I co-presented a session on agentic AI with Peter Dooher, SVP of Digital Strategy at Servco, hosted by TRUE Hawaiʻi.

What made this one worth doing was the pairing. I covered how agentic systems actually work, and Peter brought real workflows his team had already put through their paces at Servco. He walked through an agent that booked an actual service appointment end to end from a plain language request, and an application built from a product requirements document where the agent scaffolded the project, ran it, read its own error traces, fixed itself, deployed, and then kept extending the app as new requirements came in.

The distinction we kept coming back to is that an agentic system plans, acts, reflects on what went wrong, and iterates, rather than waiting to be told what to type next. Peter had a good way of putting it. You are less giving instructions to an autocomplete and more handing work to a junior developer who sets up the repo, fixes the bugs, ships it, and stays on call for the next feature.

We left most of the time for questions, which is where these sessions earn their keep. Mahalo to Nicole Cacal and TRUE Hawaiʻi for convening it. Getting academic and industry practitioners in Hawaiʻi into the same room on this stuff is exactly the kind of exchange the state needs more of.