What is behaviour change in training?

Behaviour change in training means that people work differently after training, not just that they know more. A course can raise test scores without changing what anyone does on Monday morning. With AI agents, the behaviour that matters is visible: whether reviewers open the facts before approving, write reasons when they decline, and escalate what they should not decide.

Why is it the real goal?

Because knowledge without behaviour change protects nobody. A reviewer who can define provenance in a test but still approves a held payment backed by an emailed invoice has passed training and failed the job. Businesses that measure training by completion rates or test scores alone are measuring the wrong thing.

Agents make the right thing measurable. Every verdict and every human decision on held work is filed in an append-only record. Before training, a team lead can see how often reviewers approve without opening sources, how often they decline without reasons, how long held items wait. After training, the same measures show whether anything changed.

What gets in the way?

Four common obstacles:

• The old habit is easier. Approving quickly is faster than checking provenance, and queues reward speed.
• The environment does not support the new behaviour. If the rota leaves one reviewer with eighty items, they will cut corners, however well trained.
• Nobody notices the difference. If the team lead never reads the record, careful and careless reviewers look the same.
• People do not believe the reason. Some resistance is about the agent itself; the patterns are covered in what is resistance to change.

Training alone rarely overcomes these. Behaviour changes when training, the work environment and feedback all point the same way.

An example

Take a hypothetical retailer's refunds team. Before training, the record shows reviewers approving most held refunds within seconds and rarely writing reasons for edits. The training covers why refunds are held and how to read the facts. Afterwards, the team lead changes two things in the environment: the queue now shows the agent's reason for holding at the top of each item, and she reviews a sample of decisions weekly with each reviewer. A month later, time per decision has risen slightly, written reasons have become the norm, and two reviewers have caught refunds to accounts with open fraud flags. That is behaviour change, read from the record.

How is it different from change management?

Change management is the wider discipline of moving an organisation from one way of working to another: communication, sponsorship, structure, timing. Behaviour change in training is one outcome within it, focused on individuals doing their work differently. A change programme can succeed at the organisational level, with the agent live and the team restructured, while individual reviewers still work in the old way. Both are needed.

There is a parallel on the agent side. An agent's behaviour lives in its brief, sources and bounds, as argued in the brief is where an agent's behaviour lives. Changing an agent's behaviour means changing those, deliberately, and recording each change in the AI agent change log. Changing a person's behaviour takes training plus an environment that supports it.

How do you measure it?

Pick two or three behaviours that matter, measure them in the record before training, and measure again a month after. The guide to training teams for AI agents shows how to build that into a programme.