HSBC and SAP discuss embedding treasury services in ERP and AI prerequisites
Summary
Amber Henderson-Smart and Falk Rieker discuss the shift from manual file-based treasury operations to embedded connectivity within ERP systems. They highlight HSBC's 'Payments for HSBC' app in SAP and the standardization of multi-bank connectivity. Rieker outlines how real-time data and AI improve cash management and fraud detection. He identifies four prerequisites for AI adoption: a single finance database, robust connectivity, documented policies, and an AI governance model. The speakers conclude that AI shifts the treasurer's role from execution to strategic monitoring and policy definition.
Key takeaways
- 1 Yeah, I I think controls are becoming the new reconciliation checks.
Unidentified speaker (speaker_2) · 8:38
Episode sections
- 0:22 – 1:33
Introduction and Partnership Context
Amber Henderson-Smart introduces the episode recorded at Sibos 2026, welcoming Falk Rieker from SAP. She notes the HSBC-SAP partnership began in 2011 and sets the stage for discussing real-time treasury and AI.
- 1:33 – 2:54
Shift from File-Based to Embedded Connectivity
Falk Rieker explains that corporate expectations have shifted from manual file exchanges and phone calls to demanding banking services embedded directly within ERP and TMS platforms for immediate decision-making.
- 2:54 – 6:34
Standardization and Multi-Bank Connectivity
Henderson-Smart highlights HSBC's volume of ERP projects and the launch of the 'Payments for HSBC' app in SAP. Rieker cites a client's struggle with disparate bank documentation as the driver for standardizing multi-bank connectivity.
- 6:34 – 8:08
Real-Time Data and AI Use Cases
Rieker identifies cash/liquidity management, payment error resolution, forecasting, and fraud analysis as key areas where real-time data and AI create value by reducing the time between an event and actionable insight.
- 8:08 – 10:22
Controls and Governance in Automated Workflows
The speakers discuss how controls replace traditional reconciliation checks in automated environments. Rieker emphasizes the need for end-to-end governance across the ERP, connectivity layer, and bank side.
- 10:22 – 15:44
Foundations for AI in Treasury
Rieker outlines four prerequisites for AI adoption: a single finance/treasury database with high data integrity, robust connectivity, clearly documented treasury policies with defined risk parameters, and an AI governance model.
- 15:44 – 19:02
Practical AI Applications and Human-in-the-Loop
Rieker describes current AI use cases including intraday position analysis, yield optimization recommendations, and workflow initiation based on policy limits. He stresses that humans currently supervise agents to ensure outcomes are correct before scaling.
- 19:02 – 21:23
Evolution of the Treasurer's Role
Rieker concludes that AI shifts the treasurer's role from execution to monitoring agents, defining policies, and performing strategic judgment, requiring an elevated expert profile to validate AI outputs.
Claims worth verifying
HSBC and SAP formed a partnership going back to 2011.
Unidentified speaker (speaker_1) ·0:42
How to verify: Check HSBC or SAP press releases or partnership announcements from 2011; verify via company investor relations pages or news archives. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.
Primary sourceHSBC is doing around 4,500 ERP projects per year.
Unidentified speaker (speaker_1) ·2:54
How to verify: Verify via HSBC corporate communications, press releases, or annual reports discussing client integration volumes. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.
Primary sourceHSBC launched the 'Payments for HSBC' app in SAP a few weeks before the recording.
Unidentified speaker (speaker_1) ·4:13
How to verify: Check SAP App Center or HSBC press releases for the launch date of this specific integration; verify via SAP marketplace listings. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.
Primary sourceHSBC has 11 or 12 clients live on the standardized multi-bank connectivity.
Unidentified speaker (speaker_1) ·5:30
How to verify: Verify via HSBC press releases or case studies published around the time of the recording (Nov 2026) detailing client adoption. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.
Primary sourceAI output quality is dependent on the quality of the input data.
Unidentified speaker (speaker_1) ·9:53
How to verify: This is a general industry principle; verify via HSBC or SAP white papers on AI governance and data integrity in treasury. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.
Treasury policies must be documented with clear risk parameters and constraints to be used by AI.
Unidentified speaker (speaker_2) ·11:11
How to verify: Verify via SAP or HSBC thought leadership publications on AI governance in treasury; cross-reference with regulatory guidance on algorithmic decision-making. For recent events, verify 2026-08-11 through 2026-11-11. Use another period only when the speaker explicitly names it.
Current AI usage in treasury involves humans supervising agents, not fully autonomous execution.
Unidentified speaker (speaker_2) ·17:53
How to verify: Verify via SAP or HSBC case studies or white papers describing current AI implementation models in corporate treasury. 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 HSBC's annual ERP project volume of 4,500 against HSBC's public financial reports or press releases.
- Check SAP's official documentation for the 'Payments for HSBC' app features and ISO compliance standards.
- Review HSBC's Sibos 2026 press materials for details on the partnership timeline and scope.
- Assess the feasibility of 'single finance database' requirements for large enterprises with legacy systems.
- Investigate SAP's AI governance models and how they handle risk parameters in treasury workflows.
- Compare implementation timelines for standardized vs. bespoke multi-bank connectivity solutions.
- Look for case studies on 'human-in-the-loop' AI validation in corporate treasury environments.
- Check for any conflicts of interest given the joint nature of the podcast between HSBC and SAP.
Full transcriptexpand
- 0:00
- Unidentified speaker (speaker_0): Welcome to HSBC Global Viewpoint. The podcast series that brings together business leaders and industry experts to explore the latest global insights, trends, and opportunities.
- 0:12
- Unidentified speaker (UNKNOWN): Make
- 0:12
- Unidentified speaker (speaker_0): sure you're subscribed to stay up to date with new episodes. Thanks for listening, and now on to today's show.
- 0:22
- Unidentified speaker (speaker_1): Welcome to HSBC Global Viewpoint. I'm filming this episode live on location at Cybos where industry leaders come together to explore the forces shaping finance. I'm Amber Henderson Smart, Global Head of Client Connectivity at HSBC, and I'm pleased to be joined by Folk Riker, SVP and Global Head of Banking at
- 0:42
- Unidentified speaker (speaker_1): SAP. Thank you, Folk, for joining me today. HSBC and SAP formed a partnership going back to 2011. And in this time, a lot has changed when it comes to connectivity and how Treasury looks. In this episode today, we're going to explore how HSBC and SAP are helping our Treasury clients to
- 1:02
- Unidentified speaker (speaker_1): move toward real time Treasury and, of course, AI in Treasury as well. Okay, so let's kick off with the first question. So, Folk, since HSBC and SAP started working together, connectivity has evolved probably from what was a very technical project really into more of an enabler for Treasury optimization and
- 1:22
- Unidentified speaker (speaker_1): finance within the business. Maybe to start by bringing the audience up to speed. What has that shift meant for ERP providers like SAP and for banks?
- 1:33
- Unidentified speaker (UNKNOWN): Yeah,
- 1:33
- Unidentified speaker (speaker_2): I think the expectations from the corporates have changed completely yeah, over the past 15 fifteen years. Uh when you look back how Treasury worked fifteen years ago,
- 1:44
- Unidentified speaker (UNKNOWN): uh
- 1:44
- Unidentified speaker (speaker_2): where we sent files back and forth, uh we had a lot of uh Excel sheets. And uh when we wanted to determine the Interday p uh position,
- 1:54
- Unidentified speaker (UNKNOWN): then
- 1:54
- Unidentified speaker (speaker_2): we had to pick up the phone, yeah, and call a couple of our counterparts to see if the payments are coming on order or payments are made.
- 2:02
- Unidentified speaker (UNKNOWN): And
- 2:02
- Unidentified speaker (speaker_2): uh I think that has changed dramatically. Yeah, when you when you look at the expectation today from a corporate treasurer, he wants to have banking at his fingertips. Yeah. He wants to have it right where he is making his daily decisions: if it is payments, if it is cash, if it is liquidity,
- 2:22
- Unidentified speaker (speaker_2): if it is
- 2:23
- Unidentified speaker (UNKNOWN): uh the
- 2:24
- Unidentified speaker (speaker_2): hedge strategy, foreign currency. positions, you know, everything he wants to have banking services right at a sense.
- 2:34
- Unidentified speaker (UNKNOWN): Yeah,
- 2:34
- Unidentified speaker (speaker_1): absolutely. I think what's interesting for us too, we look after the integration with our clients. And I think per year now we're doing around 4,500 ERP projects, and definitely seeing a shift of clients wanting to work in the bank portal or multiple bank portals if they're multi-banked to moving into operating into their uh ERP, TMS, and internal ecosystem. So
- 2:54
- Unidentified speaker (speaker_1): definitely echo uh that shift. So just with that in mind, clients increasingly expecting banking to be within the platforms that they're using. Um, what do you think banks and ERP providers need to think about to make that experience simple, controlled, and scalable?
- 3:11
- Unidentified speaker (UNKNOWN): And
- 3:11
- Unidentified speaker (speaker_2): the first thing you know we have to have in mind when we want to satisfy the expectations of the corporates is obviously that we create business value. The pure technical connectivity is table stakes. Yeah. So to make this an experience that really results in better decisions,
- 3:31
- Unidentified speaker (speaker_2): in faster decisions, in fewer arrows, we need to embed these banking services and banking. data
- 3:40
- Unidentified speaker (UNKNOWN): in
- 3:40
- Unidentified speaker (speaker_2): the Treasury workflow.
- 3:43
- Unidentified speaker (speaker_2): And that's what is essential. Yeah, that SAP and HSBC make sure we embed this information in workflows, which are predefined, yeah, with the respective controls, so that it is an experience that results ultimately in predictable outcomes.
- 4:03
- Unidentified speaker (speaker_2): in governed processes and obviously in less effort for the treasurer. So at the end of the day, we need to make the life of the treasurer better.
- 4:13
- Unidentified speaker (speaker_1): Absolutely. And I think the other thing also is the technology teams, right? Supporting our treasurers. So how do we make the integration much easier? I know we've partnered together with the multibank connectivity with SAP, and actually a few weeks ago we launched the uh uh payments for HSBC, I think we called it in SAP app. So basically clients can automatically generate their ISO
- 4:33
- Unidentified speaker (speaker_1): compliant payment files without even needing to move outside of the ERP sort of ecosystem, which is quite uh significant.
- 4:40
- Unidentified speaker (UNKNOWN): And
- 4:41
- Unidentified speaker (speaker_2): and that's a great example that you bring that up, because you know the one of the triggers for multi multi-bank bank connectivity was we obviously talked to some of our corporate clients and I had the privilege to be part of that process and talk to a couple of corporates. And then uh I never will forget I went to one large American customer of
- 5:00
- Unidentified speaker (speaker_2): ours, and uh
- 5:02
- Unidentified speaker (UNKNOWN): he
- 5:02
- Unidentified speaker (speaker_2): showed me all the folders he had that he needed to read through for the different SAP releases, how he connects with Bank A. And then he showed me another room. where he had more documentation, how he connects to other banks. Yeah. And from from that we really said, you know, this
- 5:22
- Unidentified speaker (speaker_2): doesn't add any value. Yeah. So we need to standardize this. And it's great to see that you're building on this uh standardized connectivity.
- 5:30
- Unidentified speaker (speaker_1): Absolutely. I think we've got uh 11 or 12 clients live now and we're seeing the acceleration. So what used to probably be multiple months, if not sometimes a year. Again, very dependent on our client and sort of where they're at. But now we can do that in a matter of weeks, which is significant.
- 5:45
- Unidentified speaker (speaker_2): And and just think about the maintenance effort that goes away on your side as well as on the on the corporate side.
- 5:51
- Unidentified speaker (speaker_1): Absolutely. And I think again, you know, we're going to talk a little bit about things like AI, real-time, being able to bring in real-time balances Time shortly. I think having that reusable uh connectivity that can evolve is really fundamental to making it easier for the treasurer to get value. Um, so thinking about real time, so I think that's now shorthand for modern treasury. Real time
- 6:11
- Unidentified speaker (speaker_1): treasury we we hear about for a number of years. Um, but you know, information arriving faster only helps if it's really changing an outcome. Um, so where do you see real real-time time data really creating an advantage for clients? Um, and is there any anything that you think needs to happen inside banks or ERP providers to uh
- 6:30
- Unidentified speaker (speaker_1): really enable that value for the treasurer?
- 6:34
- Unidentified speaker (speaker_2): Yeah, well where do we see the the first results? Um
- 6:38
- Unidentified speaker (UNKNOWN): uh
- 6:38
- Unidentified speaker (speaker_2): certainly around everything around uh cash and liquidity management. When you think about um, you know, you want to determine your intraday cash position, you want to opt optimize your liquidity position. This is exactly you know where real-time information is so valuable. Another area is around payment
- 6:58
- Unidentified speaker (speaker_2): execution and not only successful payment execution, but also in analyzing payment errors. Yeah, uh there uh having a faster time to resolution is extremely valuable, and then everything around forecasting. Yeah, these are areas.
- 7:15
- Unidentified speaker (speaker_2): where I believe real time makes a huge difference. Fraud is another topic. Yeah. Payment fraud patterns analyzing of this and uh smart payment routing is another area where AI can be extremely uh helpful. So there are a lot of use cases where AI can Um help.
- 7:34
- Unidentified speaker (speaker_1): Absolutely. And I think uh it's not just about faster data, I think that gives a really helpful foundation. But if you all the examples you gave is really about reducing the time between something happening, a treasurer knowing about it, and being able to take action. And I think that's for us why we've focused a lot on building APIs and connectivity to deliver real-time balances into the ERP,
- 7:54
- Unidentified speaker (speaker_1): but also, like you say, exception information, payment status, And rejection, etc., so uh clients can take action very quickly. Um, I think what else is uh interesting here is really on the controls within the workflow.
- 8:08
- Unidentified speaker (speaker_2): It's a very important topic, yeah.
- 8:11
- Unidentified speaker (speaker_1): There's often a you know a fear or a worry. You mentioned fraud as well. Um, you know, what's your view around controls? Uh historically, clients or even today still coming into the bank portal, they're using the authentication two two-factor, factor, they have their permissions. Now, as we're bringing the bank more into the uh the ERP, we're more reliant on some controls in the
- 8:30
- Unidentified speaker (speaker_1): ERP. What's your view on how we align that across sort of that connection from ERP into bank for the client?
- 8:38
- Unidentified speaker (speaker_2): Yeah, I I think controls are becoming the new reconciliation checks. So uh when you think about the old world, yeah, with uh the with the Excel sheets and the batch processing and all this, where you know shifted shift data the from left. to right. You had all these reconciliation checks. Now we are in a very automated scenario, and in an automated
- 8:57
- Unidentified speaker (speaker_2): scenario, we have less uh reconciliation needs, yeah, because everything is automated, uh less manual, I should say. But you need to have the controls, that's very important. And it's not just the controls in the ERP system, it's the controls in the connectivity layer. It's also the controls on
- 9:17
- Unidentified speaker (speaker_2): the bank side, obviously, the terminal checks. Yeah, that need to happen. And you need to have that entire loop at the end of the day documented so that you have the the end-to-end governance uh of uh the respective treasury transaction.
- 9:33
- Unidentified speaker (speaker_1): Yeah, absolutely. And that sort of observability and audibility as well, which I know we're gonna going move to into some. questions on AI becomes increasingly important as well. Um, so to speak about AI, so again, we spoke about real time visibility. I think real time delivery of data is really important for agentic use cases as well. Uh, we were chatting earlier and you
- 9:53
- Unidentified speaker (speaker_1): know made the comment that you know the AI output will only be as good as the input. So we definitely see the connectivity and the delivery of trusted data and the speed of that really enabling. that experience. Um, so if we just think about uh moving toward AI assisted treasury, uh, what foundations do you think clients need to start putting in place
- 10:13
- Unidentified speaker (speaker_1): to enable AI in treasury? And do you have any views around how they should maybe think about prioritizing um that investment?
- 10:22
- Unidentified speaker (speaker_2): I think the the first one, and it sounds like table stack. stakes but it's actually the foundation. It's that finance and treasury database yeah you need to have. Um ideally in a single system.
- 10:38
- Unidentified speaker (speaker_2): We offer one of
- 10:39
- Unidentified speaker (UNKNOWN): them
- 10:41
- Unidentified speaker (speaker_2): in for real time information. I think that is so crucial. Uh you need to have that data quality and data integrity to run AI. Yeah. So that's number one.
- 10:53
- Unidentified speaker (speaker_2): Number two is the connectivity. Again, you know, we we want to have the best connectivity possible. AI can't build that uh connectivity. If there is no connectivity, there is no AI between banks and EMP system.
- 11:09
- Unidentified speaker (UNKNOWN): Uh
- 11:09
- Unidentified speaker (speaker_2): the third one,
- 11:10
- Unidentified speaker (UNKNOWN): uh,
- 11:11
- Unidentified speaker (speaker_2): in terms of foundation is around the treasury policy.
- 11:15
- Unidentified speaker (UNKNOWN): Now,
- 11:15
- Unidentified speaker (speaker_2): this is a topic, you know, I see in some organizations very detailed, very well done, very mature. In other organizations, it's
- 11:27
- Unidentified speaker (UNKNOWN): in
- 11:28
- Unidentified speaker (speaker_2): the minds of some experts, yeah, but not very good documented.
- 11:33
- Unidentified speaker (speaker_2): That means when you want to hand over this treasury policy.
- 11:39
- Unidentified speaker (speaker_2): to a machine,
- 11:42
- Unidentified speaker (speaker_2): you run
- 11:45
- Unidentified speaker (speaker_2): run into issues. So you need to have your risk parameters clearly defined. You need to have the instruments you want to use for your treasury uh clearly defined and mentioned mention the constraint as well. Yeah. What can you do? What can't you
- 12:00
- Unidentified speaker (UNKNOWN): do?
- 12:01
- Unidentified speaker (speaker_2): And
- 12:03
- Unidentified speaker (speaker_2): you
- 12:04
- Unidentified speaker (speaker_2): want to define all the things around your treasury um uh strategy in detail so that really you give the guardrails to the to the machine to the artificial intelligence. If you don't do that, then you know it can be can be a problem, as other examples
- 12:24
- Unidentified speaker (speaker_2): have shown recently, which are
- 12:26
- Unidentified speaker (UNKNOWN): uh
- 12:27
- Unidentified speaker (speaker_2): talked about in the press. Yeah, so we want to have these guardrails fixed. And then fourth topic: AI governance.
- 12:34
- Unidentified speaker (UNKNOWN): You
- 12:34
- Unidentified speaker (speaker_2): need to have an AI governance model in place that means workflows, clear role definition of who does what, and audit, audit trail, and how how do you deal with decisions
- 12:49
- Unidentified speaker (UNKNOWN): uh
- 12:49
- Unidentified speaker (speaker_2): across your enterprise. that are not covered by the treasury policy?
- 12:54
- Unidentified speaker (UNKNOWN): Yeah,
- 12:54
- Unidentified speaker (speaker_2): so
- 12:55
- Unidentified speaker (speaker_3): sort of uh things like what
- 12:56
- Unidentified speaker (speaker_1): thresholds or mission.
- 12:57
- Unidentified speaker (speaker_2): Exactly
- 12:58
- Unidentified speaker (speaker_3): if you have an
- 12:59
- Unidentified speaker (speaker_1): AI agent taking action, where do you need humid review and approvals and those types of things? Exactly. Really, all you're saying is if you go back to the basics of what
- 13:07
- Unidentified speaker (speaker_1): SAP as an ERP provides, you think about the data the bank delivers over secure connection, I think these are foundations how you HSBC and SAB can offer clients today that helps them get ready for AI. And it sounds like you know, data is one thing, but of course, leveraging the you know workflow, the controls you have within the SAP platform. We also
- 13:27
- Unidentified speaker (speaker_1): offer with our new model context protocol options for clients to use their entitlements within the bank portal uh if they don't have that same structure on their side. So I think we've seen clients of different levels. of readiness, but I completely agree with you in terms of the data integrity and the security, the controls, the workflows, but then of course that
- 13:47
- Unidentified speaker (speaker_1): AI observability as well. How do we make sure the AI model is operating as we expect, which model was used, how would it reason? You know, did the right approval happen Will and so on. And I think that's the difference with AI over a normal technical implementation. You know, you would definitely test something before you deploy it. With AI, you don't just test
- 14:07
- Unidentified speaker (speaker_1): before you deploy, you need to continue to be sort of testing and observing as well. So I think I agree that AI governance would be quite new for customers to be thinking about.
- 14:17
- Unidentified speaker (speaker_2): Yeah, and you know.
- 14:20
- Unidentified speaker (speaker_2): Customers shouldn't think about AI as a separate project. Yeah. AI is part of your treasury workflow at the end of the
- 14:29
- Unidentified speaker (UNKNOWN): day.
- 14:30
- Unidentified speaker (speaker_2): And you need obviously you should reuse what you what you can reuse from your existing workflows and from your existing data. But it might not be good enough. Yeah. And this
- 14:43
- Unidentified speaker (speaker_2): looking at uh AI, the first thing you need to have in place is the data and the controls. If you don't have these two things in place, you better do AI assisted treasury.
- 14:56
- Unidentified speaker (speaker_1): Yep,
- 14:56
- Unidentified speaker (UNKNOWN): yep,
- 14:56
- Unidentified speaker (speaker_1): absolutely. Yeah. Agree. So thinking a little bit around practical value in treasuries. We spoke a little bit around getting ready for AI, some of the foundations. Again, both of us well placed to support our clients in that. Um, but if a treasurer is sort of thinking about where do I prioritize those use cases? you know, where do I start? We
- 15:16
- Unidentified speaker (speaker_1): like to say, you know, start with the smallest thing possible and build from there. Where do you see the role of AI playing? I think we've both seen cases of treasurers using AI to interpret information, process large amounts of historical data, maybe do forecasting and so on. Um, But are you seeing any clients sort of willing to move to using the their
- 15:36
- Unidentified speaker (speaker_1): AI to take action? What's your kind of observation? in speaking with clients around practical use and value in the treasury setting?
- 15:44
- Unidentified speaker (speaker_2): Yeah, I I see customers uh using it to take action. So turn that insight into action. Um, you know,
- 15:51
- Unidentified speaker (UNKNOWN): they
- 15:52
- Unidentified speaker (speaker_2): use AI to analyze, you know, the intraday positions and then, you know, they they look obviously as well, uh, what instruments can be taken to address the this shortfall on the cash side for uh for example. And then um, you know, you you look at the market conditions and you do that all automatically with AI,
- 16:12
- Unidentified speaker (speaker_2): and obviously the bank data, which is instrumental uh to to to do the AI.
- 16:19
- Unidentified speaker (speaker_1): Got a complete picture. I also like to say you've got a huge amount of internal context that sits in the ERP. The bank obviously gives you a bank context as you say, you also bring in the market data. So you have a multi sort of faceted context to in order to enhance the decisions.
- 16:33
- Unidentified speaker (speaker_2): So so and then you can come up with recommendations. For example, uh say I move money from account A to account B to optimize my yield. So that's one form of how AI is already used. The other the other uh uh area is workflow initiation that gets automatically
- 16:53
- Unidentified speaker (speaker_2): initiated based on the Treasury policy and predefined processes and approvals that obviously need to happen. That, for example, you say there is a shortfall, or there is a certain FX limit um exceeded. Yeah. I need to I need to do
- 17:13
- Unidentified speaker (speaker_2): f FX uh FXH to address this. So these things can be automated in advance when certain limits are exceeded and then can be executed automatically. That's for me AI as well, based on real-time information. And then you can have also AI in a more end-to-end orchestration. Yeah.
- 17:33
- Unidentified speaker (speaker_2): Uh Let's think about cash concentration for example or intercompany settlements. So, which you know is uh multiple workflow flows in a in a row. So if you have the right governance in place, if you have the human in the loop or the human on the loop, yeah, which I still see as
- 17:53
- Unidentified speaker (speaker_2): default, by the way, when I talk about AI. So what what I experienced is not everything runs automatically and no human looks at it.
- 18:05
- Unidentified speaker (speaker_2): We have the humans supervising what the agents do. And you know, once we have the assurance this is actually the right outcome, what the agents say, and this is repeatable, then we can scale it.
- 18:21
- Unidentified speaker (speaker_1): Yep. No, it makes absolute sense. And I think if you go back to that real real-time time data delivery, it becomes really important for that agent to be able to act on the latest information. But I think the other bet advantage is pulling together information, like you said, from different sources, being able to do that really quickly and allow treasurers to take a much faster path to action while, you know, sort of reviewing the
- 18:41
- Unidentified speaker (speaker_1): output. But I think what will be interesting as those policies and controls improve to our earlier point on the machine readable, it'll give a more safeguards in order for agents to potentially take action with without sort of someone always reviewing and checking, but operating within a more deterministic outcome. So you can be very specific with what you uh what you would
- 19:01
- Unidentified speaker (speaker_1): want to see.
- 19:02
- Unidentified speaker (speaker_2): And this is a very good topic. I think what what we're seeing now in applying AI in treasury is that the role of a treasurer changes. Yeah. So from purely execution. executing tasks and doing research, yeah, to monitoring
- 19:20
- Unidentified speaker (UNKNOWN): uh
- 19:21
- Unidentified speaker (speaker_2): the the agent, defining the policies,
- 19:25
- Unidentified speaker (UNKNOWN): and
- 19:25
- Unidentified speaker (speaker_2): doing the strategic judgment. Yeah. So we need we need from a from a profile, we need treasurers that can judge if the outcome of the agent is correct or not. Yeah. That's very important. Yeah. So that means as well that just
- 19:45
- Unidentified speaker (speaker_2): executing task is not good enough in an AI world. You have that you need to have that elevated expert profile and train people accordingly so that they know that this output from the agent is correct.
- 20:00
- Unidentified speaker (UNKNOWN): Do
- 20:01
- Unidentified speaker (speaker_1): you think there's a risk of uh people becoming a bit complacent in really checking the output of the agent?
- 20:08
- Unidentified speaker (speaker_2): I don't think so. I just see more transaction volume.
- 20:12
- Unidentified speaker (UNKNOWN): Okay.
- 20:13
- Unidentified speaker (UNKNOWN): Okay.
- 20:15
- Unidentified speaker (speaker_1): Um okay, excellent. So I think uh just to sum up today, thank you so much for uh being here, folk. Uh real-time treasury is not simply about moving information faster, it's about using trusted data. that the bank provides into the enterprise context that SAP is supporting and implementing effective controls on both sides for
- 20:34
- Unidentified speaker (speaker_1): clients. So they're able to turn an event into the right action much faster. I think we're saying AI is definitely here. It is being used and definitely more to come. And it's going to help our treasurers further, you know, enhance uh what they do and as you say, move into a more strategic role with. within the organization, taking out the noise from operational tasks and
- 20:54
- Unidentified speaker (speaker_1): execution uh that that is there. Um, so thank you so much. Very exciting times. Um, and thanks for being here today.
- 21:01
- Unidentified speaker (speaker_2): Thank you, Ember. Thanks for having me.
- 21:03
- Unidentified speaker (speaker_3): Our pleasure.
- 21:04
- Unidentified speaker (speaker_2): Thank you.
- 21:05
- Unidentified speaker (speaker_0): Thank you for joining us at HSBC Global Viewpoint. We hope you enjoyed the discussion. Make sure you're subscribed to stay up to date with new episodes.
More from HSBC Global Viewpoint
Asia beyond the headlines
Episode discussion: HSBC analysts challenge Western narratives on China's growth and Asia's structural shifts
Speakers’ claims: Piers Butler hosts Jing Liu and Ines Lam to discuss Asia's economic trajectory beyond mainstream headlines. Jing Liu argues that China's 4-5% growth remains significant for its size, driven by a transition to high-quality growth where high-tech sectors now exceed the property sector's peak GDP contribution. Ines Lam highlights regional challenges including demographic aging and AI-driven labor displacement risks. The speakers emphasize interconnected drivers across Asia, such as fiscal health and industrial restructuring, while noting China's strategic avoidance of large-scale stimulus due to local debt concerns.
Under the Banyan Tree - Live from Hong Kong 2026
Episode discussion: HSBC analysts discuss AI capex risks, currency outlooks, and US-China tech divergence at Hong Kong event
Speakers’ claims: Fred Neumann, Paul Mackel, Yuqian Ding, and Frank Lee discuss market volatility, noting Nasdaq declines amid rising bond yields and tax concerns. Mackel attributes US dollar strength to Federal Reserve rate hikes and equity inflows, while highlighting RMB outperformance and slow yen recovery. Lee projects US hyperscaler capex could double to $1.5 trillion, raising debt-to-equity ratios but remaining manageable due to high AI profit margins. Ding contrasts US frontier performance spending with China's efficiency-focused approach, noting similar LLM outcomes. Lee identifies optics as a key supply chain overlap where China holds 70% market share, while Van der Linde concludes with optimism on long-term AI investment potential.