AI Audit and Transformation
Evidence decides where your AI starts. Move from wondering where AI fits to knowing which workflow pays for itself and which ones we’d tell you to leave alone.
What the audit examines.
Most teams come to us with AI already scattered across the business, much of it running in personal accounts. The audit is where it gets properly set up, with your data back under your control.
Without evidence underneath the decision, you’ll only ever fund pilots that never reach the people doing the work.
- The workflows your people actually run, mapped from the work itself
- Where the hours go, costed so you can check the business case
- Where your data lives, and who is allowed to see it
- What your regulators and insurers expect of the work you automate
- Where the risk sits, and who signs off when it goes wrong
- Which workflows come first, and which we would leave alone
We run this as an independent, builder-neutral assessment: a workflow earns its place on your roadmap on the evidence, and the reason it did is written down next to it.
“Their team quickly understood the unique challenges of our business … and delivered a thorough, evidence-based strategy.”
Four questions decide every workflow.
Run them against your own shortlist while you read. Every candidate is scored on all four, including the ones we end up turning down.
Payback
What does this work cost you today, and what would you get back? We sit with the people who run it and count the hours the work actually takes, so the number in your business case is one your finance team can check.
Data
Does the information this work needs actually exist, and is anyone allowed to see it? We find where it lives and what shape it’s in, so you know what would have to change before anything reads it.
Systems
What does this work already pass through, and what would it take to wire in? We map your systems and the joins between them, so you’re buying automation that runs inside the operations your team already works in.
Accountability
Who owns the output, and who answers for it when it’s wrong? We settle what your regulators and insurers expect and where the risk sits, so accountability is written down before anything goes near a customer.
A workflow that fails on data does not reach your roadmap, however good the idea behind it was. Telling you that in the first week is part of what the audit is for.
Your first AI automation within 4 weeks.
Prove and Scale land together when the scope allows.
Identify
We sit with your team and map where the hours actually go
- Workshop with your team
- Workflow mapping
- ROI baseline agreed
Build
We set up your platform and agree the rules before building anything on it
- Platform and guardrails in place
- Wired into your systems
- First working version
Prove
We count the savings against the hours you measured in week one
- Live on real workflows
- Measured against baseline
- Outcomes signed off
Scale
The same team carries on into the next workflow, with nobody new to bring up to speed
- Production rollout
- Your team trained to extend
- Next workflow scoped
What you leave with.
The manifest
- A written report your board can read in one sitting Every workflow we looked at, ranked, with the reasoning next to it.
- A prioritised 12 month roadmap Costed against what the audit found, and yours to keep.
- A platform your team is set up and trained to use Set up in your accounts, with access rules agreed.
- The evidence behind every call The hours we counted, so your team can check the working.
- Your data connections mapped Where your information lives and how it moves.
- The first working system, handed over Built to hand over and keep.
What we ruled out
The no-list
The work we judged and turned down, with the reason written next to it. In two of our seven audits so far it ran to its own list, one of them eight items long.
Inside the written report.
The report is written for the people who will have to fund the work, so it opens with the ranked list and the reasoning, and the counted hours sit behind that for anyone who wants to check them.
Every workflow we put in range gets its own entry: what it costs you today in counted hours, what the four criteria said about it, what would have to change before it could run, and where it lands in the 12 month roadmap. The roadmap is ordered by what the audit found and costed against it, so you can take it to a board without us in the room.
Who the audit is for.
The audit works where there is enough of something for the hours to be countable: claims, applications, tickets, case files, the report somebody assembles by hand at the end of every month. It also needs a team who can point at a week of their own work and say where it went, and somebody senior enough to act on the answer inside the year.
It is a poor fit if what you want is a document that satisfies a board and nothing after it, because the first build happens inside the same engagement and needs someone on your side to receive it. It is also the wrong starting point if your question is whether an AI system you already run is behaving itself, which is an assurance question and begins with the AI Charter.
Thorlux could check our numbers against their own dashboard.
Thorlux runs SmartScan, a lighting sensor estate that had been collecting occupancy data for years without turning it into anything a director would sit down and read. We ran a three week discovery sprint to work out whether AI could close that gap, and what it would take to try.
Week one was three workshops with the Lighting Controls Director, the Technical Lead and the Analytics Lead, working out who each report was for and what that audience needed to see. Then we went into the database itself, mapping five tables and running proof-of-concept SQL against roughly 6.5 million occupancy records for Thorlux HQ. The numbers our queries returned matched the live dashboard for the same site, which is how everything after it earned the right to be believed.
Week two benchmarked seven candidate language models on cost per report across four report types. Week three produced an AI strategy report and a pilot technical specification, with three sample reports generated from real Thorlux HQ data, so the recommendation could be judged on output rather than on slides. Thorlux’s own engineering team took the specification forward, and the pipeline was specified to run in a UK region.
One workflow is enough to start. Tell us where the hours go
What happens after the audit.
OpenKit becomes your business’s embedded AI team: the people who ran your audit stay on to deliver your 12 month roadmap.
Governance runs beneath all three. The rules for what AI may and may not do are written while the audit is still running: how the AI Charter works. Teams shopping for AI strategy consulting usually want what the audit produces, and the consulting page sets out how that work is grounded.
Score your own readiness first.
It is an indicative self-assessment, and the audit is where the same picture gets grounded in your actual tools and data.
FAQ
Do we need to know what we want before an audit?
No. Working out where the value sits is the audit’s first week, not something you have to bring with you.
What if AI isn’t the answer for us?
Then the audit says so. The work we judged and turned down is a named part of what you get, with the reason written next to each item.
Who from our side needs to be involved?
The people who actually do the work being mapped, for the workshop week. After that we need one point of contact and not much more of your time.
Can we buy the audit on its own?
The audit opens the engagement rather than standing alone. The first build happens inside the same three to four weeks, so you leave with working software, not a slide deck.
What does it cost?
A fixed fee, agreed before we start. Tell us which workflows you want in range and we’ll quote it.
Start your audit.
Tell us which workflow eats your team’s week and we’ll come back with how we would scope it. Not ready for that? Score yourself with the free AI readiness check first, then just fill in the form when you are.
We reply within one working day.