00 AI consultancy
For owners who want AI working in the business but are not sure where it fits. We look at how the work actually runs, tell you what is worth doing and what is not, and quote a fixed build. We build what we recommend, so the advice has to survive production.
01 Advice, or something built?
The question most owners ask first is whether they need someone to advise or someone to build. With us it is one team. We sit with the work, draw it, tell you where AI pays and where it does not, and then build what we recommended. Because we build it, the advice has to survive production. We cannot recommend something clever that we would not want to run ourselves.
It also means we can say when the answer is not to build. We pass on roughly half the projects that reach a discovery call, usually because there is no specific problem yet, the data is not in a usable state, or someone wants the person taken out of the loop altogether. If the task takes twenty minutes a week, or the process changes every month, or nobody owns it, we will tell you. The first call needs one concrete process or pain to look at, so come with the task that hurts, not a general wish to use AI. We do not hand over a strategy document and leave. Our note on the three kinds of AI project we turn down has the detail.
02 Where AI usually pays first
These are the four we meet most often. Each has its own page with the detail.
If you are not sure where to start, hold a task up against three lines. A task that passes all three is usually worth drawing. One that fails any of them is usually better fixed another way.
Website chat, WhatsApp and email sorted, answered from approved wording and passed to the right person, with a person on anything unusual.
Read moreFiling, chasing, triage and data entry handed to workflows that log every step.
Read morePolicies, manuals and files turned into an assistant that shows its source or says it does not know.
Read moreReports that build themselves, and the dashboards, portals and internal tools for when a spreadsheet stops being enough.
Read moreRun a letting or property management business? The front desk and the back office have their own page: AI for letting agents and property managers.
03 Why we draw first
Most automation projects fail for a boring reason. A team automates a process that nobody examined first. The copying and the chasing are wired up faster, but the detours, the double entry and the workarounds are wired up too. Now they run at machine speed, and they are harder to see.
We have watched it happen. A workflow gets built around how one person remembers the job, not how the work actually moves. Six months of history is missing from the plan. The exceptions that eat the most time were never counted. The tool works exactly as specified and still saves almost nothing.
That is why we draw before we build. A process you can see on one page is a process you can question. Before we automate a single step, we want to know which handoffs are real, which are habit, and which quietly disappear once the work is redrawn. The drawing is cheap. The wrong system is not.
Some owners call this an AI audit or an AI readiness check. For us it is one page, drawn from one real case, that you can read and argue with. It is the first step of an engagement, not a product on its own.
04 How we work
01 · Observe
We ask which task costs you the most time or goes wrong most often, and where you think AI could help, because you know the business better than we do. Then we sit with the people who do the work and follow one real case end to end. We count the handoffs and note every workaround, including the ones nobody admits to in a meeting.
02 · Draw
One page: every step, every system it touches, and a rough minutes-per-week figure against each one. The mess stays in. That is the useful part.
03 · Decide
We split the steps that need human judgement from the steps that only need speed, and mark the approval gates: the moments where a person signs off before anything moves.
04 · Automate
Each automated step maps back to a line on the page and a number, so you can see what it was meant to save and check whether it did.
05 Build, buy or leave it
Build it, when the steps follow a pattern and the saving is worth the work. Buy something off the shelf, when a tool you can rent already does the job. Or leave it, when the task is small enough that automating it would cost more than it saves. We will tell you which, and why. The comparison with screen-clicking robots, extra admin staff and doing nothing is on our AI automation page.
06 How we quote
Three to five weeks
Eight to twelve weeks
The build is quoted fixed once the map is drawn, so the build price you approve is the build price you pay. Model usage and hosting are passed through at cost and itemised on top. If you want us to host and run it afterwards, that is a rolling monthly arrangement, quoted separately.
07 What you keep
We are a consultancy, not a platform you rent forever. The process map, the automation and the documentation belong to you from the start. We can host and run it for you, or hand the whole thing over with the notes to operate it. If you ever want to bring it in-house or connect it to something new, nothing is locked behind our login. How we wire AI into the tools you already run is on our AI agents page.
08 Case
A professional services firm was losing about four hours a week to invoice triage. Before building anything we mapped which system owned which fact. That one decision kept the bookkeeping software out of scope and removed half the integration risk.
CaseProfessional services firmUK · anonymised
How we replaced four hours a week of manual invoice triage with an AI workflow that captures, classifies, files and reports, without disturbing how the finance team works.
Finance operations case studyNames redacted out of respect for our clients.
09 Questions
The same answers we would give on a call, written down.
A good one starts by finding out how the work really runs, not by choosing a tool. We follow one real case from start to finish and draw it on a page, with the handoffs and the minutes against each step. Then we tell you which steps are worth automating, which need a person and which are better left alone. After that we build the first one.
It depends on the work, so we do not publish a rate card. Every engagement is a fixed quote after a discovery call. The first piece of work, mapping one process and automating its first step, usually takes three to five weeks. Once the map is drawn we quote the build as a fixed figure, so the price you approve is the price you pay. Model usage and hosting are itemised on top.
With one process that repeats, has a visible cost and follows a pattern someone could describe: invoice intake, enquiry handling, chasing documents, weekly reporting. Not with a tool, and not with a company-wide plan. Pick one, draw it, prove the saving in production, then widen. A useful question to ask yourself: which one task would you most like off your plate? Our first call needs that one process or pain to look at. If nothing specific comes to mind yet, our note on six first moves for a UK small business is the better place to start, and you are welcome to come back when one task stands out.
For some things, yes, and we will say so. A general chat assistant is good for drafting, summarising and thinking aloud. It is not connected to your inbox, your CRM or your files, it does not log what it did, and nobody is accountable when it gets one wrong. When the work has to run every day, touch your systems and leave a trail, a built automation earns its place.
Both. We are a consultancy that builds, which is why the advice has to survive production. We do not sell the map on its own. It is the first step of an engagement, drawn so the build can be quoted fixed, and it is yours to keep once it exists.
Then we say so, at the map and before a build is committed. We pass on roughly half the projects that reach a discovery call, and the reasons are usually the same: no specific problem yet, data that is not in a usable state, or a wish to take the person out of the loop altogether. A process that runs a few times a month, changes every quarter or has no clear owner is usually better fixed another way. The handoff count and the minutes per week on the map are how we say that with a number, not an opinion.
We design for it from the start. We map where personal data flows, define the lawful basis and retention, restrict access and log usage, so a data protection review has something real to check. Anything we run for you is built in the UK and hosted on managed servers in the EU, under UK GDPR.
10 Enquiry
Tell us about the one task or process that costs you the most. We will say what we would map first and whether a build is worth it.