Insights · Agentrix

By Agentrix · Updated 20 October 2026

The five stages of AI maturity, process by process

A lot of AI projects stall because nobody wrote down the starting point, so nobody can say later what changed. We score each process separately, because your invoicing and your HR files are probably at different stages.

The ladder

1. Manual. Done by hand, on paper, in inboxes or from memory, and it depends on who's in that day.
2. Digitised. The information sits in a system or spreadsheet, but people still key it in, check it and move it along.
3. Automating. Fixed rules move or check the data on their own, and staff deal with the exceptions.
4. Agent-assisted. AI reads, matches or drafts the work, and a named person checks and approves it before it counts.
5. Agentic. AI runs the process within limits you set, logs what it did and hands anything unusual to a person, who still owns sign-off.

The safety line between 3 and 4

A process only scores 4 or 5 if the safety basics are in place: an AI policy, a DPIA where personal data is involved, and a person reviewing what matters. For care providers, that lines up with DHSC's guidance on AI in adult social care (29 Sep 2026) and CQC's expectations (updated 21 May 2026).

You don't have to reach stage 5

For a lot of work, stage 4 is the right place to stop, with a person approving every output. In care, it's often the right ceiling.

A real example

GK Healthcare is a UK recruitment company, and the process was finance reconciliation, not care work. It moved from Stage 2 (Digitised) to Stage 4 (Agent-assisted): from about 2 days a week of the Director of Finance's time to one 30-minute run a week, with the Director of Finance checking the output. Read the case study.