“The difference is not intelligence. It is who holds the steering wheel.”
That line reframed the room when Regina Ofori-Boateng, PhD opened the Frontline Workshop Series, hosted by CREED Aberdeen.
Her definition of an AI agent: a model in a loop, with tools, and a rule for when to stop. It is not a chatbot, a clever prompt or a script that calls a model. An agent decides what to do, acts, checks the result, and goes again. You stop being the loop.
Key takeaways
- Cost: Using one expensive model for every step is the most common mistake. Let the strong model plan and the fast model execute. The work and the result stay the same, at a fraction of the bill.
- Scope: “An agent with access to everything is not powerful. It is unscoped.” Every agent needs a room that is locked in and three tiers of permission: act automatically, act then notify, or stop and ask.
- Ambition: “Start with one. Add a second only when you can name the thing the first one cannot do.” Seven orchestration patterns exist, and most real systems need the simplest one.
Where this is already landing
Regina showed real uses in prior authorization in healthcare, well modeling in oil and gas, client onboarding in financial services, and casework triage in the public sector. In every case, a human still signs off on the decision that matters.
About the speaker
Regina is head of agentic AI and robotics at a stealth AI startup and advises a Fortune 500 company on agentic architecture. She was previously a senior GenAI manager at KPMG UK, and her PhD is from Robert Gordon University in Aberdeen.
Watch the full session: Recording
This was session one. Watch out for more in the Frontline Workshop Series.