What did Kognitos present at AI Summit New York 2024?

Kognitos COO Drayton Wade keynoted the session 'Navigating Uncharted Waters – Identifying Systematic Risks of AI in Finance', covering deterministic policy, evidence-grade audit, and how neurosymbolic architectures de-risk AI for money-bearing workflows.

Where can I watch the AI Summit New York session?

The session highlights are embedded on this page. For the full recording or speaker deck, fill the contact form and we'll share what's available.

What systematic risks did the session cover?

Hallucination on money-bearing decisions, model drift over time, opaque vendor claims, audit gaps that surface in regulator letters, and the operational risk of fragmented AI deployments across departments without a shared governance pattern.

How does Kognitos address those AI risks in production?

A neurosymbolic runtime that keeps decision-making deterministic, an evidence-grade audit log for every action, deep integration with enterprise IdP/SoD controls, and a written-rule policy layer the business owns directly.

Is Kognitos exhibiting at upcoming industry events?

Yes — see the Events page for the current calendar. We attend AI Summit, HFS, Bank Automation Summit, AI4, Intelligent Automation Week, and key sector conferences.

Kognitos at The AI Summit, New York, 2024

Name: Kognitos at The AI Summit, New York, 2024
Uploaded: 2026-04-08
Duration: 40 min 17 s

About this panel

Kognitos CEO Drayton Wade joined a panel at The AI Summit, New York 2024 to debate systemic risks of AI in financial services. The session ran 40 minutes and was moderated by Forbes contributor Joe McKendrick, who writes on AI trustworthiness for Harvard Business Review.

The panelists

Magdalena Konik — General Counsel at AIQ (UAE energy-sector AI).
Drayton Wade — CEO of Kognitos, working primarily with financial services and finance & accounting teams.
Kieran Norton — Head of Cyber AI and Automation at Deloitte.
Hariom — VP, Risk AI at a major bank; author of Machine Learning Blueprints for Finance; UC Berkeley fintech startup advisor.

Where the panel agreed and disagreed

Most GenAI risks are nuances on existing programs: Kieran argued security, privacy, regulatory, and reputational risk programs already exist in financial services — modernise them rather than build parallel structures.
Hallucination and accuracy are the new wrinkles: Hariom pointed to hallucination, accuracy benchmarks, and ongoing pilot work as the GenAI-specific add-ons banks are layering on top of traditional AI risk practice.
Drayton's contrary point — autonomy is structurally different: in the RPA era, risk was bounded — a bot did exactly what its developer scripted. With agentic AI, the system can choose actions across multiple systems, which expands the risk surface in a way that isn't captured by traditional programs.

What's changed in vendor onboarding

AI Councils: most large financial-services organisations now have an AI Council that sits alongside InfoSec and Legal as a diligence layer. Legal representation typically sits on the council, and it weighs use case, data flow, and model behaviour.
Acceptance criteria matter more: panelists agreed AI contracts demand unusually careful scope, acceptance testing, and payment triggers — “already even before you start a project everyone thinks it's this but it's actually this” — because acceptance often gates the money.
Validation agents: a recurring theme — agents that cross-check the primary AI's output against alternative data sources before it acts. Hariom cited validation agents as an emerging best practice.

Questions answered in this video

Who represented Kognitos on the panel?

Drayton Wade, CEO of Kognitos, who works primarily with financial services and finance & accounting teams on long, complex workflows.

Did the panel agree that GenAI introduces totally new financial-services risks?

Mostly no. Kieran Norton (Deloitte) argued that the underlying risk domains — security, privacy, regulatory, reputational — already exist and should be modernised, not duplicated. Hariom (bank Risk AI) agreed but called out hallucination and accuracy as new wrinkles. Drayton Wade pushed back specifically on autonomy — in RPA the risk was bounded to what a developer scripted; agentic AI changes that.

What's an AI Council and why does it matter for vendors?

An AI Council is a new diligence layer most large financial-services organisations have stood up alongside InfoSec and Legal. It typically includes legal representation and reviews AI use cases, data flow, and model behaviour during vendor onboarding. Vendors should expect a Second-Step style review beyond the traditional InfoSec / Legal path.

What did the panel recommend for AI contracts?

Spend extra time on scope, acceptance testing, and acceptance criteria — because acceptance frequently gates payment. The panelists warned that misalignment on what was actually being delivered is the most common source of dispute in AI contracts, more so than in other technology contracts.

What are validation agents and where do they fit?

Validation agents cross-check a primary AI's output against alternative data sources or rule sets before action is taken. Hariom highlighted them as a practical mitigation against hallucinations and accuracy gaps in multi-system agent scenarios.