Trust at agentic speed: reimagining corporate identity in an AI-driven economy

Every bank at Sibos this year is chasing the same thing: agentic AI that works in production, not just as a pilot.

Miami will be full of answers about orchestration. But fewer people will be talking about the harder question underneath it. What happens when an agent acts on a corporate client’s data, and that data is wrong?

That is the question behind Encompass’s panel at Sibos 2026: “Trust at agentic speed: reimagining corporate identity in an AI-driven economies,” and it sits squarely inside this year’s Sibos theme, “Digital finance for AI-driven economies.”

Why this belongs on the Sibos floor

Sibos exists because no single bank solves trust and connectivity alone. It is where the industry compares notes on what is working, not just what vendors are pitching. Agentic AI is the newest version of that same conversation. Everyone agrees agents can move fast, but few have agreed on what makes an agent’s decision safe to act on.

That gap is exactly what this panel is built to close.

Meet the panel

Three senior leaders will debate what trustworthy agentic AI requires, from three different seats inside global banking:

• Lori Messer (US COO and Global Head, Cross Product, RBC Capital Markets)
• Sameena Shah (Chief AI, Data and Transformation Officer, J.P. Morgan)
• Rafik Majiti (Global Head of Digital of Wholesale Banking and Global Head of Corporate Data Management, ING)

If you attend one session at Sibos this year, make it this one.

The real blocker is not orchestration

The industry has spent two years asking what agents can do. Sibos 2026 is where the harder question finally gets airtime: what happens when an agent acts on bad data?

A retailer’s agent recommending the wrong product is a bad experience. A bank’s agent acting on an incomplete ownership structure, or a stale sanctions screen, is a regulatory event. Regulators are not asking whether banks use AI anymore. They are asking whether banks can prove the data behind every AI decision is accurate and current.

Every bank walking the Sibos floor this year will feel that pressure. Few will have a clean answer.

Connectivity is not the same as trust

Some of the agentic AI conversations this year center on model context protocol (MCP). It is worth being clear about what it does. MCP lets an agent pull data from different systems without custom integration for each one. That is a real engineering advance.

But MCP has no opinion on whether the data it moves is accurate. A well-built connection to bad data just delivers bad data faster. Connectivity is not the same problem as trust, and Sibos is exactly the place to stop conflating the two.

What makes an agent’s decision defensible

Corporate digital identity (CDI) is the layer that closes this gap. It combines real-time public data with verified private information into one continuously maintained risk profile per corporate customer.

For an agent, that profile is the difference between a decision grounded in evidence and one built on whatever data happened to be reachable.

Three things matter most once agents, not people, do the reasoning:

• Provenance. Can you show where each data point came from, and when it was last checked?
• Currency. Ownership, sanctions status, and adverse media all change constantly. Stale data means the agent is reasoning about a customer who no longer exists.
• Structure. Unstructured data forces an agent to guess at meaning. A human might catch the error. A fast-moving agent will not.

Get these right, and a model risk committee can sign off on agentic AI. Get them wrong, and every orchestration investment just amplifies the problem.

A preview from the panel chaired by Alex Ford, CRO, Encompass:

Corporate identity has a journey behind it: paper documents, passport-checking and, now digital profiles. But the verdict is blunt: “the capability has evolved faster than the trust.” Clients still hesitate to share IDs and corporate data, even as verification technology moves on.

Identity is about collating data, protecting it and earning trust, a discussion that edges into cybersecurity. Transparency cuts both ways with Encompass being able to surface more accurate data on a corporate than a general web search can. Personal identity verification has already built the client comfort that corporate identity now needs to match. An unusual reversal since businesses typically adopt new technology faster than individuals.

Bring your questions, we’ll bring you the answers

This blog raises the question the industry has been slow to ask: what happens when an agent acts on data no one has verified? The panel will answer it live, from three very different vantage points. Where they diverge, on how much agentic decisioning should run before a human sign off, on what good data governance looks like at scale, on the realistic pace of adoption given regulatory reality, will be as instructive as where they align.

If you lead model risk, compliance, or digital onboarding at a bank, this is built for you. A few questions worth bringing into the room:

• Can your model risk committee explain why an agent made a given decision, and who is accountable if it gets it wrong?
• Is your proprietary data infrastructure keeping pace with what agentic AI now demands of it?
• What would it take for your institution to be as transparent with corporate clients about data use as banks have already learned to be with individuals?

See you in Miami

Encompass’s EC360 platform delivers structured CDI profiles directly into the systems banks already use, so agentic initiatives inherit trustworthy data by design.

Encompass will be at Sibos 2026 in Miami Beach, September 28 through October 1. Come hear Lori Messer, Sameena Shah, and Rafik Majiti debate what trust at agentic speed really requires. Then bring your own questions to our team.

Book a meeting with our team at Sibos to get the fuller data readiness discussion ahead of the event.