How to decide what to implement.
Personal ChatGPT is a tool. AI in a company is a change to how work moves - teams, systems, decision points. The licences are usually in. The work is the same.
Improvement only counts when it is tied to strategy, built into how the work gets done, taken live, and measured. AI, automation, and agents are how that lands now - in the process, alongside your people. Not a licence. Not a chatbot.
Which function first, so the next one is cheaper.
A ChatGPT window sits next to the work. Org AI sits in the process - intake, routing, a first cut - with people on the exceptions and the judgement. That is why a licence does not change the business, and why the first question is not “which tool.”
Speed comes from starting right, not starting first. One function that is already a pain. The first sprint should make the second easier. That is the work. Not a forty-item catalogue, and not an 18-month programme.
What “worth doing” means.
- Already a pain
- Expensive, slow, or stuck in email. Not fashionable to automate.
- Feasible where you already run
- One function, in the systems you have. Not a migration.
- Someone will own it
- A person there can run the process after we leave. If not, we say so.
- The next one gets cheaper
- The first sprint should make the second easier. That is the sequencing.
Business value
Where will this materially improve performance, and is it worth doing?
Business process
How must the underlying work change?
Business adoption
What happens between a prototype and the way the organisation actually works?
Governance
Enough to deploy AI, automation, and agents confidently - not a bureaucracy that prevents it.
Implementation
Advice is not the finish line. Live is.
What this is not.
- An AI maturity score
- An 18-month transformation roadmap
- A slide deck with no decision
- A chatbot review
Projects fail operationally before they fail technically. The work still has to land.
Two ways in.
If you do not know what to attack, the audit is how you buy this. Already know what hurts? The sprint.
Opportunity Audit
Do not know what to attack.
A short paid pass through how work moves - teams, systems, decision points - to name the first function worth doing. If you then take the sprint, this work is not done twice.
Implementation Sprint
Already know what hurts.
Three weeks on the calendar. One painful function, live in your systems, left with an owner.
If you do not know what to attack, start with the audit.
Fifteen minutes is enough to see whether it is the right first step.
Book 15 minutes