Which agent approved this? Making multi-agent AI audit-ready - Webinar

For IT and AI leaders whose AI pilots stall before production: what it takes to run multi-agent AI in a real workflow, with governance and traceability built in from day one.

The pilot went well. The agents do what they're supposed to. Then someone from risk asks a simple question: which agent approved this, on what data, and who signed off?

That's where a lot of multi-agent AI stalls - in the space between the agents, where nobody designed for identity, access or traceability. Weiwei Feng calls it the four ways a workforce of agents breaks. On 12 November she takes one real workflow and puts it through all four:

• A regulator asks which agent approved this - on what data, and who signed off. Could you answer?
• One agent can call another - what stops it borrowing access it should never have?
• The rules change next quarter - do you reconfigure, or rebuild?
• When it goes wrong at 2 a.m., is a human in control - or locked out?

If any of these give you pause, this session is for you.

What you'll take away

• A test you can run on your own AI initiatives the next morning: the four questions, and what a good answer sounds like.
• What traceability by design looks like in a real workflow - who decided what, on what data, when.
• Where your people need to stay in control, and how to spot that they're already locked out.

Who it's for

IT and AI leaders, CTOs and transformation leads in mid-sized and large organisations in regulated sectors. You have AI initiatives or a working demo - and you're the one who'll have to answer when audit asks how it decides.

Who's running it

Weiwei Feng is a former Global AI Director at Capgemini, where he contributed to a group-wide AI offering and getting enterprise AI into production. He now builds Stolonic, a multi-agent engine with identity, distributed security and traceability built in, designed from the business requirements with the domain expert on the team.

Hosted by Kvadrat Kompetens