Sibos in Frankfurt brought together more than 12,500 attendees to connect, network and take stock of market direction. It involved firms from all across financial services - banks, fintechs, financial market infrastructures (FMIs) and service providers - with a programme of dynamic discussions covering a wide range of new and emerging themes.
Here were 5 things that really stuck out to me.
1. AI Agents aren’t coming: They are already here
Unsurprisingly, Agentic AI took centre stage this year at Sibos, dominating much of the discussion on panels, conference stages and across the floor.
What stood out consistently was how rapid adoption is made possible by automating specific tasks carried out by people, and leveraging governance and control frameworks that are already in place.
Though a very obvious statement to make, this became a light-bulb moment for many who had been grappling with how to work with internal risk and audit teams on addressing the perceived risk of a rogue agent or a bad actor.
It’s very much a matter of framing process orchestration (how agents and people work together) alongside the governance and auditability that is in place right now. Fast adoption therefore becomes more about solution design and tool selection, rather than long tenured implementation programmes.
2. Tokenisation moves mainstream
Many vendors I spoke to are now showcasing real use cases with well known financial institutions where asset tokenisation has become widely embedded.
Whilst stablecoins were a popular discussion topic - certainly relating to wholesale payments - what seemed a very common use case actually related to the increasing usage of tokenised money market funds.
These are becoming increasingly attractive as a cash management tool, where treasurers are looking to deploy them as an alternative to processes such as end-of-day sweeps. The real time settlement aspect also removes fails and settlement risk.
Whilst wholesale businesses slowly increase their interest in crypto assets, it is within tokenisation where the action seems to be hotting up quickly. Expect to see increased adoption of tokenised assets underpinning traditional cash and collateral management processes over the next 12 months.
Momentum is building fast in this area - if you are not thinking about this as treasurer, cash manager or collateral manager, you should be!
3. DeFi vs TradFi
Linked to the above theme is the increase in complexity businesses are facing where they are trying to bridge assets across two sets of infrastructures. This means understanding how a tokenised asset on-chain is underpinned by a fiat asset off-chain.
Alignment of where the asset is, its price and its liquidity becomes increasingly difficult for institutions to manage - and will add infrastructure cost as a result.
One FX business head I spoke to framed it like this: “You can have instant or real-time settlement, but it is actually quite costly to do that. Clients want settlement to be faster, but they also want it to be cheaper - which doesn’t work”.
And whilst much of the historic narrative around decentralised finance (DeFi) has been the cost savings and simplicity it brings, the reality appears to be that having to settle and service assets across two quite different infrastructures potentially adds more cost.
Some of the initial narrative around DeFi bringing a ‘zero ops’ or a ‘zero reconciliations’ operating model seems increasingly far-fetched - and it's actually driving a greater need for reconciliations and data integrity checks.
This blending of DeFi and traditional finance (TradFi) has been a repeat theme at industry events this year. Expect an ongoing discussion heading into 2026 and beyond around the potential headaches it creates for heads of Operations and COOs.
4. Legacy leaving leaders behind
AI adoption is no longer a nice-to-have in your estate. It is a must-have for any vendor or internal IT team. The likes of big-tech - AWS, Microsoft, Google - are investing significant amounts into their toolkits and making available AI accelerators such as Bedrock.
These allow innovation to happen much faster, but do also require some basic architecture fundamentals such as modern APIs, containerisation and orchestration to make them work. And, most fundamentally of all, require data to be available and accessible to allow models to be trained.
Where architecture doesn’t conform to these standards - and given much of the industry is still running on software built 20+ years ago, a lot of it doesn’t - this creates a major blocker to progression. Modernising architecture (or refactoring an application to create a cloud-native version) is time consuming and expensive.
This means that many organisations are unable to access the full capability that AI is now able to deliver - limiting innovation to often simple use cases that don’t move the needle.
This is not a trivial problem to solve. Whilst technology architects are trying to figure out how to blend the old with the new, some of their peers are able to make transformational leaps in capability through its adoption.
And whilst there are some use cases which can be attacked outside of the core architecture, these are low level in comparison to the value that Agentic AI can potentially bring across middle and back offices.
A successful AI strategy needs modern technology and architecture to enable it. If you don’t have this in your stack you will increasingly be left behind.
5. Agentic empathy
No, this isn’t around ensuring you are thinking about your Agent's feelings - this is about ensuring your transformation programme understands the very personal nature of Agentic AI and what this means to your organisation.
Agents are no longer solely complimentary to your teams - helping and assisting their day-to-day roles - they are now in direct competition with people working in traditional Operations and Finance roles.
Some software-as-a-service (SaaS) vendors are now talking about ‘job-scription’ pricing models, deploying teams of agents at a fraction of the cost of the equivalent people in their roles. This is no longer simply a seat based pricing model, it is a role based one…
A recent - and well publicised - report by MIT highlighted that only 5% of AI projects succeed. Whilst this may seem a staggering statistic (even taking into account fail-fast projects and those where change incompetence - not AI - was the cause), it is perhaps not surprising when considering the very personal impact AI can have on operating models.
Not painting a vision of the future - one where process orchestration between humans and agents can be clearly visualised - is unlikely to create buy-in. And driving transformation against a backdrop of fear of change is never going to bring a good outcome.
So whilst a potentially game changing innovation - indeed, perhaps the iPhone moment for the world of post-trade - we shouldn’t forget technology is only ever an enabler of transformation. It is never solely transformational itself.
This people part is probably more critical to the innovation outcome than it has ever been, and organisations who fail to recognise this - and empathise - will in all likelihood end up in the 95%.
Innovation can’t happen in a vacuum
And so we wrap up a great Sibos. Innovation is moving fast, but many of the talking points this raises have been key themes for a while now. The disconnect between modern and legacy technology, questions around integrating new and traditional finance, questions around governance and the trials of transformation persist.
Innovation, no matter how game-changing it may be, can’t happen in a vacuum.
We’ll be bringing you more insights and takeaways from Sibos over the coming weeks. Subscribe to our newsletter to make sure you don’t miss out.
