The signal this week: AI agents — software that takes action autonomously, not just answers questions — are moving from experiment to operational reality. Gartner forecasts 40% of enterprise applications will include agentic AI by end of 2026. The organisations that understand what this means for their operations are already making decisions. The ones that do not are making a decision by default.
WHAT AN AI AGENT ACTUALLY IS
Not a chatbot. Not a search tool. An agent is software that executes tasks — booking, routing, drafting, escalating, reporting — within boundaries you define, without someone manually triggering each step.
Think of the work in your organisation that currently requires a person to notice something, decide what to do, and then do it. Chasing an approval. Following up on an outstanding document. Flagging an anomaly in a report. Assigning a task when a trigger is met. That is the territory agents operate in.
THE OPERATIONAL QUESTION
Most organisations are not ready for agents — not because the technology is too advanced, but because their underlying processes are not defined clearly enough for software to follow them.
An agent cannot improve an unclear workflow. It can only execute what is specified. If your processes rely on informal knowledge, personal judgement or tribal memory, agents will expose that dependency immediately.
The prerequisite for AI that works is operations that are already well-defined.
WHAT NORTH FORGE DOES
We help organisations get process-ready — defining, simplifying and documenting how work actually flows before any technology is introduced. When organisations are ready, we design and implement AI solutions that fit their actual operations.
If you are curious about where agentic AI could add value in your organisation, reply to this email. We can tell you quickly whether you are ready — and what it would take if you are not.
Craig Sewell | Founder, North Forge AI