Fractional AI Officer
Get AI doing real work across your business.
A Fractional AI Officer is the dedicated lead who designs the systems, trains your people, and keeps the whole thing working.
In most of the businesses I look at, AI is still at the dabbling stage. Useful to a few individuals, and invisible in how the business runs.
Where this starts
What dabbling looks like
- Employees using unauthorised tools, with no company policy.
- An enterprise subscription with nobody owning how the team uses it.
- Data too patchy to get anything meaningful out of.
Staff get emails summarised or drafted - but that's as far as it goes. Hardly revolutionary.
These are all the same problem. AI is nobody's job, so it stays where it is.
Why it matters
Shallow AI has a cost
Used badly, with single-step prompting, AI produces the slop that is everywhere now. People have come to read it as a synonym for generic, and no business wants its clients looking at its work and thinking a machine wrote it in one pass.
The difference comes down to design: the prompts, the context a system works from, the checks around it, and whether the people using it know what they are doing.
The alternative to designing it is another year of dabbling, or money spent on a shiny tool nobody makes use of.
Good demos are cheap now. Systems that keep working are not.
What this is
A part-time AI and operations lead, inside your business
I take responsibility for the whole of it: what gets built, whether it works, whether your people can use it, and whether it still works six months from now.
This comes out of more than a decade running operations across a law firm, a charity and a retail co-operative: finance, HR, IT, compliance, and the reporting that tells you what is happening. Now applied to AI.
I run my own business on the systems I build: custom skills, agentic research and reporting workflows, scheduled tasks feeding a live project tracker. I have built them for clients too, including a business development agent and a communications agent that writes in the owner's own voice.
Three steps, and then it keeps going
Work out what is needed
Before anything gets built, we look at how the business runs and what is in the way. Depending on where you are, that means AI strategy, a risk assessment, work on the data layer, or a policy your team can follow.
Design and build the systems
The workflows, the tools, and the connections between them. Built piece by piece, with your team using each part as it goes in.
Keep it working
QA, training, maintenance and review. The tools move quickly and so does your business, so this is where things get tightened, extended and kept current, including as AI regulation changes.
This runs as one role across the whole life cycle. There is no hand-over at the end and no report left on a shelf.
Four things you can hold me to
Tools earn their place
Chosen on whether they do the job and keep doing it, at a price and a level of risk you can live with.
Your people can use it
Training is part of the build. A system nobody uses has not been delivered.
You keep your data
And you always have a clear route to change providers. AI tools are volatile, and you should not be tied to one of them.
It adapts
When the business changes, the systems change with it. That is what the ongoing part of the role is for.
What changes
What it looks like when it is working
- The routine work runs without anyone chasing it.
- The numbers you need for a decision are there when the decision is in front of you.
- A new person can pick things up without everything being re-explained.
- The work that leaves your business reads like your business wrote it.
- AI is part of how the place runs.
And whether the business is keeping up on AI stops being an open question.
Who it is for
Knowledge-work businesses where the admin has outgrown the team
Small and medium businesses, roughly 25 to 100 people. Owners and leaders who want AI working across the business, and who know it will not get there on its own.
Before we talk
See one of these systems at work
A Monthly Management Report, produced end to end by an agentic workflow, with a note on how it was built and what the agent did at each step.
Read the example report