speaker_1): can include formalizing an agreement, disseminating information, delivering an initiative, running a campaign, or engaging in any number of other efforts. And just keeping the lights on, that falls back to our execute operation value stream.
23:47
Unidentified speaker (UNKNOWN): So
23:47
Unidentified speaker (speaker_1): with this perspective as backdrop, consider that each value stream stage has work underway that is performed by any number of stakeholders. Stakeholders that are in general are assumed to be human resources. Customers and partners are largely represented as humans. as well, with their own personas, even though we tend to generally think of them
24:07
Unidentified speaker (speaker_1): in lots of contexts as companies, right?
24:10
Unidentified speaker (UNKNOWN): So
24:10
Unidentified speaker (speaker_1): we do accommodate the concept of an organization being a stakeholder, but we generally lean towards humans.
24:18
Unidentified speaker (UNKNOWN): The
24:18
Unidentified speaker (speaker_1): working assumption for planning purposes Uh is that an organization's human resources may be defined as AI agents at some point. Customers, at least in the immediate future, will in all likelihood continue to be humans. Of course, if we automate ourselves out of existence, the future may just be AI agents working to help our other
24:38
Unidentified speaker (speaker_1): agents.
24:40
Unidentified speaker (speaker_1): We're not quite entering that uh apocalyptic terminator
24:44
Unidentified speaker (UNKNOWN): one,
24:45
Unidentified speaker (speaker_1): two, three sort of perspective at this point in time, so I'll just put that aside. So we continue to say. assume that uh customers are real people, and again, customers can be constituents and patients and everything else that you have out there.
24:58
Unidentified speaker (UNKNOWN): So
24:59
Unidentified speaker (speaker_1): a typical example of stakeholders at a given organization include things like asset maintainers, controllers, data analysts, content developers, financial analysts, human resource analysts, product managers, program managers, administrative assistants, human resources. resource specialists, and more.
25:16
Unidentified speaker (UNKNOWN): In
25:17
Unidentified speaker (speaker_1): a well-architected business architecture, these stakeholders are defined in the stakeholder map, with each stakeholder associated with one or more value stream stages.
25:27
Unidentified speaker (UNKNOWN): In a
25:28
Unidentified speaker (speaker_1): business architecture, each stakeholder has a description which represents their persona or the role they play in a situational context.
25:35
Unidentified speaker (UNKNOWN): These
25:35
Unidentified speaker (speaker_1): aren't job titles, although they may overlap and certainly they can. Job titles are not always a good way to capture the context of what a person does on a day-to-day basis. They may be playing one role in one value stream or another role in another value stream.
25:51
Unidentified speaker (UNKNOWN): But
25:52
Unidentified speaker (speaker_1): let's take a look at some descriptions. A risk manager might be described as an individual responsible for identifying, evaluating, prioritizing, and mitigating potential financial, operational, strategic. or regulatory risks and vulnerabilities that might impact an organization's objectives, assets, or balance sheet in negative ways. It's pretty broad. Maybe you want to break
26:12
Unidentified speaker (speaker_1): that down. We'll talk more about that.
26:14
Unidentified speaker (UNKNOWN): An
26:14
Unidentified speaker (speaker_1): accountant, on the other hand, would describe an individual who records, tracks, analyzes, and reports on the finances of an organization. If an organization has multiple kinds of accountants, then
26:26
Unidentified speaker (UNKNOWN): we
26:27
Unidentified speaker (speaker_1): can be more granular in our persona descriptions. A cost accountant, for example, would differ from a forensic accountant. So we may have multiple types of accountants that play specific roles
26:39
Unidentified speaker (speaker_1): as we get into our value streams and our stakeholder work that goes on
26:44
Unidentified speaker (UNKNOWN): there. An
26:45
Unidentified speaker (speaker_1): agentic AI team might choose to decompose a subset of these stakeholders even further.
26:50
Unidentified speaker (UNKNOWN): And
26:50
Unidentified speaker (speaker_1): we'll talk a little bit more about decomposition and putting boundaries and putting constraints on the work that an AI agent might be doing in your organization. And narrowing the role that they have is one recommendation that comes from industry.
27:06
Unidentified speaker (UNKNOWN): The
27:07
Unidentified speaker (speaker_1): inclusion description of these stakeholders and their association with the value stream stages in which they participate are important in context of identifying. and formalizing the roles for our synthetic stakeholders, who are essentially our AI agents. They also form the basis for exploring and formalizing dynamic rules based routing maps, which I'll get to in a little bit. And you've
27:27
Unidentified speaker (speaker_1): heard me talk about those before, but I'm going to bring them back today because they're very relevant to what I'm talking about.
27:33
Unidentified speaker (UNKNOWN): Now,
27:33
Unidentified speaker (speaker_1): The good news is that most business architecture industry reference models, at least those delivered from the business architecture.
27:39
Unidentified speaker (speaker_1): Guild, the Industry Association that has reference models and other content, they provide a baseline for stakeholder descriptions and corresponding value stream
27:50
Unidentified speaker (speaker_1): staged stakeholder cross-mappings. We call them cross-mappings when we connect things in business architecture. Reference models provide organizations with a starting point that they can then customize and expand upon to fully represent their business ecosystem. An example, a manufacturing reference model from the Business Architecture Guild has lots of stakeholders that include
28:09
Unidentified speaker (speaker_1): manufacturing types of stakeholders, inventory managers, part managers, or material managers, and so
28:16
Unidentified speaker (UNKNOWN): on.
28:17
Unidentified speaker (UNKNOWN): So
28:17
Unidentified speaker (speaker_1): prior to getting to the AI agent discussion, there are two more connections to stakeholder that even advanced business architects and mature business architectures often fail to map out.
28:27
Unidentified speaker (UNKNOWN): And
28:27
Unidentified speaker (speaker_1): I want to mention these because they're particularly important as we're getting more rigorous and more formal in our architectural pursuits.
28:36
Unidentified speaker (UNKNOWN): The
28:37
Unidentified speaker (speaker_1): first one is the stakeholder-to-business unit mapping. And that's a natural mapping from uh from the stakeholder over to the business unit, uh, the knowledge base uh that uses or leverages the guild's meta model.
28:51
Unidentified speaker (speaker_1): will be able to support that.
28:53
Unidentified speaker (speaker_1): So
28:55
Unidentified speaker (speaker_1): the idea here is that a stakeholder with certain responsibilities isn't isolated to one or another area, but more likely to appear across multiple business units.
29:05
Unidentified speaker (UNKNOWN): So
29:05
Unidentified speaker (speaker_1): an eco-wide view of stakeholder establishes potential AI agent reuse, which should be explored as part of an agent planning. So, does an organization really require, for example, two dozen agentic cost accountants that have identical descriptions, responsibilities, and constraints? Probably not. So
29:25
Unidentified speaker (speaker_1): if we're going to define a cost accountant and what it does, we would like to make that stakeholder available globally from an enterprise perspective.
29:37
Unidentified speaker (UNKNOWN): Now,
29:37
Unidentified speaker (speaker_1): the good news is that if a business architecture is not constrained to one or another business units. It's across the ecosystem. That's the good news. And if we pursue that approach and we define the stakeholder map and we define the breakdown of the organization map, and we associate
29:57
Unidentified speaker (speaker_1): these particular stakeholders with one to many business units. So for example, a cost of accountant might be working over in this one area with manufacturing, they might work in another area for sales and so forth. Then we have more of a global view from an ecosystem perspective of where stakeholders play a role.
30:17
Unidentified speaker (UNKNOWN): When
30:17
Unidentified speaker (speaker_1): I say this is somewhat topic, it is, and there are some organizations that already are advanced enough to be able to pursue this type. of mapping. There are also many, many organizations that are stuck in an endless merry-go-round arguing whether or not their capability map is principled or not, and not really thinking that they need value
30:37
Unidentified speaker (speaker_1): streams and so forth. So for those organizations, a quick aside, you need to get going and get moving and uh you know stop debating all this stuff internally and go and pick up a reference model and start start doing some work.
30:50
Unidentified speaker (UNKNOWN): The
30:51
Unidentified speaker (speaker_1): second stakeholder connection is even more nuanced. When a stakeholder is participating in a value stream stage, that stakeholder is engaged in delivering certain capability outcomes, which in turn deliver value items within the value stream stage. The stakeholder capability association, the relationship there,
31:11
Unidentified speaker (speaker_1): only works within specific confines of a value stream stage. because capability behaviors that lead to a capability outcome may differ based on the situation. An example I like to use is a marketing value stream, for example, may be seeking customer preferences based on a history of content impressions, while an
31:30
Unidentified speaker (speaker_1): agreement-oriented value stream might involve a security team seeking to determine data sharing rights.
31:39
Unidentified speaker (UNKNOWN): In
31:39
Unidentified speaker (speaker_1): both situations, or both cases, the capability is customer preference management, or
31:44
Unidentified speaker (UNKNOWN): it
31:44
Unidentified speaker (speaker_1): could be agreement preference management. So assuming a customer has a set of preferences and then they've got six agreements,
31:52
Unidentified speaker (UNKNOWN): the
31:52
Unidentified speaker (speaker_1): agreement preferences may be unique to each agreement, but the customer preference may override it. Those are business rules you have to establish. Stakeholder to capability associations require exploration if the human stakeholder is replaced by an AI agent, where the agent blends with the capability automation or maybe even becomes that degree of automation. So,
32:12
Unidentified speaker (speaker_1): what outcomes will an AI agent deliver? And what's their relationship with a capability? So, starting to look at these mappings Mapping sooner versus later is a very good idea. And I haven't seen a whole lot of business. architectures out there that are looking at either of the business unit or the
32:32
Unidentified speaker (speaker_1): stakeholder to capability mapping at the level I'm describing. The last architecture and design topic, as we lead into the agentic AI discussion, involves dynamic rules based routing, which is a design discipline that is constrained by value stream stage boundaries. Once in a stage, the routing map serves as the event and rule rule-driven driven. state machine, with work
32:52
Unidentified speaker (speaker_1): tasks being assigned to various stakeholders within that stage. Being nonlinear, there may be several work tasks or what we call them work items in business architecture in play across multiple stakeholder work queues at the same time.
33:06
Unidentified speaker (UNKNOWN): And
33:06
Unidentified speaker (speaker_1): that's not a problem because dynamic rules-based routing maps are nonlinear. They can have multiple things happening, and that's all. normal. Once a combination of states is satisfied, the work may move out of this the stage and on to the next stage. But while the work is underway, everything is constrained by the state restrictions
33:26
Unidentified speaker (speaker_1): that allowed you to enter the stage and force you to remain in the stage until such time that another set of states are reached.
33:33
Unidentified speaker (UNKNOWN): Right?
33:34
Unidentified speaker (speaker_1): So for a quick example, on a take-a-trip value stream where a passenger is taking a trip. the passenger remains in the arrive at destination stage until the trip segment is completed and the plane lands, or they get off the boat, or they get out of their out of the Uber car and so forth. Even as the trip itself remains active because it's a
33:53
Unidentified speaker (speaker_1): multi-leg trip, and they're just going to go and wait for their next flight. So they'll go back to the depart stage. That's all state-based.
34:01
Unidentified speaker (UNKNOWN): These are
34:02
Unidentified speaker (speaker_1): two states. that are checked and rechecked to allow for the navigation within and across the take-a-trip value stream.
34:08
Unidentified speaker (UNKNOWN): So
34:08
Unidentified speaker (speaker_1): you get a little bit of taste on how some of this works with a real value stream in the in the sort of the real world. And again, you want to see how this works in business architecture, go to the business architecture guild. Uh you can join as an individual, or maybe you have a corporate membership. Either way, works, and you can download the reference model for use in your organization. And these take-a-trip value
34:28
Unidentified speaker (speaker_1): streams that would be inside the it's in multiple reference models, but certainly the transportation model.
34:34
Unidentified speaker (UNKNOWN): So,
34:35
Unidentified speaker (speaker_1): what I want to do is uh move on to the next discussion, which is um AI agents and having them take the place of our stakeholders that we've been talking about today. But before I do that, I wanna want to go and take our second break.
34:50
Unidentified speaker (UNKNOWN): You're
34:51
Unidentified speaker (speaker_1): listening to the Strategy Execution Podcast with William Ulrich, co-author of N10 Strategy Execution from Inception through Solution Deployment.
34:58
Unidentified speaker (UNKNOWN): You
34:59
Unidentified speaker (speaker_1): can get our book on Amazon.com or other outlets. We'll be back right after a short break.
35:12
Unidentified speaker (UNKNOWN): The
35:12
Unidentified speaker (speaker_0): groundbreaking new book, In-to-In In-to-Ind Strategy. Execution, from Inception Through Solution Deployment, tackles initiative failures and wasted investments head-on. Co-authors William Ulrich and Charles Bowman take readers on a guided journey that not only helps leadership formulate a strategic vision, but lays the foundation
35:32
Unidentified speaker (speaker_0): for making that vision a reality. The book uses practical examples, analogies, and case studies. to chart a path forward for C-suites, implementation teams, and everyone in between. Get a copy today and join them on the journey.
35:48
Unidentified speaker (speaker_3): Voice America is on LinkedIn.
35:51
Unidentified speaker (speaker_0): Connect with us today.
35:58
Unidentified speaker (speaker_0): We're back on the Voice America Business Channel, where William Ulrich is breaking down the hidden barriers that keep organizations stuck in repeated failures. Let's continue.
36:10
Unidentified speaker (speaker_1): Welcome back to the Strategy Execution Podcast. I'm William Ulrich, co-author of End-to-End Strategy Execution from Inception through Solution Deployment. You can check out this podcast at VoiceAmerica Media.com or subscribe to it on Apple Podcasts, Spotify, or wherever you get your podcasts.
36:26
Unidentified speaker (UNKNOWN): So
36:26
Unidentified speaker (speaker_1): I took a long lead in discussion on signing AI agents to play the role of stakeholders in your business ecosystem, specifically one bound by architecture design constraints that protect the organization and its partners and customers while delivering desired results. So I wanted to provide that framework of thinking as
36:46
Unidentified speaker (speaker_1): we went into the next discussion that we're starting now, because it's very relevant. And as you'll see by some of the folks who talk about agentic deployment in your organization, that you'll see some of that. But I did talk about some of the baseline. One is setting up an environment where you can start to manage your agents. I'll
37:06
Unidentified speaker (speaker_1): talk more about that as we wrap up today. But let me take a look at some of the literature around AI agents and the roles that they can play.
37:15
Unidentified speaker (UNKNOWN): So
37:16
Unidentified speaker (speaker_1): in a piece published on IBM's website, it was titled Flexible by Design, Reliable by Proof: Building and Evaluating AI agents that work in the real world. Alex Draley wrote that the future of automation is a deliberate balance of agentic and deterministic approaches, designed for adaptability and governed for
37:35
Unidentified speaker (speaker_1): trust and evaluated by proof.
37:38
Unidentified speaker (UNKNOWN): The
37:38
Unidentified speaker (speaker_1): piece added a key point. Deterministic systems have limits. They are brittle by design. A workflow that receives unexpected results, and this is my perspective,
37:49
Unidentified speaker (UNKNOWN): right?
37:50
Unidentified speaker (speaker_1): Can stall. If a condition isn't explicitly defined, actually this is a quote from from uh from Alex Traley. If a condition isn't exactly explicitly defined, the system simply cannot proceed. Deterministic automation works best in structured, stable environments. It struggles with ambiguity, edge cases, and change.
38:10
Unidentified speaker (speaker_1): And if if you're not clear on exactly why I said that, I'm going to get into why business processes are not the ideal here. You can always go back to my podcast that discussed that particular topic.
38:22
Unidentified speaker (UNKNOWN): So
38:23
Unidentified speaker (speaker_1): at these points, I want to, the podcast was the October 21st. podcast episode where I expressed the bear's shortcomings of flow-based business process models. Specifically, they are too rigid, too deterministic, and unable to function in the real world because they rely on a mythical happy path and an endlessly spiraling tangleweb of exception
38:42
Unidentified speaker (speaker_1): processing that never covers every exception.
38:45
Unidentified speaker (UNKNOWN): The
38:45
Unidentified speaker (speaker_1): piece by Straley Stray reinforces the point, the very point I was making in that podcast. that business process models are not the viable foundation for AI deployment. Yet AI literature, systems integrators, and the press,
39:00
Unidentified speaker (speaker_1): and the all-knowing analysts that are out there seem to continue to insist that people should have their business processes defined before they get into pursuing agentic AI, which, in my view, the opposite is true.
39:13
Unidentified speaker (UNKNOWN): So
39:14
Unidentified speaker (speaker_1): you can listen. or re-listen to my October 21st podcast or just apply common sense to this. Business processes are not viable, not a viable foundation for AI deployment because their very essence is the antithesis of what Straley warns against.
39:29
Unidentified speaker (UNKNOWN): In
39:29
Unidentified speaker (speaker_1): that same October podcast episode, by the way, on business processes, I suggested an alternative called dynamic rules-based routing, which is an event-driven, stable state-based approach framed within the context of business architecture value streams, what I've been talking about here all along. Dynamic rules-based routing maps address the
39:49
Unidentified speaker (speaker_1): limitations of flow-based BPM by handling, leveraging business object states, and corresponding business rules to deploy nonlinear, collaborative work management environments that address
40:01
Unidentified speaker (UNKNOWN): the
40:01
Unidentified speaker (speaker_1): overall requirements of one. what needs to happen in the context of that particular piece of work. Straley lays out in the piece I cited about on IBM's website some clarity around all this, right? So which is great because I'm glad to see that at least some of the literature out there is starting to push people the other direction.
40:21
Unidentified speaker (speaker_1): Now, the Australia Straley piece didn't say come out directly and say don't use business processes,
40:26
Unidentified speaker (UNKNOWN): but I
40:26
Unidentified speaker (speaker_1): think it was very, very clear what was being said when you read between the lines. Dynamic business rules-based routing maps accommodate everything as normal. Again, you've heard me say this if you listen to my other podcasts. There are no happy paths and no exceptions. Things just happen.
40:42
Unidentified speaker (UNKNOWN): You
40:42
Unidentified speaker (speaker_1): have rules engines that provide the navigation based on a finite number of states, a very finite number of states if you're within a given value stream stage. They can only exist in that context. That puts a lot of boundaries on things. And the boundaries are what we're looking for when we're trying to put limits on agents.
40:59
Unidentified speaker (UNKNOWN): So,
41:00
Unidentified speaker (speaker_1): what do we mean here by putting limits? Dynamic rules-based routing maps provide the constraints on the actions taken by a particular stakeholder, human or otherwise, in context of a given value stream stage. And why is this relevant? Because according to an article titled Agencies Urge AI Agent Access Limits and Weigh Human
41:19
Unidentified speaker (speaker_1): Oversight by Gail Emery on Meritalk.com, AI agents should receive minimal privileges rather than automatically inheriting the access rights of the employees using them. In my view, this is where automation of risk and security-related security related capabilities associated with sentient business objects comes come into play. These capabilities are
41:39
Unidentified speaker (speaker_1): inherent in most business architecture reference models and should be a major automation focal point for organizations deploying AI solutions.
41:46
Unidentified speaker (UNKNOWN): But
41:47
Unidentified speaker (speaker_1): another recommendation from Gail Emery's piece is that workflows
41:51
Unidentified speaker (speaker_1): involving multiple agents divide tasks among multiple AI agents where For each agent only has the privileges needed for its specific actions. The stakeholder assignments in a dynamic rules-based routing map align very well to this concept. They're very specific to a given task
42:11
Unidentified speaker (speaker_1): or work item that's being assigned to a queue that's assigned to a stakeholder.
42:17
Unidentified speaker (speaker_1): And it's not a full process that's being modeled. It's essentially a series of work tasks that happen in some sequence that allow you to transition. through the states needed to that you have to transition to. This separates the privileges across the routing map. So certain tasks are assigned to certain stakeholders.
42:37
Unidentified speaker (UNKNOWN): Again,
42:37
Unidentified speaker (speaker_1): tasks are called work items in business architecture, and they're assigned to work queues.
42:42
Unidentified speaker (UNKNOWN): And
42:42
Unidentified speaker (speaker_1): those are assigned to stakeholders with limited authority.
42:45
Unidentified speaker (UNKNOWN): So
42:45
Unidentified speaker (speaker_1): a strategy execution book, if you're looking for more details on this, you can go to chapter six and then you can also look at chapter nine. But chapter six talks about it under work management design. Decomposing stakeholder personas in the business architecture stakeholder map to the point where they meet these constraints provides another basis for addressing these limits that we're
43:05
Unidentified speaker (speaker_1): talking about putting in here. Applying the dynamic rules-based routing map design, which is controlled by states, through to the implementation of a hierarchical state machine, allows architecture. architects and designers to implement these constraints. Again, you can take a look at our book and the chapters I was referencing. So finally,
43:23
Unidentified speaker (UNKNOWN): I
43:24
Unidentified speaker (speaker_1): want to discuss the idea of managing AI agents, which is a whole nother topic that I want to make sure I cover today. Because it's not something a lot of folks think about or talk about, but it is certainly something that's that's beginning to be written about. Agents need to be documented.
43:41
Unidentified speaker (UNKNOWN): And
43:41
Unidentified speaker (speaker_1): again, business architecture helps with that because we have a stakeholder map, right? So if we're talking about what some folks are are at least sharing out there,
43:51
Unidentified speaker (UNKNOWN): um,
43:52
Unidentified speaker (speaker_1): in a piece called Leaders Will Manage AI Agents Before They Know How, Kate Kellogg, professor of work and organization studies at MIT Sloan, calls managing AI agents. agent supervision ambiguity. She said that with a genetic AI, a manager is acting as a supervisor of agents. And at this point in time, people don't
44:12
Unidentified speaker (speaker_1): know how to do this. Plain and simple. It's a new thing. We don't even have that role probably defined in our organization, although we should start thinking about it.
44:20
Unidentified speaker (UNKNOWN): A
44:20
Unidentified speaker (speaker_1): career spent Spend supervising people comes with no instructions on how to supervise agents, according to Kellogg. Essentially, saying that there are no instructions for managers.
44:29
Unidentified speaker (UNKNOWN): So,
44:30
Unidentified speaker (speaker_1): what can a leader do? Well, Professor Kellogg offers a couple of answers. First, encourage people to have a growth mindset, right? So we can't box ourselves into the way we've been doing things in HR for the last 30 years.
44:43
Unidentified speaker (UNKNOWN): The
44:43
Unidentified speaker (speaker_1): second is about volume. People need to be running a lot of agents. That's a gimme, right? That's a gimme.
44:49
Unidentified speaker (UNKNOWN): And
44:49
Unidentified speaker (speaker_1): think about how those agents work in parallel. Well, I just talked about agents working in parallel in value streams under the context or design of a dynamic rules-based routing map. We're not constraining them, we're allowing them to work in parallel as states are set and reset and reset
45:08
Unidentified speaker (UNKNOWN): as
45:08
Unidentified speaker (speaker_1): things are moving forward.
45:10
Unidentified speaker (UNKNOWN): So
45:10
Unidentified speaker (speaker_1): I want to point out that for formal architecture, as organizations pursue the use of agentic. AI, you're going to start to document this, and then the dynamic rules-based routing map design is a spin-off from the business architecture itself. When all agents are documented formally in the business architecture, which should be housed in a business knowledge base, a formal knowledge
45:30
Unidentified speaker (speaker_1): base, those stakeholders or agents, as they will become, are mapped to the value stream stages where they play a role.
45:37
Unidentified speaker (UNKNOWN): So
45:37
Unidentified speaker (speaker_1): there's greater clarity on that role definition that they're going to play. Operationally, then, deployment of detailed dynamic rules-based routing maps needs to be built into an organization's data architecture. This may not be apparent. So when you're starting to build out your data architecture based on your business architecture, the kinds of things I'm
45:57
Unidentified speaker (speaker_1): talking about need to be in your data architecture. And where would agents fall? in terms of the information being managed within the context of one, your information map and business architecture, and two, your data architecture? Well, in the information map, each stakeholder rolls up to a short list of business
46:17
Unidentified speaker (speaker_1): objects. One is a human resource, another is the customer, and a third one is a partner. Those categories, those three categories, and again, they can vary by the company and business you're in and so forth. But those three categories are aligned to the business objects that are being manipulated by the capabilities and that are
46:37
Unidentified speaker (speaker_1): represented in the information map. So when we say where's that information managed, they're managed at the object level through information concepts and the capabilities that work against them. So it's it's a closed loop, it's it aligns, it is rigorous, and it works well. So you actually have an architecture that can actually answer a lot of
46:57
Unidentified speaker (speaker_1): these questions in advance if you bother to make the effort to invest in these advanced concepts that I'm talking about today, and stop arguing about whether or not something should or shouldn't be in your capability map.
47:09
Unidentified speaker (UNKNOWN): So
47:10
Unidentified speaker (speaker_1): again, AI agents need to be documented, constrained and by roles, bound by formal. constraints, and engaged in formal work management designs that are not constrained, but allow for the latitude for agent decision making where the AI agent is a proxy for today's human knowledge worker. Remember, knowledge workers are people who know more about their role or their job
47:30
Unidentified speaker (speaker_1): and uh don't need to be tied down to a fixed, predetermined process path. Building out AI solutions in the absence of these constrained yet flexible man work management environments that are framed by formal architecture is a path more more organizations must explore sooner versus later. So you need to put in place these
47:50
Unidentified speaker (speaker_1): constraints and the architecture necessary to put some management structure over what you're doing. And I don't see that today. I see a very small number of people who are organizations who are thinking like that, and I see a whole bunch. It's just the Wild West, right? Just shooting from the hip, right? So there's enough technologies deployed today where
48:10
Unidentified speaker (speaker_1): architecture, and I use the term loosely, is so well defined and so convoluted that it's amazing that things work at
48:16
Unidentified speaker (UNKNOWN): all.
48:17
Unidentified speaker (speaker_1): So
48:18
Unidentified speaker (UNKNOWN): uh
48:18
Unidentified speaker (speaker_1): let me sort of try to wrap up a few of the things I covered today as I close out my discussion. I want to be clear, first of all, that I'm not by any means looking to eliminate humans from the loop. On the other hand, if organizations are going to continue spending tens of millions or hundreds of millions of dollars on AI, they should impart some rigor on the overall approach
48:38
Unidentified speaker (speaker_1): and have a vision as to where they're going to land. The dynamically enabled value stream is one such landing place and works from a strategic perspective
48:47
Unidentified speaker (UNKNOWN): as
48:47
Unidentified speaker (speaker_1): opposed to throwing as much spaghetti at the wall as possible. and trying to see what sticks.
48:53
Unidentified speaker (speaker_1): So the messaging here is formalize things in architecture. When we're getting to the point of understanding where things are and how they're documented, the full business architecture, the 10 domains, not just capabilities, not just even value streams, not just value streams and information and capabilities, but the full business
49:13
Unidentified speaker (speaker_1): architecture with the organization map and the breakdown of business units, the stakeholder map and the stakeholder breakdown and decompositions and relationships over to business units and over to the value streams that they play a role
49:28
Unidentified speaker (UNKNOWN): in. And
49:29
Unidentified speaker (speaker_1): of course, over to the capabilities where they're playing those roles in value stream stages. That's a formal structure. My guess is that it's a lot more than. you have today in terms of trying to document and figure out where a Gentec AI is going to be deployed. Now, if you keep going down the path that you're on with no formalization, no documentation,
49:49
Unidentified speaker (speaker_1): and just keep starting randomly deploying agents across your business organization without any kind of cohesion to bring it all together, I am not, won't be totally surprised if I start to see these AI agent wars. breaking out inside your organizations
50:06
Unidentified speaker (speaker_1): with one agent trying to get control over another agent. You've got to put some boundaries on it. And as much as leadership's been fighting architecture and wants to see it destroyed, at least that's my view with what I see in terms of what leadership's doing in some organizations and decoupling their architecture teams and sending people with their walking papers.
50:25
Unidentified speaker (speaker_1): As much as I see that.
50:27
Unidentified speaker (speaker_1): they need to stop doing that. They need to turn a page and start adopting architecture in the formalisms we're talking about. Otherwise, you're going to be in a Wild West situation.
50:37
Unidentified speaker (UNKNOWN): So I'm
50:37
Unidentified speaker (speaker_1): going to wrap up today. You've been listening to the Strategy Execution Podcast with William Ulrich, co-author of End-to-End Strategy Execution from Inception through Solution Deployment. Send your comments and questions to W M M U L R I C A. H. That's W Michael Michael U L R I C H my last name. At T S-G
50:57
Unidentified speaker (speaker_1): Consulting Inc., that's T as and Tom, S and the Sam G Consulting INC, Tsg Consulting Inc.com.
51:07
Unidentified speaker (UNKNOWN): You
51:07
Unidentified speaker (speaker_1): can check out the podcast at voiceamericamedia.com Voice America Media.com or subscribe to it on Apple Podcasts, Spotify, or wherever you get your podcast.
51:14
Unidentified speaker (UNKNOWN): You
51:14
Unidentified speaker (speaker_1): can get our book on Amazon.com or other outlets. I want to thank you for joining me today, and I will see you all next week.
51:33
Unidentified speaker (speaker_0): Continues uncovering the truth behind strategy execution and how to finally get it right. See you next week.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from End-to-End Strategy Execution

All insights