CogniveilDemo

Company · Newsroom · 28 September 2026

Trust at machine speed: The closing keynote at Finelevate Digital Summit 2026

Dimitris Dimitriadis closed Finelevate Digital Summit 2026 in Skopje with a keynote on governed AI, instant payments and four futures for finance in 2030.

On 16 September 2026, the Finelevate Digital Summit brought banking, insurance and fintech leaders to the Skopje Marriott Hotel. Stefan Andonovski, Minister of Digital Transformation of North Macedonia, opened the programme. 15 sessions followed, on AI in finance, instant payments, data governance and cyber resilience.

Dimitris Dimitriadis, Founder & CIO of Cogniveil, closed the summit with a keynote on why governed AI decides who wins in finance.

Dimitris Dimitriadis closed Finelevate Digital Summit 2026 in Skopje

He started by handing a pebble to everyone in the room and asking them to put it in their shoe. He did not say why.

Finelevate Digital Summit 2026

Two shifts and a weak signal

The keynote set out two shifts already under way. Money is starting to move on a customer's behalf rather than through them, with software acting on instructions the customer set in advance. And settlement is becoming instant. Fraud checks used to happen in the hours between the instruction and the money arriving. A payment can now be irreversible in 10 seconds, which means the fraud window and the settlement window are the same window.

Then came a weak signal, a small thing that tells you the rules are changing. Banks have spent decades answering one question about every account. Who is on the other side of it? The next applicant may not be a person at all. It may be software that opens an account for someone. The institution will need to know who sent it, what it may do, and how to prove both.

The two shifts and the signal put the same question to an institution. Can it prove what it did?

Regulation is moving the same way, with DORA already in application and the AI Act phasing in, but the question arrives before any supervisor asks it.

Four futures for 2030

One axis is how fast an institution can act. The other is how well it can prove the action afterwards. Together they give four futures for 2030.

Safe and irrelevant. Slow and provable. Every decision waits for a person, and customers leave.

Fast and fragile. Fast and unprovable. Trust collapses at the first incident nobody can explain.

Stalled. Slow and unprovable. Endless pilots, the most expensive way to stand still.

Governed velocity. Fast and provable. Systems act at machine speed, and every action carries its own proof.

Dimitris called the fourth one the winner's quadrant, and was direct about what it costs. It is slower to start than the alternatives. It forces an institution to write down who is allowed to decide what, a question many have left comfortably vague. And it means turning down some automation that would otherwise pay for itself this year. Fast and fragile is rarely an accident. It is usually what happens when a firm takes the cheaper path on purpose.

Three questions

He left the room with three questions. For every AI decision made today, can you show who made it and why, within 10 seconds? Which systems can move money on their own, and who signed off on that? And when software applies to open an account next year, how do you check it?

Then he asked the room to take the pebble out, and to notice the relief.

A weak signal behaves the same way. Stand still and you do not feel it. Start moving and it makes itself known, and the faster you move, the sharper it gets. Most institutions will deal with this one exactly the way the room dealt with the pebble, by removing the discomfort rather than asking what it was telling them.

"Most institutions in that room can already move at machine speed. Very few can show their work afterwards. Closing that gap takes more than a technology project. It is a decision about how the institution wants to be governed, taken long before the first system goes live."

Dimitris Dimitriadis, Founder & CIO, Cogniveil

Governed velocity is the future Cogniveil is built for. The governance sits around the AI rather than inside the prompt, so the controls hold no matter which model is running and an instruction written into a prompt does not widen what a user can reach. Access follows a person's verified identity and assigned role. Answers carry the sources that support them, and when the approved sources do not support a claim, the system says so rather than filling the gap. The human keeps the decision.

Speed is the easy part. What a financial institution has to defend is the answer, weeks later, in front of someone who was not in the room.

The three questions Dimitris left in Skopje are worth answering before the next system goes live. See how governance works at Cogniveil

Cogniveil is a governed AI platform for regulated industries, built for legal, audit, compliance, risk, finance, procurement and onboarding teams. cogniveil.ai

Continue

More from Cogniveil