The talk in brief
AI agents are moving from answering questions to taking actions: changing bookings, issuing refunds, approving requests, making decisions on a customer's behalf. Most of the conversation is about capability. Very little of it is about authority.
This keynote reframes autonomy as a transfer of responsibility, not an efficiency gain. Every agentic deployment hands authority to a system, and authority comes with obligation. It is the most important talk for any organisation currently evaluating AI agents.
What it covers
Autonomy is granted, not emergent
Systems don't become autonomous. Organisations decide to let them act. That decision deserves the same scrutiny as delegating authority to a person.
Autonomous systems fail more expensively
When an assistant gets an answer wrong, a customer is annoyed. When an agent takes the wrong action, the business owns the consequence. What that means for design and governance.
Boundaries, escalation, and reversibility
Where agents should act, where they should ask, and how to design decisions that can be explained and undone.
Accountability you can explain to a customer
"The system decided" is not an answer a customer, regulator, or board will accept. What organisations need in place before they grant authority.
What the audience leaves with
- A clear framework for deciding what AI agents should and shouldn't be allowed to do
- The governance questions boards should ask before approving agentic AI
- A practical view of escalation, oversight, and reversibility by design
- A way to talk about agentic AI with confidence, without hype or fear
A 60 to 90 minute board briefing version is available for directors and senior executives who need to understand the governance obligations of agentic AI before they approve it.
Frequently asked questions
What is Ben Farrell's agentic AI keynote about?
It reframes agentic AI and autonomy in customer experience as a transfer of responsibility, not an efficiency gain, and covers what organisations must get right before AI agents act on behalf of customers.
Who should book the agentic AI keynote?
Executives, boards, legal, and risk audiences, particularly organisations evaluating AI agents. A board briefing version runs 60 to 90 minutes.