← 返回

2026.10.07 | TB | Fintech | AI in Core Banking Migration

原标题：AI in Core Banking – Migration Timelines & Monetisation in Temenos, Thought Machine, Finastra & Avaloq

#financials #software

Third Bridge · 2026-10-07

Analyst-led • 22 Sep 2026

Companies: MSFT US, TEMN CH, LLOY GB, ORCL US, INFY IN, FISV US, FIS US

Mohammad Mohtashim Khan (MMK) Managing Director & Head, Private Banking & Wealth Management Operations at NatWest Group plc

Philip Atkinson (PA) — Third Bridge Sector Analyst

Agenda

Key Insights

Introduction

PA [00:00:04]: Welcome to Third Bridge's Interview entitled AI in Core Banking – Migration Timelines & Monetisation in Temenos, Thought Machine, Finastra & Avaloq. I'm Philip Atkinson, and I'm here with Mohammad Mohtashim Khan, Managing Director and Head, Private Banking and Wealth Management Operations at NatWest Group.

Before we get started, could you please state either I agree or I disagree to the following statement: You understand the definition of material non-public information and agree not to disclose any such information, or any other information which is confidential, during this Interview.

MMK [00:00:47]: I agree.

PA [00:00:48]: Could you share an introduction to your background, how you have been interacting with core banking vendors and AI and what your role encompasses?

MMK [00:01:04]: As you've pointed out, I work for NatWest Group, specifically within the private banking and wealth management division. I joined the bank in 2018, but prior to that, I had a very long career in management consulting, working for EY and Capgemini, working with a lot of the leading private banks and retail banks, helping them with technology and strategy. My first five years in NatWest was spent on doing the same thing, essentially working on strategic projects, strategy and transformation and how we accelerate the bank's digital and technology transformation. The last three years has been specifically focused on operations and trying to modernise operations, not only from a location strategy perspective, so ie, building capabilities in different parts of the world, but also driving automation, driving simplicity and driving the adoption of AI and AI-enabled tooling.

The way that I have interacted with core banking is in my previous roles, running projects, which we were looking at core modernisation or digital platform modernisation, but more recently, a lot of my focus is on agentic operations and how we modernise the way that my team interacts. They're primarily Avaloq users, which is our core banking system, but they also use the satellite systems that are in and around Avaloq, things like Finastra Total Messaging, FIS IntelliMatch, PayUK platforms for payments, Fiserv and TSYS for cards and also some of the digital platforms we use, including Backbase, Microsoft Dynamics and Additiv. Essentially, a lot of my job, even though it's operations, is about technology and how we take a step forward on technology to simplify the end-to-end customer journey, which involves my operational teams as well.

Questions

00:03:06] What would be your top three priorities for banking IT spend allocation over the next 2-3 years?

MMK [00:03:26]: If I boil it down, my top three priorities are resilience and regulatory change, so to make sure that the service is really safe and secure. Data and AI foundations, in lieu of my role, I'm also the Chief Data Officer from a business perspective, so I'm responsible for data management. The third thing will be simplifying the client and advisor journey. Essentially, the third one is enabled by the first two things. First, around resilience and regulation. This isn't really discretionary. We run a best-of-breed architecture. As I said, we've got Avaloq at the core and then a number of satellite systems that support specific capabilities. The risk really lies within the integration points, not in a single system. Operational resilience means that we have to evidence and that we can stay within impact tolerances across the whole chain. Add the complexities of moving to an ISO 20022 structured data changes on SWIFT, regulation changes around consumer duty, outcomes monitoring and cybersecurity. You have a floor of spend that really comes off the top before anyone talks about innovation.

Second is data and AI foundations. Most banks have learned that AI is only as good as the data underneath it. Avaloq is our golden source. It's got a rich set of book of records, positions, transactions, but it isn't really designed to serve real-time govern data to models and agent. It is really quite complex. We are investing heavily in the data layer, building data products and data capabilities, which help us do things like event streaming off the core, data lineage, entitlements and then also a lot of stuff which is not really apparent around the usage of AI, like building model risk and governance tooling and also investing in core capabilities, things like MCP servers, which can help us accelerate the scale-up of AI.

The third element, which is really, as I said, enabled by the first two is essentially within the private banking and wealth management construct, making life easy for the relationship manager and the customers, investing in orchestration for CRM and workflow for client portals, where we're looking at the onboarding journey, KYC journey, suitability and advice journeys and actually enriching and stitching together the platforms in a better way, which builds capacity not only for operations, but also for advisors and relationship managers. Essentially, those are the three main things. There's margin pressure, which means cost to income is important. Therefore, simplification is really driving the agenda for as well as building a great digital experience so we can acquire younger customers.

00:06:42] What is your outlook for budget discussions? What are the top contention points when planning for 2027 as we head into Q4 2026?

MMK [00:06:58]: We are right in the midst of budget planning at the moment, and I expect Q4 to be tighter and much more forensic than previous years. I'm not necessarily seeing a cut in budget here, but a harder scrutiny of how value is delivered. If I look at the main points of contention, there are actually four of them. The first, run vs change. Run costs keep creeping due to vendor uplifts that are often inflation-linked, cloud consumption, cyber tooling and licences that scale with accounts or assets, so AUM-based fees. Every pound of creep squeezes the change budget. The second area of contention is mandatory vs discretionary. Regulatory and resilience items get ring-fenced, which leaves smaller and more contested discretionary pots. Essentially, you're trying to prioritise what is more important vs the other, and it's a difficult debate. The third is in-flight programmes. Obviously, a lot of the modernisation programmes we have around CRM and orchestration, they are multi-year, but in order to get funding for future years, you really need to prove what's the delivery, what's the benefit? That links with the fourth point, which is benefits realisation.

CFO office wants savings to come out of the P&L, not just capacity release, so there's a lot more scrutiny. Now if you take AI specifically, it's been a priority, and it will remain a priority, but I think the tone around that has really shifted. It's now showing the value of AI and the unit economics rather than experimenting with things. It's less about POCs. It's more about can we scale something? Consumption costs. We've built models where they're more visible. You know what tokens and seat licences we are paying for things like Microsoft Copilot. We know what we're paying for the hyperscalers for providing us access to capabilities. What's really being now looked at is what is the actual value of it, cost per case handled, cost per document processed or per seat licence. As AI funding is expanding, it's being much more outcome focused. Also, I think what's changing is previously, AI was a centre of excellence. Now it's moving to the business, which I think is a healthy change because it really forces people to think about the value we're adding rather than using a central part of money to actually experiment with.

Now overall, I see budgets still going up modestly in 2027 across the industry, whether that's within the UK or Europe, but in real terms, I would say they're probably flat because most of that growth is going to come from inflation or a CPI-based vendor uplift, but I think the mix within that is going to change. I think there's probably going to be a lot more investment in AI and data. Security and resilience will go up, while legacy run and discretionary change will be squeezed. There is less of a cushion for it because of the interest rate environment and the pressure to pass on the interest rate benefits to the consumers. I think it will be a tough year 2027, a year where we'll need to prioritise.

00:10:50] You mentioned initiatives around SWIFT [Society for Worldwide Interbank Financial Telecommunication] and the new ISO [International Organisation for Standardisation] standards. Would you say the ring-fenced regulatory budget is increasing compared to prior years?

MMK [00:11:17]: Yes. It is increasing prior [compared] to previous years. That's mainly driven by the amount of regulation that has been loaded on over the last two years, specifically within consumer duty and the payment space, but not only on that, but also if you look at wealth business, it is getting more complicated to deliver advice and much more expensive. Therefore, investment is going on into platforms that bring us to a much more sustainable model.

PA [00:11:46]: Would you expect those elevated costs to persist over the next 3-5 years, or are these a near-term pressure to implement an update?

MMK [00:00:56]: I think the pressure is higher near term. It will ease, but it will not go away in the next 3-5 years. It's going to be a consistent theme. As more and more fintechs enter the market, there's going to be more and more scrutiny around how banks and fintechs, different operators work. Therefore, I would still see it persisting for a period of time.

00:12:29] You mentioned that you expect legacy run costs to be squeezed. Could you walk us through that process? What might be targeted for that squeeze?

MMK [00:12:45]: I think a lot of it is going to be things on legacy infrastructure. As we go towards cloud modernisation and also platform modernisation, a lot of the spend is going to shift. If you're thinking specifically about core banking, we are now at a phase where core banking is quite stable. It meets the need of the regulation. We're probably going to look to maintain spend around core banking or try to squeeze that spend, and then move the spend on the things that sit around it, for example, the digital layer or the AI and data layer, which means from a vendor perspective, if you say Avaloq, and you were looking to build core capabilities within Avaloq to help us manage our workflows and you were making, say, X number of pounds on an annual basis, what that means is that from '27 onwards or even in '26, you're making X minus N because we're doing that investment ourselves in the tertiary systems rather than enhancing the core. We're keeping the core stable.

PA [00:14:01]: Do you expect banks to in-house some of these core functionalities?

MMK [00:14:12]: Exactly. I think what AI has enabled us to do is it's enabled us to code and build capabilities in-house in a much more scalable way, which we weren't previously able to do. Essentially, what that helps with is that the impact of that is that, say, if you were to take the vendor stack, their licence fees are pretty stable, and I think they'll continue to grow with inflation, but the development and the professional services fees are the bits where they will feel a bit of pressure because I think we are now better capable of doing some of that stuff internally than we were probably about 4-5 years ago. The one area in AI where as a bank, we have made significant improvements is our ability to code and actually do development in-house. That will really come into the mix in the next few years when we think about budgets.

PA [00:15:17]: Do you think that will be a theme across the industry, and therefore across all similar banks?

MMK [00:15:26]: Yes. It's already a theme if you're, say, not focusing on a particular name, but if you pick up some of the big fintechs, they are already vendor agnostic when it comes to core development.

00:15:40] What specific functionalities do you think can be most easily moved in-house?

MMK [00:15:55]: I think it's less about the functionality. Let's take it, the first thing that you're moving in-house is actually the ability to build and enhance the product. If you take Avaloq, you needed PL/SQL developers, you needed Avaloq expertise and a lot of that expertise was concentrated essentially in Switzerland mainly, parts of the UK, but mainly in Switzerland. The unit cost economics around getting a Swiss developer to come and build that was pretty high, because you need specific capabilities. What AI has enabled us to do is that PL/SQL coding, that programming and that specific domain expertise is now much more easily done in-house. You don't need to hire somebody from a Synpulse in Switzerland to be able to come in as a contingent worker and charge GBP 1,400, GBP 1,500 per day. That's the first element. The second element that is being in-housed is now that we have a stable core, what's happened is that if you look at the legacy development of core banking, in the 2010s, when we adopted Avaloq, it was all about having a consolidated system, a monolithic block that did everything.

Once we achieved that utopia and simplified the architecture by having one core banking system, which provided multiple capabilities from CRM to lending journeys to credit card journeys, the issue that arose was that it was very difficult to decompose that and build the digital layer on top of it. The next phase was between probably 2020 and 2026 has been focused on that digital orchestration, whereby we've decomposed capabilities and built an ecosystem of platforms such as Backbase or Additiv or Dynamics where a lot of the orchestration of the journeys is happening. Now the third phase is essentially building an overlay on top of all of these solutions to be able to build enhancements to the overall experience, both from an advisor and from an operations and a customer perspective, which brings all of these systems together. Previously, that was difficult because we didn't have API layers. Now we have got API layers with the advent of MCP and AI and the ability to contextualise data and information, which is not exactly complete and take action on it, ie, agentification of the architecture, I think that's the next five years. That's where I think a lot of the change is going to happen and the area where we're probably going to invest in.

00:18:51] Regarding core banking spend with vendors such as Avaloq, you mentioned that licence fees will remain stable and grow with inflation, while development and professional services will face pressure from in-housing. Could you break down the typical annual spend? Out of a GBP 100 spend, how much goes to the licence vs development and professional services?

MMK [00:01:11]: If I were to take a broader view, essentially, if you look at IT spend as 100%, roughly about 35-40% of our total tech spend sits on core and adjacent software. You have to include the adjacent software. These are the elements that make it work together. This covers licences and also the change book. Today, probably you could split as 50/50. 50% of that is your change spend, application maintenance and application development, and the remaining is licences. Then if you look at the second element of it, you've got your workflow systems or your front office systems, that's about 15-20% of your total spend. Then you've got your messaging platforms for your payments. Again, that's about 15%. Then data and AI probably today sits at 10% of the total. Then you've got a smattering of other things like financial crime systems, compliance tooling, card processing platforms, reconciliation, which makes up the remaining 30% or so. Now if you are working in a tier 1 bank like I am, a lot of that spend will be done in-house. You will have your mainframe core or you will have your vendors like your TCS and FIS or Infosys. Those are in the mix.

Essentially, over the next 2-3 years, I expect the core share to drift downwards, and the front office to go up. Essentially, to answer your question, at the moment, if you've got, say, a total spend of GBP 100, of that, roughly about 20% of it is going on core licensing subscriptions and 20% is going on development, a part of which will go to a vendor and a part of which we'll do ourselves. It's not completely 100%. It's very difficult to break down what we're paying the vendor vs us. That's 40% spend just on core banking. The remaining 60% is other things, including your front office platforms, payments, data, card processing, fin crime, recs, those sort of things.

PA [00:22:03]: 20% of that 40% spend on core banking is going to come under pressure. Am I following that correctly?

MMK [00:22:14]: Yes. The 20% that's licensing is probably going to be okay, but a portion of the 20% that is where we subscribe on services, and tooling, connectors and things like that that's where I think they're probably more at risk. Because the licence you need, in a regulated business, there is a requirement to have a core banking system, but as I said, that core is now coming to a point where it's stable where you're not developing it. If you're not developing it and you're developing in and around it, then the spend for the vendor is naturally going to go down.

00:22:55] What is your outlook for the bill for core banking vendors such as Avaloq or Finastra? Do you see spend plateauing as in-housing and inflation offset each other, or do you expect downwards pressure from the shift in-house?

MMK [00:23:15]: I think there's downward pressure. If I were to say that today, it's 40% spent on core banking, I see that in the next 2-3 years coming down to 35% and then 30% in five years' time. There will be an equivalent increase in the front office and the data and AI platforms.

00:23:40] You mentioned that budget allocation for AI, data and front office will increase. What does this mean for new AI initiative launches? Do you expect a trial period for new vendors, or do you see a consolidation around specific vendors to scale existing proofs of concept?

MMK [00:24:22]: I think there are three areas where AI will grow. There are certain domain areas where I think we will look to actually get AI-specific vendors. A good example of that is financial crime and fraud, where there is quite a mature use of machine learning. If you take Featurespace, for example, which I think now belongs to Visa, it is a payment filtering platform, and they have their own proprietary data models. We don't want to go and develop those models ourselves. I think there will be an aim to look at specialist AI providers who can give us those capabilities, which we don't really see as strategic, but they are essential to keep the bank safe and they are a hygiene factor for any banking organisation. Then you will have a second set of AI features that are going to come from the core banking providers. Those features are going to be most likely features that enable us to better understand the data that is within those particular systems. For example, Avaloq might come up with a enhanced view of tax reconciliations, which helps with the tax reporting module within Avaloq. I think that will be a limited amount of capability or a limited amount of value that we will look to see.

A majority of AI development, I see being done directly by the banks using their partnerships with the hyperscalers. That could be in areas such as operations and document intelligence or advisor platforms. Then you will have any of the client servicing journeys that you're doing. Now underneath that lies what I would call your copilot infrastructure. This is where AI is not the agent, but AI is an assistant.

Most companies will partner either with a large player like Microsoft, which is, I think, the predominant player here, but you also see that Anthropic and OpenAI also have their copilot propositions. These are basically your chat solutions, which help you manage your diary and you can ask them a question or format your e-mail, things like that. That's more around the productivity layer. Productivity layer, you're picking off the shelf with a large player, your financial crime fraud, you're getting specialist solution. You're getting some solutions from the core providers where it's really investigating and managing data within the core, but a large chunk of the development I see happening with the hyperscalers and the banks working in partnership, building the AI solutions themselves.

00:28:22] You mentioned that core modernisation and migration are ongoing topics. How do you think modernisation is viewed today, particularly with the advent of generative AI? Do you still see this as a priority, or will spend move towards those AI initiatives?

MMK [00:28:59]: Modernisation is viewed very differently from what it was like 10 years ago. As I mentioned, 10 years ago, we started with the view that we're going to have one monolithic platform that is going to bring simplicity into the architecture, and we're going to do everything on that platform. The challenge with that is that you tend to get a lot of risk with that in a large organisation and core banking replacement platform projects are infamously complex. Today, what's really looking at it is people are moving from replacing the platform to upgrading the capability continuously. I think what gen AI does is it reinforces that shift. Previously, the default was big bang, rip the legacy core, replace it with a next-gen platform, accept a 5-7-year programme with serious risk. There are a lot of use cases now of those programmes running over budget, causing outages and therefore, then being paused. The biggest example of that is what happened to TSB a few years ago when Sabadell were trying to move them from the Lloyds platform onto their own platform, which was catastrophic for the business.

Today, the view is about progressive modernisation. You keep a stable regulated core, you wrap it with APIs and events, modernise the capability by capability. You take, say, for example, onboarding or payments and then you add best-of-breed components. We realised for private banking and wealth, Avaloq's depth in portfolio management, custody, multi-jurisdiction booking was really hard to replicate. The case for replacing wasn't really there. The better return was actually to focus on modernising by building the ecosystem of systems, which helped us orchestrate client journeys, so things like Microsoft Dynamics or Backbase and to improve the client advisor experience much more quickly than to spend seven years. Now if you take gen AI, the three main impacts that it will have, first, it lowers the cost of understanding the legacy. Much of any migration cost is reverse engineering. Understanding what the old system does, so the code, the parameters, undocumented rules, gen AI is really good at code compression, documentation, the test generation, data mapping, which makes the migrations possibly less risky and cheaper.

Secondly, it reduces the pressure to migrate at all. If you can put an intelligent layer on top of the core, an agent that reads data, triggers workflow and briefs the advisor, then much of the core that's holding us back, that argument weakens. The user experience is no longer bound by using green screens. Third, it really raises the bar on data. AI exposes poor data quality very quickly. Modernisation is being reframed as a data problem rather than an application problem. Net effect, I think, is that gen AI is an accelerator for the engineering work and a reason to be more selective about what you actually replace. When boards want modernisation that delivers value, they're not really thinking about ripping up everything. They're thinking about how do I get that value as quickly as possible on a quarterly basis rather than at the end of a five-year period.

00:32:53] You mentioned that private banking's migration story was less convincing because of the depth that Avaloq had. What do you think about retail or corporate banking with those less specific use cases? Could traditional migration stories be more convincing for the core banking solutions?

MMK [00:33:20]: They are not ubiquitously beneficial where you have got large tier 1 banks or tier 2 banks, which are established. I think most of them will go with the model that I've explained, which is rather than replacing the platform, you upgrade capability continuously, but if you are a fintech or a start-up who wants to accelerate the journey, the likelihood is that you will do a full core replacement or if you have a system which is really end of life. Now I don't see any examples of that in the UK, but obviously, other parts of the world, you have got within Europe, say, eastern and central Europe, you've got core banking systems, which are from the '70s or '80s haven't been modernised. Then there is a business case because what you can't really do is build the overlay. You don't have that issue with Avaloq because you can build APIs and they're building an MCP gateway. It's a pretty modern architecture, but where the architecture is inhibiting, you may want to do core replacement, but that is definitely less prevalent than the model that I've explained where you'd want to basically rip everything apart and put in a new one.

A couple of things that happened. We're talking about Thought Machine. For example, JPMorgan and Lloyds have both chosen Thought Machine for different use cases, but what they're doing with it is they're not really ripping out the entire core. They've chosen Thought Machine and they've said, "Okay, we're going to actually transition use case by use case." They don't think it's going to be 100% fit for everything. They've chosen a couple of areas where they strategically want to migrate to Thought Machine and build capability over there. That's a much more likely scenario within the large tier 1 banks where you'll choose a new core banking system, but you're not going to move everything to it. You're going to take a couple of use cases, build it up slowly and see whether there is a good way of managing the transition and the migration.

00:35:48] Do you see the migration revenue growth stories for core banking players such as Temenos as a convincing proposed strategy, or do you see gaps and difficulties from a customer perspective?

MMK [00:36:17]: Let me just say, demand for modernisation hasn't gone away. If anything, it's broader, but its shape has changed. What you're moving towards is smaller units, business-led and increasingly vendors being judged on their ability to improve the data and AI landscape. There are three things I'd like to call out. First, the scope of the project has narrowed. Instead of having one enormous core replacement, banks will run a portfolio of smaller projects, focusing on things like, say, a data platform with a business case of its own that can be stopped or scaled. Secondly, what's happened is the buyer has changed as well. Previously, core banking was the purview of the CIO and now business heads. For example, 10 years ago, somebody in my role would never have been a part of a core banking platform discussion, but today, I co-own the decision because the value case is around driving business productivity. The third thing that you probably need to look at is SaaS and managed services are default consideration. If you keep those bits in mind and the business case that they're building, I think there is still a business case, but it will manifest itself in a slightly different way.

I think different vendors, depending on how they approach it, will have different business cases for it. For example, full core replacement will stay selective, mainly greenfield and as I said, end of life. The other area where you could have greenfield is post-merger consolidation. I think you used the term hollowing out. Tier 1s will keep hollowing out progressively rather than swapping. Secondly, AI readiness will become a formal selection criteria. We've already started evaluating vendors on do you have open APIs, what's your agentic strategy or what model you run, how you govern it? I think the third thing will be built-in AI will be expected. As I said, a small amount of AI use cases we would expect the core vendor to provide, which could be monetised. It covers for some of the loss of income that you have from a scaled migration reduction. Banks will want core vendors to embed things like anomaly detection or intelligent exception handling or natural language queries on positions or automated documentation as a core banking use case. That could lead to extra revenue even though you don't have those large migration use cases, you're doing these smaller projects, but with the smaller projects, you're actually enhancing your proposition and therefore, creating a case for increasing your licence fees, which will offset some of the loss of earnings.

The practical implication I see for this is vendors will include AI as a part of the platform and compete on openness because banks won't really accept AI that works inside the vendors' sandbox. I think it is a difficult time for core banking vendor providers, but at the same time, this is an opportunity for them to rethink about how they monetise their business because I do think that the era of big platform migrations is probably coming to an end where projects are going to be much more focused, but at the same time, you've got the opportunity to land and then expand your presence there if you do a really good job at use case number one. I don't think there's anything stopping them from building 10 extra use cases on the platform.

00:40:27] Which core banking vendors do you see as market leaders or best-positioned to capitalise on AI and monetisation changes for tier 1 and 2 banking customers?

MMK [00:40:50]: I think for tier 1, I think the largest vendor is still in-house because a lot of the large banks has its own mainframe platform like Nationwide, but among the third parties, I think FIS and Fiserv are probably the strongest in the US market and also in parts of Europe, but you've also got the likes of Temenos and Oracle, who are credible players. As I said, in the next generation, you will have Thought Machine, which has some big clients like Chase and Lloyds, but they're mainly focusing new propositions. I think if I were to say it's the in-house and then Temenos and Oracle probably are the three biggest in Europe, followed by the likes of your TCS, Infosys and Thought Machine. When it comes to tier 2, I think probably Temenos, along with the remnants of the old Finastra Fusion business. I know I don't know whether you know or not Finastra and they sold all of their universal banking. I think that they've got a pretty strong proposition. I think they're now owned by private equity, and I think they will continue to do really well.

The other player, again, that is doing really well is Mambu, which is another fintech along with Thought Machine. Now when I look at private banking, then I think Avaloq and Temenos are probably the leaders followed by a couple of niche players like ERI Olympic. I think those are probably the ones that are setting the mark. It's pretty consolidated as a market. You don't have tons and tons of players that are here, but at the same time, the pressure to translate into a market competitive position is difficult at the moment with a lot of… If you take private banking, Temenos and Avaloq are probably between themselves have 60% of the market share, but between them, there's a very limited difference. Then you've got a splattering of other players that are there that provide it, but in total, there are probably about five players that service private banking and wealth. Similarly, retail tier 1, majority of it or 50% of it will be in-house. Then the rest, you'll be using FIS and Fiserv in the US and then Temenos and Oracle in most of Europe.

00:43:36] Do you think vendors such as FIS, Fiserv, Temenos, Oracle, Avaloq and Finastra will be able to translate their current strengths into market share gains and revenue growth, particularly given what you discussed regarding budget priorities and AI?

MMK [00:44:04]: I think if you think about just the market in general, this is moving away from what you would see as a logo win and moving to more of a building strategic partnerships and long-term revenue streams working with the bank. As we said, migration revenues are going to slow down. The way that these vendors are looking at it is basically to try to get into a tier 1 bank, start with one product or one use case and then expand over the years. Rather than thinking about I'm going to get big bang a lot of money, the move is towards building strategic partnerships and going in and actually pitching a low-value product or a proposition, building trust with the client and then actually doing the work. There's an old adage that farming is easier than hunting, and that's the model that they're going towards. The growth pool is very much contested. AI orchestration, data spend, a lot of it is going to in-house teams and hyperscalers.

The core vendors' challenge is actually to go in and actually take a slice of all of this. The vendors that I see really doing really well are the vendors that can combine their technology, so they've got a full proposition. The ones that we mentioned, they've got a pretty complete proposition, so they can do different types of products, vendors. Avaloq has chosen just to focus on wealth management, which is fine because they can deal with the complexities of wealth management, but if you take Temenos in a retail bank, commercial bank or wealth manager, Temenos has a good set of capabilities. They are open to being cloud native. Having a cloud-native SaaS model where you can move cost from a CAPEX to an OPEX, which helps with the current interest rate environment.

The ability of having a platform that has open APIs and they fit within a composable estate and then adding a credible layer of AI to accelerate the migration, those are the players I see winning the share. The people who can say, "Okay, you buy our platform, it's not going to take you five years. We're going to come and we're going to do a project in 18 months. In 18 months, you're going to see a value coming out of your investment. You don't have to take the risk of a full replacement and a seven-year business case." I don't know whether that was clear or not, but what I'm trying to say is propositional completeness, the flexibility to work with a vendor, building a AI and a data proposition that protects some of the revenue from an in-house or a hyperscaler and then accelerating your delivery model into a cloud-native SaaS model, which you can implement quickly, I think that's the key to success.

00:47:34] Regarding specific AI functionalities, what do you think core banking vendors could provide that banks would prefer to adopt vs looking in-house?

MMK [00:47:50]: I think the things that they really need to focus on is the data that sits within the core itself, for example, reconciliations. If you look at Avaloq today, Avaloq is not one system. It's not one monolithic system. My teams have to reconcile data between one part of Avaloq and another part of Avaloq. A question I ask myself is, should I be doing this? Should I be having a 10-person team consolidating data across the same system? We shouldn't be. Anything that sits within the core allows us to better understand what's in the core, do the reconciliation in the core, allow us to extract the data from it, allow us then from a technology perspective to develop and accelerate the migration. When I say AI, it's not only about AI within the system, it's about how you use the AI to accelerate the migration. Do I have data discovery tools that are there? Do I have a AI-enabled migration engine that will be able to accelerate my migration and reduce the pain that a core migration can cause? It's about looking at the full sum of capability within the core, how do I implement it quicker? How do I make sure that the customer finds it easy to use my platform and isn't having to hire extra people just to take data from one part of the platform and paste it into another?

00:49:32] Do you see data clean-up and reconciliation as a one-off reformatting task, or could this be an ongoing service?

MMK [00:49:44]: It's an ongoing thing, because you always need it. You need it for the one-off migration. As I said, I'm thinking that these vendors, it's difficult to sell a big bang project. You're selling a service where you would take smaller bits on over a period of time. You migrate, demonstrate, deliver value, move on to the next use case. Therefore, you have a perpetual revenue stream. Rather than earning your GBP 10m or GBP 20m for migration in year one, you're earning GBP 4m per year, but you're earning it for 20 years because you're slowly enhancing your platforms and giving more capabilities. It becomes more of a strategic partnership rather than I've now sold you the platform, bugger off. If you want to speak to me again, I'm going to charge you GBP 1,500 per day for you to get my consultant in to do the work for you.

PA [00:50:43]: If core vendors start offering reconciliation services, it creates a monetisation revenue stream and secures a longer lifecycle within the bank because it is easier to work with them and build AI on top of their systems. Would you agree?

MMK [00:51:02]: Exactly.

00:51:02] You mentioned a movement towards best-of-breed point solutions, such as for fraud and financial crime. What are the near-to-mid-term priorities where you expect AI functionality to take priority?

MMK [00:51:27]: I think the biggest use case I have for AI is agentic operations. Today, AI is mainly about assistance. We're all implementing copilots that help people do their work, but it's more moving to agents that complete work under tight controls and banks will pay for the outcomes, not just for the table stakes features. What I mean from agentic operations is the ability for end-to-end handling of things like onboarding, KYC refreshes, corporate actions, payment repairs, RX exceptions with humans doing the approving rather than the doing, which means that you need less and less people. The second area would be advisor augmentation at scale. Things like meeting prep, portfolio insights, suitability and consumer duty. Again, core banking vendors are specifically well placed to play a role here, even though the hyperscalers and the likes of Microsoft are trying to win this battle because they've got competing solutions like Dynamics. This is about making sure that pre-meeting prep and personalisation of communications happens quickly.

I think the third thing will be specifically within wealth is personalisation of insights and also AI-assisted execution of things like investing. All of these are in order of priority to major banks at the moment. The fourth area is risk and compliance, so things like real-time surveillance, interpreting regulatory change, control testing. The fifth would be your engineering and change where I think a lot of it has also been already in place. The areas where I think core banking vendors can do quite a lot is the agentic operations. For example, the biggest challenge I have is how to make sure that Avaloq processes are effectively executed with as few people as possible. I think they have a role to play in the third element, which is the personalisation. Making data available in the platform and the portal to enable things like self AI-assisted directed investing, so directly into Avaloq. Avaloq can build those tools and give them to us, because all of the data sits there. The execution happens within Avaloq, but at the moment, I don't see that forthcoming from their part. We are having to work with other third parties to help us with that.

Then I think, again, on the risk and compliance, you could have point solutions, but if it's automated control testing within the core banking system, they should be giving us the tools which allow us to do it automatically. At the moment, we use third-party tools, which means that they're leaving money on the table and integrating a UiPath is much more difficult if they had their own testing solution embedded within the core banking platform, which they can obviously charge extra for. I think the main thing is AI needs to deliver a measurable outcome and the vendor needs to be able to carry accountability for it.

PA [00:54:55]: It sounds as though a lot of the AI functionalities or use cases you mentioned would operate directly within the core banking software workflow. Does this lead you to favour core banking vendors if these solutions are available?

MMK [00:55:10]: Yes. If it's based on core banking data. If it's multisystem, then you build it yourself.

00:55:16] You mentioned that third-party vendors such as UiPath are used because core banking vendors do not offer agentic use cases today. If a core banking vendor introduces this agentic capability in 1-2-years' time, do you think banks would switch away from companies such as UiPath to consolidate under their core vendors, or is it already too late?

MMK [00:55:52]: I think it depends on the capability and how it's priced and how it works. I think it's 50/50 here. I don't think the opportunity is lost for core banking vendors.

PA [00:56:04]: On one side, there's an opportunity for core banking vendors, but there is also a risk that current third-party revenue going to UiPath is not as secure as it might seem.

MMK [00:56:15]: There's a lot happening. Literally, if you look at my inbox, every day, I've got 10, 15, 20 e-mails coming from different vendors in my inbox and on LinkedIn trying to set up calls in the same area. The challenge is I'm not getting the same calls from Temenos and from Avaloq. I think the concern more is do they understand that this is at stake here? That they've got a role to play here because I think they're really focused on building the core capability. They talk about building AI capabilities, but it's a bit slow at the moment, but they've got an opportunity, a window of opportunity really in the next 24-36 months to actually assert their positioning here because they will lose the migration revenue. How do you replace it? This is the natural way of replacing it.

00:57:08] You mentioned that for the vast majority of core banking vendors, spend is going to be decreasing as customers move functionalities in-house. If current trends continue, what would be your outlook for the decrease, and what could that spend look like? Is that a 5-10% decrease annually, or how should we contextualise that based on the functionalities we have today?

MMK [00:57:55]: If I were to take the same baseline of GBP 100 as the base spend, and today, I'm spending 40% of that or 40% of that on core banking, I expect that to reduce to about 35% in three years' time and 30% in five years' time. Now this is the spend that I'm making across the licences, which are probably going to increase a bit because of inflation, and then a reduction in what we pay in application maintenance and development and professional services and also things like connectors where we're using some small services or tools from a core banking provider because they always have extra modules that you pay for. Now that 10% is going to possibly see an increase of an equivalent amount in data integration and AI platforms. I think today, we spent about 10%, but we're probably going to double that in five years' time. 20% of our spend is probably going to be there. Then an increase of probably a 5-7% in three years and probably a 10% increase in, say, front office CRM workflow type of orchestration type of portals.

Now if you look at the reduction of that 10% of revenue that a core banking provider would get, there is an opportunity for them. Already, if you look at, say, Temenos, Temenos has bought Additiv. The reason why they bought Additiv is because Additiv gives them the ability to extend beyond the core into orchestration. Then similarly, if you look at Avaloq, Avaloq have started to build Avaloq AI Studio and they're building their own MCP because they feel that rather than us going to a third party to get an AI or MCP governance layer and a governance platform, we should be able to, as Avaloq users get that. There is a route to replacing that 10% and actually getting a bigger market share. The only thing that is against these core banking vendors is the time required to get there. I don't think their propositions are complete, so they will have to get into acquisition mode to be able to buy the capability and then integrate it quickly before we build everything ourselves.

I think the main challenge here is the time rather than anything else because the hyperscalers, they're promising us the world. Anthropic has come up and said that "We are going to give you a financial planning capability. You don't need to get a third platform. You can use Cowork for your Copilot, but we can give you financial planning using Fable 5.0," whatever new foundational model they're currently using. I think the essence is that the hyperscalers want to come into and the AI companies, they want to come into workflow management. These companies need to get there before and make ingress into the AI capabilities, so protect their revenue share.

PA [01:01:36]: Looking at existing customer spend, the worst case is a 10% decrease in revenue. The best case is maintaining flat revenue by replacing those losses with AI. Is that correct?

MMK [01:01:52]: Yes. AI and also orchestration.

PA [01:01:55]: Thank you so much. It's been an absolute pleasure going through core banking with you, and I think you articulated things really nicely.

MMK [01:02:13]: Thank you very much. Much appreciated.

Transcription ends at 01:02:19 of the recorded material