AI adoption in capital markets has entered a more exacting phase. The early cycle of pilots, productivity tools and isolated use cases is giving way to questions of operating model, governance, data architecture, control evidence and return on investment.
A-Team Group’s AI in Capital Markets Summit London 2026 will examine that shift through a practical agenda focused on moving AI from experimentation into controlled enterprise deployment. The programme opens with J.P. Morgan Asset Management’s ProxyIQ journey, a practitioner account of building an AI-powered platform to modernise proxy voting and challenge legacy market infrastructure.
From pilots to production
The central question running through the agenda is how firms scale AI with appropriate controls in regulated markets. The opening panel on safe, scalable deployment will examine AI control planes, standardised data models, guardrails, kill switches, AgentOps and integrated compliance oversight.
Usman Khan, founder and chief executive officer of APEX:E3, frames the challenge around orchestration: “The hard problem in enterprise AI was never a single capable agent, it’s getting many of them to work together safely, at scale. ALICE is the orchestrator agent built to do exactly that: securely coordinating other agents, with compliance and control designed into the architecture rather than bolted on.”
Frank Callaly, chief technology officer at Neueda, adds: “Safe and scalable looked achievable when AI was a co-pilot. As agentic AI moves into trade surveillance, credit, and regulatory reporting, the question changes entirely. The firms that will deploy safely at scale are those investing now in the human oversight capability that autonomous AI demands.”
Callaly will also lead a keynote on closing the AI capability gap, looking at why traditional training programmes often fail to create lasting behavioural and organisational change. The session will focus on the skills, operating models and leadership behaviours needed as AI becomes embedded in day-to-day workflows.
Measuring value
The summit will also address a board-level question: how to measure AI return on investment. A dedicated panel will examine repeatable ROI frameworks, chief financial officer scrutiny, build-versus-buy decisions, risk reduction metrics and the organisational changes needed to make technology teams accountable for value creation.
The agenda frames AI value through efficiency, risk reduction, control evidence and decision quality. In regulated markets, measurable value may also come from reduced manual intervention, better control evidence, improved anomaly detection, faster investigation triage and more consistent decision workflows.
Data as the foundation
Data readiness is another major theme. A buy-side and sell-side panel will examine how firms move from storage-led architectures to trusted, standardised data models that can support scalable AI. The discussion will cover data fabric, lake and hybrid models, alongside lineage, quality monitoring, anomaly detection, privacy and security.
Nicolas Hourcard, co-founder and chief executive officer of QuestDB, says: “Old data infrastructure wasn’t built for the scale and speed AI agents demand across trading, surveillance and backtesting. Firms are rebuilding this layer now to be AI-ready, so agents can actually interact with the data.”
Stéphane Rio, chief executive officer and founder of Opensee, points to the role of governed translation between raw data and business use: “The technical debate around data fabrics versus lakes versus hybrid architectures misses the point – what financial institutions actually need is a semantic layer that turns raw, fragmented data into business-ready information, ensuring every AI query is correct by design, auditable, and replayable. Without that governed translation layer, you’re just giving AI agents access to complexity they can’t be trusted to interpret. And even if you are ready to trust them, you still can’t prove correctness within the time required to act on the results.”
Matthew Cheung, chief executive officer of Ipushpull, adds: “AI readiness in capital markets is a standardisation and governance problem. AI is only as good as the data it sees, and firms can’t scale AI on top of data they can’t govern, audit or trust at the point of use. Firms that will move fastest are those treating standardised, controlled data models as the foundation layer for AI, not an afterthought.”
Governance and trust
The afternoon programme turns to explainability, bias mitigation, model risk, shadow AI, data leakage, deepfakes, stress testing and incident response. A Lloyds Banking Group case study will examine red teaming and bias metrics before production deployment, while a governance panel will address how model risk management must evolve for non-deterministic and agentic systems.
The programme also addresses sovereignty. APEX:E3’s Khan says: “Not long ago, an institution’s technology stack and data were its most closely guarded assets, kept firmly under lock and key. The question worth asking is why we’ve grown so comfortable sending that same code and data to external providers, when sovereignty over it should be the one thing a firm never gives away.”
A connected agenda
The later programme includes case studies, a future-focused keynote on quantum computing and Champagne Roundtables covering enterprise AI operating models, decision intelligence, buy-side trading, portfolio risk, post-trade, T+1, model risk governance and AI-enabled fraud.
For attendees, the value lies in examining how peers, practitioners and technology providers are turning AI from isolated experimentation into controlled enterprise capability. The practical challenge is how to design AI-supported workflows that are measurable, explainable, governed and capable of delivering business value at scale.
Early booking is recommended. For more details on the agenda, speakers, and to secure your place, please visit the official event page HERE.
Subscribe to our newsletter
