Ten keynotes for the executives who have to decide

Most organizations have proven that generative AI can produce output. Far fewer have proven it can be trusted to act. This session sets out what changes when AI stops answering questions and starts completing work: the data and identity prerequisites, the emerging interoperability protocols, the human oversight model, the security perimeter around tools and actions, and the operating model redesign that has to happen alongside the technology.

AI strategy documents are plentiful. Operating models that actually deliver are rare. Drawing on enterprise-wide deployments inside regulated institutions, this session lays out how to structure ownership, funding, platform decisions, product accountability, risk partnership and adoption mechanics so that AI becomes a durable capability rather than a portfolio of pilots.

AI performance is a downstream symptom of data condition. This session works through the specific data characteristics that determine whether an AI program scales: semantic consistency, lineage, quality at the point of capture, governed access, privacy controls and the contextual metadata that models depend on. It draws on the consolidation of more than forty data warehouses and a twenty-petabyte estate into a single modern platform.

AI introduces a risk surface that legacy controls were never designed to manage: prompts, tools, model behavior, non-human identities, agent-to-agent traffic and autonomous action. This session frames AI security as a continuous control plane rather than a periodic review, and sets out what boards, CISOs and CIOs should require before autonomous systems touch sensitive data.

Boards are being asked to approve significant AI investment while the evidence base is still forming. Built from monthly board reporting experience inside a large regulated institution, this session gives directors a concrete question set covering capital allocation, value measurement, model risk, cyber and data exposure, regulatory posture, and who is accountable when an autonomous system gets it wrong.

The largest source of disappointing AI returns is applying new capability to unchanged processes. This session covers function-by-function redesign: what work gets eliminated rather than accelerated, how roles change, where decision rights move, and how to manage the workforce and change dimension honestly. It draws on an enterprise-wide AI deployment that reached 88 percent weekly active usage alongside process redesign.

Regulated institutions face a harder version of the AI problem: the same ambition, with model risk management, supervisory expectations, privacy obligations and audit evidence attached. This session covers what it actually takes to get AI into production in a bank, including governance design, model risk partnership, client experience, fraud and financial crime, capital considerations and the remediation discipline that keeps programs credible with examiners.

Transformation programs fail commercially long before they fail technically. This session sets out how to build an investment case that survives contact with the CFO: value drivers that can be measured, the difference between cost avoidance and cost reduction, capital efficiency as an underused lever, and how to report progress in terms a board and an investor recognize.

There is a meaningful difference between an enterprise that uses AI and an enterprise that is organized around it. This session examines that gap across structure, culture, platform strategy, talent model, governance and capital allocation, and what incumbents can realistically adopt from companies that were built this way from the start.

A perspective session for audiences planning beyond the current cycle. It covers where agentic systems, multimodal interfaces, enterprise memory, model routing and autonomous workflows are heading, which shifts are likely to be durable, which are likely to be absorbed into platforms, and what leaders should be positioning for now.