Leading Transformation

Evolinq CEO Noam Shakuri claims AI procurement agents cut buyer time by 60% and reduce errors

38 min Listen to the original Episode page
ProcurementArtificial IntelligenceSupply ChainEnterprise SoftwareManufacturingCost Reduction

Summary

Noam Shakuri, CEO of Evolinq, argues that manual procurement is inefficient and costly, citing $16 million in annual losses for average manufacturers due to downtime and safety stock. He introduces Evolinq's AI agents, which mimic buyer behavior via email integration to automate administrative tasks while keeping humans in the loop. Shakuri claims the system is 60% more accurate than leading LLMs and cites a defense electronics case study where it managed a 40% demand spike without production stops. The company offers a four-week onboarding process and charges based on purchase orders and users.

Key takeaways

  1. 1
    The average manufacturer in automotive, defense, electronics, semiconductors, and aerospace loses approximately $16 million annually, split between $13 million in production downtime and over $3 million in safety stock costs.

    Unidentified speaker (speaker_0) · 4:44

Episode sections

  1. 0:00 – 0:41

    Introduction and Problem Statement

    The episode introduces the inefficiencies of manual procurement, citing a case where an airplane load of components arrived by mistake due to an email error. The host highlights that buyers managing thousands of orders face shortages or excessive safety stock, leading to significant financial losses.

  2. 0:41 – 4:10

    Evolinq Solution and Onboarding Process

    Noam Shakuri, CEO of Evolinq, explains that their AI agents mimic buyer behavior by being copied on emails, learning the process within a week. The system integrates with existing ERPs and spreadsheets without requiring a new portal, automating up to 90% of administrative tasks while keeping humans in the loop.

  3. 4:10 – 12:25

    Industry Economics and Cost of Errors

    Shakuri claims average manufacturers in sectors like automotive and electronics lose $16 million annually due to downtime and safety stock. He notes that automotive plants can lose $1.3 million per hour of downtime, and electronic manufacturing operates on thin 5-6% margins.

  4. 12:25 – 14:27

    Technology Architecture and Security

    Evolinq operates on a secure private cloud with SOC2 Type 2, GDPR, and ITAR compliance. The company emphasizes that customer data is not used to train global models, ensuring privacy, while the agents are trained on anonymous, aggregated data from thousands of buyer-supplier interactions.

  5. 14:27 – 20:03

    Performance Benchmarks and Accuracy Claims

    Shakuri states that Evolinq’s AI models are 60% more accurate than leading LLMs like Anthropic’s Opus and OpenAI’s GPT. He mentions an upcoming public benchmark to validate these claims, positioning the company as the most accurate model for supply chain tasks.

  6. 20:03 – 22:19

    Implementation Timeline and Pricing Model

    The onboarding process takes four weeks, progressing from manual triggering to full autonomous operation. Pricing is based on the number of purchase orders and users, with specific agents for RFQs and invoices costing extra. A three-month trial period is offered at a discounted rate before annual agreements.

  7. 22:19 – 33:22

    Case Study: Defense Electronics Manufacturer

    A defense electronics client saw a 40% spike in demand during a conflict while losing two buyers. Evolinq’s agents managed 4,000 orders and 300 suppliers in five days, reducing order delays by 41% and cycle times by 34% without stopping production lines.

  8. 33:22 – 37:59

    Future Vision and Contrarian Take

    Shakuri argues that the current 90% automation goal is just the beginning, with a future vision of 98% autonomy including risk detection and resolution. He advises manufacturers to implement AI immediately rather than building in-house solutions, citing speed and global supply chain insights as key advantages.

Claims worth verifying

RH Group reported that AI agents from Evolinq cut buyers’ time on day-to-day follow-ups by more than 60%.

Unidentified speaker (speaker_2) ·0:41

How to verify: Check RH Group press releases or case studies published between 2026-08-18 and 2026-11-18; verify via RH Group investor relations page or TechCrunch articles from January 2026. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

TechCrunch named Evolinq one of the top logistics and manufacturing startups in January 2026.

Dr. Petra Doerken ·2:11

How to verify: Search TechCrunch Startup Battlefield archives for January 2026 winners or finalists list. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Evolinq’s AI agents are 60% more accurate than Anthropic’s Opus and OpenAI’s GPT models.

Unidentified speaker (speaker_0) ·14:56

How to verify: Wait for the public benchmark mentioned by the speaker; verify via Evolinq’s official website or published research papers post-2026-08-18. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Average manufacturer in automotive, defense, electronics, semiconductors, and aerospace loses around $16 million a year ($13M downtime, $3M safety stock).

Unidentified speaker (speaker_0) ·4:44

How to verify: Cross-reference with industry reports from Gartner, McKinsey, or Deloitte on supply chain costs; verify calculation against public financial data of large manufacturers in these sectors. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

An automotive plant can lose $1.3 million for every hour a line stops; a grounded aircraft can cost around $150k an hour.

Unidentified speaker (speaker_0) ·5:25

How to verify: Verify against FAA or automotive industry association reports on downtime costs; check public filings of major automakers or aerospace companies for disclosed downtime losses. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

Electronic manufacturing runs on about 5-6% margins.

Unidentified speaker (speaker_0) ·5:45

How to verify: Check SEC filings (10-K) of major electronic manufacturing services (EMS) providers like Flex, Jabil, or Sanmina for gross margin percentages. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Primary source

Evolinq has SOC2 Type 2, GDPR, and ITAR compliance.

Unidentified speaker (speaker_0) ·12:58

How to verify: Request compliance certificates from Evolinq; verify SOC2 via Attestation reports; check ITAR registration on DDTC.gov. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

A defense electronics client saw demand spike by 40% during a war with Iran; Evolinq reduced delays by 41% and cycle times by 34%.

Unidentified speaker (speaker_0) ·22:31

How to verify: Verify via Evolinq case study publication; cross-reference with news reports on the conflict and procurement challenges in the defense sector during that period. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Evolinq aims to automate up to 90% of procurement tasks currently, with a future goal of 98%.

Unidentified speaker (speaker_0) ·11:26

How to verify: Verify via Evolinq product documentation or whitepapers detailing automation capabilities and roadmap. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Evolinq was trained on hundreds of thousands of communications tagged by its team.

Unidentified speaker (speaker_0) ·14:07

How to verify: Request technical documentation or whitepaper from Evolinq detailing training data sources and volume. For recent events, verify 2026-08-18 through 2026-11-18. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify the RH Group report cited at [0:25] to confirm the 60% reduction in buyer time and the methodology used.
  • Check Evolinq's SOC2 Type 2, GDPR, and ITAR compliance certificates to validate security claims made at [12:58].
  • Review the upcoming public benchmark mentioned at [14:56] to independently assess the claim that Evolinq is 60% more accurate than Anthropic's Opus and OpenAI's GPT.
  • Analyze the defense electronics case study details at [22:31] to understand the specific metrics for the 41% drop in order delays and 34% cycle time reduction.
  • Investigate Evolinq's pricing model structure for purchase orders and users to calculate potential ROI against the claimed $16 million annual loss avoidance.
  • Search for independent customer reviews or third-party audits of Evolinq's AI accuracy in supply chain contexts beyond the CEO's claims.
  • Confirm whether the 'anonymous, aggregated data' training method at [12:25] truly prevents data leakage or model contamination as claimed.
  • Evaluate the four-week onboarding timeline at [20:03] against industry standards for enterprise AI implementation to assess feasibility.
Full transcriptexpand

0:00Unidentified speaker (speaker_0): In one day they just received an entire airplane full of orders and components. Like someone forgot to cancel it over email.

0:08Dr. Petra Doerken: How many of your purchase orders still live in someone's inbox? And what happens when one of them slips through?

0:15Unidentified speaker (speaker_0): Just imagine that you're a buyer, you need to manage four thousand orders at any given moment in front of hundreds of suppliers. They are left with one of two bad choices.

0:25Dr. Petra Doerken: Welcome to leading trends. Transformation. I'm Dr. Petra Delkin. In episode two, we talked to Tao Rosenberg, CEO of a manufacturing company, RH Group, and he told us about an AI solution that changed the way his company runs procurement. Oh,

0:41Unidentified speaker (speaker_2): which basically supplies us with smart agents for our supply chain. Basically, to reduce more than 60% of the buyer's tactic time, the emails and tech. day-to-day tactical operations.

0:53Dr. Petra Doerken: That made us curious. So we went straight to the source and talked to the CEO of that company that built the AI solution. So today that CEO shares with us how procurement even in the most advanced manufacturers are still running on emails and spreadsheets.

1:08Unidentified speaker (speaker_0): Everyone told me if you do anything manual more than once you need to automate it. And then I saw this the supply chain world and procurement. I saw everything completely manually.

1:19Dr. Petra Doerken: And how an AI agent can learn your buyer's work by simply being copied on their emails.

1:26Unidentified speaker (speaker_0): Watch and learn how the buyer is managing their process and then mimic the exact same thing.

1:31Dr. Petra Doerken: The early warning signs that are hiding in your supplier's emails. The

1:36Unidentified speaker (speaker_0): tone of the supplier, whether his tone is nice and friendly or not.

1:41Dr. Petra Doerken: And the mistake that most manufacturers make when they try to bring AI into procurement. Today's guest is Noam Shakuri, co-founder and CEO of EvoLink, the company that build AI agents that work as digital buyers for manufacturers. Before EvoLink, Noah NOAM spent over a decade leading engineering teams working in cybersecurity and risk intelligence, working with machine learning way before AI became a buzzword. In 2025, TechCrunch listed EvoLink as one of its top logistics and manufacturing startups startup in the TechCrunch technology startup battlefield. Stay with us. I

2:25Unidentified speaker (speaker_0): think not coming from the industry is more an advantage than disadvantage.

2:34Unidentified speaker (UNKNOWN): The

2:34Unidentified speaker (speaker_0): story of FivoLink is a crazy story. And my background is in high tech. For more than a decade they led engineering team in cybersecurity, Uh risk intelligence, AI, and machine learning, long before GPT and long before AI became a buzzword. And back then, I used to work with an electronic manufacturing company, one of the largest globally. And one day they just received an entire airplane full of orders and components that arrived by mistake. Someone forgot to cancel it over email. And that was a crazy story. It's like it's not like you ordered some. something by mistake on Amazon. You ordered tens of millions of dollars over email. And for me, it was shocked because like when I was in high tech, everyone told me like, if you do anything manual more than once, you need to automate it. And then I saw the supply chain world and procurement, and I saw everything completely. manually. And that back then, that was the thing that

3:44that was the reason we decided to found Evolving. We would like to solve the entire area of supply chain and procurement and help those companies to make sure on-time Time delivery.

3:58Unidentified speaker (speaker_3): What makes you think that the problem is big enough to build a whole entire company around? that? Because I mean, a lot of manufacturing leaders may not think that procurement is a technology problem yet.

4:10Unidentified speaker (speaker_0): Yeah. And

4:11Unidentified speaker (UNKNOWN): look,

4:12Unidentified speaker (speaker_0): when you look at manufacturer, they manage tens or hundreds of thousands of purchase orders a year, completely manually, and with a lot of disruptions every day because of the geopolitical situations and shortages. global shortages, they are left with one of two bad choices. The first one is like is to experience shortages.

4:34Unidentified speaker (UNKNOWN): And

4:34Unidentified speaker (speaker_0): a shortage means production downtime, operational downtime. And that's the biggest loss of a con that the a company can have. And the second is to try to avoid it. And then you need to have a lot of safety stock. You buy and store a lot of material. materials that you don't really need, just in case, and they pay for it for all of it. Now, look at the numbers. The average company in the script manufacturing will look at automotive companies, defense, electronics, semiconductors, and uh aerospace. They lose, in average, around 16 million dollars a year. About 13 million on production. downtime, and more than 3 million on safety stocks. And just to realize that, an automotive plant can lose $1.3 million for every hour that the line starts. An aircraft can that that is grounded can cost around $150k an hour. And those industries live with a very thin margin. margins. Electronic manufacturing runs on about I think five-six percent of margins. So

5:45every one of those losses losses really helps those companies. It's huge.

5:50Unidentified speaker (speaker_3): Sorry, I are you saying that because they are doing everything still very manually, right? Email and spreadsheets and phone calls. Are you saying that is why they are losing all this amount? of money?

6:03Unidentified speaker (speaker_0): Yeah, just imagine that you're a buyer, you need to manage now four thousand orders at any given moment in front of hundreds of suppliers, and you need to coordinate with each one of them to receive confirmation, to schedule deliveries, to receive documents, to ensure the compliance, and you need to pay attention for each and one of the orders and every single mistake can cost huge losses for the companies. So yeah, the money I'll work that's the one of the main reasons for that.

6:34Unidentified speaker (speaker_3): I see so it's a manual process so it's prone to human errors. That's what okay. And then the other thing that I when I talked to when we interviewed interview Tao, tal he mentioned that your system works work within the the spreadsheet. essentially and you know from what I understand you know you have no portal no chatbot nothing like that so like what does it look like really how does it work?

7:00Unidentified speaker (speaker_0): That's the question everyone asks and like the honest answer is it feels exactly like hiring a new buyer to your team. On day one we sit together and define what the agent is going to manage. Every day the agent reads the information from the company's system of record. It's the ERP, supplier, portal, whatever the company is, and it decides which orders need attention based on both MRP or the system-owned system own analysis. Then it does exactly the same what the buyer does. It emails the suppliers on behalf of the buyer with the buyer's signature, the company logo, the buyer, the buyer in CC. It doesn't look like a robot. It's take exactly and mimic exactly the same process of the buyers. Whenever the supplier replies, it might be with free text, PDFs, Excels, Excel, screenshots, attachment, whatever, the agents. read it. It analyzes each one of the responses, the status of each one of the order lines, make sure the uh on-time delivery, the quantity, compliance, etc., and whether

8:15everything is correct and uh and approved, someone needs to approve it with a single click by the end of the day or on the next day. But everything Evolution does, it's basically executing process in front of the vendors. Everything is supervised by the buyers. They can monitor, they can watch everything, and eventually they can decide which things they would like to do manually and which to the agents. But the agents act exactly the same as that.

8:45Unidentified speaker (speaker_3): Is this rule based or how how is it trained? Like how how does it make decisions decision, for example, um you mentioned for example ensuring others arriving on time. So does it make decision to email or make phone call or you know, like is that a rule based decision or is that uh you're training the model based on where is the data from?

9:08Unidentified speaker (speaker_0): Yeah, so it's a mix. We already have a lot of best practices that we bring with us to the companies and are very recommended. But the company can configure anything they want. They can configure which supplier, which buyer we should act on behalf of him, to whom to send alerts and escalations, which warehouses rely on which holders, orders and how the system can recommend. recognize it. Basically, the company can configure everything in Evelyn. Configurable. From the people and the stakeholders, we need to escalate issues, which range of delays is approved, or which we need to escalate. After that, what price changes are approved and the agent can manage autonomously. and what price issues and price ranges the system need to escalate and to whom. Now, that's the first thing. So the first one is configuration of the company. The

10:15second one is the agents. The agents know how the organization's And a management managing process by watching them. You need to, at the beginning of the on-board. onboarding, you need to put Evil Link on CC on your emails, on your escalations, on a couple of processes. It's a very simple onboarding. Take less than a week. And by that, Evilink watch and learn how the buyer is managing their process and then mimic the exact same thing. And the third one is like the agents can improve some of the decisions. that it would like to do. So we also have some decision improvements of how to improve the relationship with the suppliers. How to like we have anti-spam algorithms that aims to not spam your supplier and improve the relationship because we can communicate only on the important. things and yeah and more other things.

11:19Unidentified speaker (speaker_3): Does it replace the buyer? Or it's does it it sounds to me that you still need a human in the loop, right?

11:26Unidentified speaker (speaker_0): That's a great question. So we aim to automate up to 90% and that's what we see with our companies. But it doesn't need to replace the buyer. It's basically taking all the administrative tasks, all the manual work needs to send emails, read emails, enter information, and reduce errors and human prones, and basically make the buyer focus on the most important things, on the escalations needs to be managed, on the alternatives that needs to find be found, on making a better relationship. with the suppliers. So even help the buyers to focus and be a real buyer, like handle procurement, relationship and everything needed to to actually bring uh uh cost reduction to the companies and not being you know um

12:21Unidentified speaker (UNKNOWN): the

12:21Unidentified speaker (speaker_0): one that's making all the administrative tasks because that's not procurement.

12:25Unidentified speaker (speaker_3): So is this right running on-prem or is this running like with in-cloud deployed within the client's environment or is that running from your end and then accessed by the client?

12:39Unidentified speaker (speaker_0): Yeah, so EvoLink is on the cloud. We're running on a secure environment, having all the certificates uh certificates needed for that, including SOC2 type to 2, GDPR for the European companies, ITAR, which is NIST behind the scenes. We have all the compliance needed. Also, the agents and the AI is in our own private cloud. Nothing can learn on the customer's data, no information or data of the customer. is being shared with others or being trained on. Everything is in our own private cloud, and we're basically creating an environment for each one of the customers. So every customer gets his own secure private environment.

13:32Unidentified speaker (speaker_3): So then earlier you mentioned that you bring in the industry best practice. So how do you, if you don't use um if you don't use the client's data in order to train your model, then where do you get the information or data to get the best practice?

13:47Unidentified speaker (speaker_0): Yeah, so it's mixed with uh experience of the team and a lot of anonymous data that was trained at the beginning of of the journey of evolution eviling and keep being trained all all the time. So just for example. our agents were trained over hundreds of thousands of different communications and conversations between buyers and suppliers that were tagged by our team. We have the best AI researchers and AI engineers, and they basically created the best AI models for procurement. And just for that, we are going to publish a benchmark soon, a public benchmark to measure um to measure LLM models. It's very common in different areas of LLMs, and Evolink was found 60% more accurate than any other LLM. We compared it to an a Opus of uh Anthropic to a soul of um of a

14:52of an OpenAI, and evilink is 60% more accurate than any other model. And for supply chain, it's the most important thing to help those companies to have an accurate information and data about their orders. And that's what we do basically.

15:07Unidentified speaker (speaker_3): Very nice. So I'm still curious, So right? So there's no dashboard or portal where they log on, so for example, you mentioned that the buyer can make decisions as to what they want to jump in and do manually. So how do they interact, other than you you mentioned about CCing Evo Link agent, but how do they interact with the agent?

15:29Unidentified speaker (speaker_0): Yeah. So the first the first interaction is using emails. Again, just like another employee, another team member that's putting you on CC, interacting with you whenever you need something. The second one is our platform. We have

15:45Unidentified speaker (UNKNOWN): two

15:46Unidentified speaker (speaker_0): platforms actually. The first one is for the tactical buyers, the the those that are managing the agents and doing the administrative tasks. And the other for the management. For the buyers, we can integrate, we have add-ins for um. the excels for the uh the emails outlook etc we're trying to be in every environment that the buyer already using in the in the day to day so it's very easy for them to to get to know evil to educate on it because there is no new platform they need to learn they know there is no something very complex for them to study. Then you have the management platform. The management platform is a control and command platform giving visibility over everything, all the orders, all the email conversations and histories, all the documents that were interacted with the suppliers, and supplier

16:47evaluation page, giving you logs on of everything. that the supplier did during the year, and what's the reason that Evilink evaluated the supplier with that score. A buyer evaluation page to be able to set goals and KPIs with your teams. Action items that need to be done, and more and more. So the management platform is basically the platform that the manager can open every Monday, start the week with it, see all the action items, see what's wrong or what they need to strengthen during the week, and then take actions with with the team.

17:27Unidentified speaker (speaker_3): Oh very cool. Very cool. And then you mentioned that you have work with a a range of industry like electronics, you mentioned automotive aviation, I think even banking. So those are very different industries and they have completely different procurement complexity and risk tolerance and requirements. So do you how do you build a platform that serves all those industries? And what do you see as the commonality between them and what are the significant differences?

17:55Unidentified speaker (speaker_0): Yeah. Although it looks very different, basically it's the same. Procurement is procurement and it's exactly the same process eventually from a company to another. Um what change are the rules? Like procurement basically, you place an order, you need a supplier to confirm it, you need it to arrive on time, the right place, with the right compliance and prices, right? You you you have deliveries, you need documents, procurement is the same process. from with uh in different industries but what changes the rules are the rules in aviation and defense compliance is everything so you need certificates you need ITAR you need a country of origin and documents around it in electronics it's more about speed and volume thousands of parts prices that change every week. And in automotive, it's like the on-the-on-time delivery, right? The production line can never stop. Huge losses for them. But the process is the same. So, what we've built, we built one core engine and rules on top of it. So every customer can configure their own thresholds, their own escalations, compliance.

19:15checks, and basically it's the same agents, just trained for different jobs, like a buyer who moves from electronics company to uh aerospace companies. But it's completely the completely the same, the skills are the same, but the rules are different.

19:31Unidentified speaker (speaker_3): So tell me tell me about like the risk or you mentioned that when you deploy um the agent in a new client for example, they would then the agent would then go in and the buyers basically CC the agent, right? And including them in escalation and things like that. So for a week or so. So then do you come out from that point beyond the POC, do you I assume that the accuracy is still gonna increase over time. So the beginning, like the risk at the early stage, like how do you manage that? And what are we looking at? What are we talking about?

20:11Unidentified speaker (speaker_0): Yeah, so the onboarding is is being managed is being managed step by step. We're enabling and turning on the automation, um, and but in steps. On the first week, the buyers only call. copy the system, the system trained over the data of the company for their specific use, and they starting to see a lot of analysis of the system. On the second week, the buyers starting to manually trigger the system to track and to um start communications with the vendors. Okay, so they choose. which orders are going to be sent through the systems, the system is going to coordinate, etc. The third week, we already moved to a full operation with all the suppliers, all the orders after they already trust in the system. They saw the results of the system, they started to gain gain some trust, and that's when we moved. move to the full operation. On the fourth week, we turn on all the automation. The buyer doesn't need to keep and trigger which orders to send. The system does it autonomously. The buyer doesn't

21:30need to check everything of the system because he already trusted. The system starting to escalate automatically. issues to different stakeholders within the company. And more and more. So on the fourth week of the POC, the customer already sees the full operation of EvoLink operating in front of 100% of the suppliers and operate also internally with the different stakeholders.

22:01Unidentified speaker (speaker_3): So we have heard um you know. a successful case, right, about Ivan Link from Tao Rosenberg with his company, which is electronics manufacturers. Can you give us another example of a successful case? And what was the kind of situation that they were in and how things have improved, or what didn't work?

22:19Unidentified speaker (speaker_0): One of our first customers was in electronic manufacturing for defense. And during the war with Iran, their demand was spiked by 40%, 40% more purchase orders. And at this exact same time, two of their buyers quitted. Two buyers both both gone in the middle of the war. They have a lot of purchase orders, and that was a crazy story for us. Now, this is a company that cannot stop production. Their customers are defense programs. So we came in and Evolin basically became their procurement team. Within only five days, the agents were managing 4,000 purchase holders, following up with 300 suppliers and escalating only the real problems to the management. And the results in those in this company was that the delays order dropped by 41%, and other cycle times dropped by 34%. They got through the crisis without stopping a single production line. And all their measurement, all the KPIs of procurement were improved, and they did it without hiring and training new buyers. And all of that in the middle of the war. wall. And that was

23:37a really great case study for us.

23:41Unidentified speaker (speaker_3): So in the past, um, when you talk about software, it is per seat kind of situation. Now you are in the next generation of SAS, if you even call it that. How do you price your product or what does it cost for your customer? You know, for someone who going in and go, Okay, I'm considering to

24:02Unidentified speaker (UNKNOWN): to

24:02Unidentified speaker (speaker_3): you know get uh evilink into my system how should they plan for this how do they figure out what is it gonna cost them

24:10Unidentified speaker (speaker_0): yeah so the price is regarding the amount of purchase orders and use the users that using the system so that's the main the main price uh parameters also every one of the agents that we have for RFQs, for invoices, for um order intake, every agent is an additional cost. And then we build a package for a company that is fit to their needs regarding the amount of purchase orders, etc.

24:46Unidentified speaker (speaker_3): So there's an initial one-time onboard cost per agent, and so essentially you have multiple agents that are doing different aspects of procurement side. And so there's an initial onboard setup fee, essentially, is that right? One time. And then there's an ongoing per and that based on the amount of work. So outcome-based, essentially.

25:06Unidentified speaker (speaker_0): Yeah. We have a very discounted one-time fee for the onboarding, training, etc. We also give a very three month um the first month free, free of charge. So the company will be able to test the system. They can keep and use the system for another three uh two months, giving them a total of three months of trial period in payment, but um they pay for it, but still they do not commit to anything, um, and it's a very discounted price. After they see the ROI, after they are already ensure that Evolin can help them, we keep and move to an annual agreement. And I'm proud to say that hundred percent of our customers kept uh keep to the annual agreement. Every one of them saw And so the impact of Evolv Evelyn during the three months.

26:00Unidentified speaker (speaker_3): So when they try to bring AI into their procurement Have you seen other situations where they m your clients may have made mistakes in the past that might have made them act differently with you or things like that?

26:14Unidentified speaker (speaker_0): Yeah, I think that I think that most companies

26:17Unidentified speaker (UNKNOWN): are

26:18Unidentified speaker (speaker_0): trying to repair so many things for AI and event eventually getting a solution of a dashboard. A dashboard and a chat they they they can uh communicate with and talk with. Now I think the most important thing now in the decade of AI, you need to look for real solutions. AI is a solution that takes decisions that can handle every Ever any kind of data. You don't need to prepare infrastructure for that, you don't need to move so many things in your company for it to work. It's supposed to just work because it's a software, a system, or a model that actually can learn. Learn just like anyone that you hire to your team. And that's the most most important thing. When you look for solutions, first look for something that actually can learn and work for you without you need to change anything internally. You don't need to prepare data, you don't need to do changes. And second, is make sure that it's secure enough, that it has all the guardrails in private place, that it cannot train on your data, and make sure that it's secure enough so so you'll your data will be secure as is also.

27:37Unidentified speaker (speaker_3): So a lot of companies. think about the buyer relationship with the supplier is a very human oriented connection, right? There's a lot of human relationship elements. So how do you see this automation? I think that could create some hesitation for people to consider putting AI into that. So how do you see that automation and the human judgment? That what is a the balance?

28:03Unidentified speaker (speaker_0): Yeah. I think today buyers lose control because they don't have they don't have time. When one buyer managed thousands of suppliers and thousands of holders, most of suppliers only heal from him when you need something. When something is late, when you need something urgent, etc. But with Evilink Evil Link, every supplier gets a real attention every day, every order. The emails go in and out in the buyer's name. And the buyer now has much more time to develop the relationship with the supplier to be able to less spam it. Because nowadays buyers sending tons of emails females to the suppliers. because they need to get a status. But with Evilink it's not the case. Evilink communicate only when needed and the buyer get the uh the ability to actually connect with the supplier to improve the relationship.

29:04Unidentified speaker (speaker_3): You mentioned earlier about data, uh insight actually. And I also remember the conversation with Tao where he kinda kind of hinted about the fact that once you have gone in you have even more data and more insight than you expected. So for people that before they put Evil Link in, they recognize certain factors as important, but once you put Evil Link in, you certainly have so much more data and be able to see. So could you share some of those insights?

29:32Unidentified speaker (speaker_0): Sure. So, yeah, that's correct. Even If only it can get data not only about the others, but a lot of different elements that you you have never looked for. Like times of responses, how much time it takes for your buyers internally or to the suppliers to respond, the tones, the tone of the supplier, whether his tone is is nice and friendly or or not. Different elements of signals about disruptions. You start Evolin can start and see changes in the supplier behaviors in later, slower responses, lead time increases, price increases, and much more. So there is a lot of metadata about the conversations that now Evolin can gather for you and also. for us, like we're also showing a lot of insights to the company about disruptions and risks,

30:33etc. But the company can also have this data as well.

30:37Unidentified speaker (speaker_3): So today there's a belief that the coding, you know, technology is no longer the barrier. So what would be the consideration for a company to think about like why don't we just go to, for example, Cloud Code or Codex or you know, like some kind of vibe coding and do this in-house versus go to you and pay you?

30:58Unidentified speaker (speaker_0): Yeah, that's a great question. First of all, why build if you can just take something ready, right? Because building a solution is not is not cheaper. You need to build a solution that you can actually trust on to be able to um give your people the the and note your people that they can live and not manage, like stop managing procurement because they can trust on the system. Second, you need to keep and um modify it and uh manage it. And third, Evolink evolution give gives you the ability to not only get insights about your data and manage your data, but also give you insights from the global supply chain. Without using any private data, we're also enriching this the network of Evolink with a lot of information from the global supply chain. So you can see risks Can uh before this happens. You can see recommendations about what you need to do. Um so it's not only about managing things internally, but also getting a lot of information externally. But except that EvoLink build something can take you months, maybe weeks, to get something very light.

32:17With Evolink, in a couple of days, you can already turn it on, get all the time improvements, the best practices, the best models, the things that And yeah, you pay for it, but it's not more expensive than building it.

32:37Unidentified speaker (speaker_3): How is that compared to the human salary?

32:39Unidentified speaker (speaker_0): I think it's not the right comparison because the right comparison is how much is is going to cost you when you need to take an order through a flight and not on uh on a ship? How much you're going to pay and what's the cost of a shortage? That's the real cost of of procurement that doesn't be managed right. And I think that's the comparison. And comparing to that, Evolink is super cheap.

33:06Unidentified speaker (speaker_3): You mentioned that you aim to automate up to 90%. ninety plant. So how much of that is realistic? right now? You can be there right now versus how much of that is still aspirational? And what part of it are still aspirational?

33:22Unidentified speaker (speaker_0): Yeah, so it's everything is real. It's happening. It's now. It's running and live in huge manufactured airline companies. Like it's already live. It's here. Everyone can use it with a really short onboarding and configuration. And the aspirational part is about what we keep doing after the execution because Evilink is not only an execution system, we're executing, and that's what we are doing today with agents, but it's execution, detection of risk intelligence, about disruptions, and etc., and autonomous autonomy resolution of the the issues. So eventually what you are going to see is an execution agent detecting risks, finding alternatives, find for you everything you need to for making decisions, and giving you the ability to resolve it with a single click. So you're going to get

34:22Unidentified speaker (UNKNOWN): a

34:22Unidentified speaker (speaker_0): full autonomous solution, and then it's not going to be 90%. it's going to be more of 98%.

34:29Unidentified speaker (speaker_3): Wow. You mentioned that your background is in tech, but not really in manufacturing or aerospace or electronics like or automotive like these clients are. So, how did you get the insight? Oh, even in procurement, right? That's not even your background either, not in the supply chain. How did you get the background to build this? Did you do did you hire specific people into the team early on so that because that's one of the key thing right now that I see in AI adoption is that there are a lot of builders who are in love with technology but doesn't have the reality grounded in real workflows and in in real m industry. So how did you do this at the beginning?

35:13Unidentified speaker (speaker_0): Yeah so I first exposured to physical supply chain a couple of years ago, I think about five years ago. And when I realized how huge is the problem, I started to lift it. I was a buyer myself for a couple of days, but I tried to experience that. I started to work with a lot of different manufacturers to experience that. And of course, I brought the best brains, the best people to be with me near me, because that's the most important in any journey of a startup. But I think not coming from the industry is more an advantage than disadvantage. Because we don't have limitations. We never knew before Evolink what people used. to do or what they were limited to do. We just were were able to rebuild everything from scratch. And it's a very visionary thing, but we actually succeeded to to implement it. And that's the amazing thing. We took the imaginary Imaginary the vision that we succeeded to draw and

36:26implemented. it with the right product to the to the uh industry. So it's kind of a mix, but yes, we do have the best people from the industry as well.

36:37Unidentified speaker (speaker_3): So before you go, I'm curious about your contrarian take. What what is the manufacturing procurement industry is getting wrong right now that you you wish that someone would say it out loud?

36:49Unidentified speaker (speaker_0): I think as Nike Nike say, just do it. AI is real, is here. Everyone can implement it. It's very easy whether you use

37:00Unidentified speaker (UNKNOWN): the

37:01Unidentified speaker (speaker_0): biggest LLMs like Cloud, GPT, or others, or vertical solutions that give you ability to solve things. But it's very easy. It's in the first of writing a prompt and you have everything. out there. Just implement, improve every process of the company in every department and every place.

37:23Dr. Petra Doerken: Thank you for listening to Leading Transformation. We are always interested in seeing how other leaders navigating technology adoption and operational transformation within their own organizations. What's working, what isn't, what surprises them. So if you are an enterprise leader, and you are open to share, we'd love to hear from you. It could be a private and confidential conversation, or you could be a guest on the show. So reach out to us directly on our LinkedIn profiles or on the show website, bleeding transformation podcast.com.

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.
All insights