You can watch replay of session or download the resources by clicking the button below.
You can watch replay of session or download the resources by clicking the button below.
The next instalment of our masterclass series shows marketing and compliance teams how to build AI agents for financial promotions review and approval. It is run in partnership with the ICA.
Matthew Huntington, Head of Marketing and Distribution Compliance at Ninety One, unpacks the FCA’s new CCI product summary. It becomes mandatory for every UK retail fund by 8 June 2027, and it comes with no template and no exemption from the financial promotion rules. Matthew covers what that means for review and sign-off, where marketing and compliance responsibilities now overlap, and where AI agents can take on the volume.
Ashi Bajwa, Regulatory AI Architect at Zango, builds a review agent live, with human-in-the-loop checkpoints at each stage:
Financial promotion identification. The agent judges whether a piece of content is a financial promotion at all. It separates promotions from education, image advertising and exempt communications before they reach the review queue.
Claims library. The agent extracts the claims in each promotion and matches them to claims your firm has already approved, along with the evidence, caveats and expiry behind each one.
Suggested fixes. The agent drafts amends that make a promotion compliant. They’re written in your firm’s own language and grounded in the rules and past decisions your reviewers rely on.
All registrants get 48 hours of sandbox access afterwards to test the workflow on their own marketing assets.
How to design a financial promotions workflow as a sequence of AI agent steps with human-in-the-loop checkpoints
How the CCI product summary changes review and approval for fund marketing
How to identify whether content is a financial promotion, and which borderline calls have to stay with a person
How to build a claims library from past approvals, so approved claims are reused rather than reviewed from scratch
How to generate suggested fixes grounded in real regulations



