AI readiness assessment
Answer ten questions about your processes, data, team and the way work actually happens. You get a level, a clear view of your weakest link, three concrete next steps and the engagement type that fits your situation.
No email, no sign-up. The result appears right away.
10
questions
4 min
to complete
0
data sent
Question 1 of 10
Pick the description closest to how your company works today. You don't need to check anything — if you don't know, choose the nearest option.
Your result
0%
Foundations
The things every implementation depends on are not in place yet: a described process, data you can use, and a person accountable for the outcome. That's a normal starting point, not a verdict. Projects that start with a model rather than a process end as a demo nobody uses. Start with one process and one dataset; we'll pick the tool later.
Pick one process that repeats every week and takes the most time. Write down how it runs today.
Record the baseline: cases per week, minutes per case, errors per hundred. Without those three numbers you can't judge any improvement.
Name the process owner and the person who will judge results. Without them the project ends at the first demo.
Ready for a pilot
Processes, data and team are orderly enough to test one AI use case within limits and measure the outcome. The risk is no longer the technology; it's scope that grows too wide. A pilot on one process, with a baseline agreed up front, will show within weeks whether to continue — or not.
Choose one use case: one task, one team, one measurable outcome. Everything else waits.
Prepare representative examples — dozens to a few hundred — and name the person who will judge correctness.
Write down what counts as success and what would make you stop, before you start rather than after.
Scale what works
Processes, data and team are mature. The biggest risk isn't capability but dispersion: several initiatives running at once without shared rules, permissions or measurement. Value now comes from two or three use cases taken into daily work, not from a longer list of ideas.
Pick the two use cases with the highest impact and take them to production. Put the rest on a waiting list.
Agree shared rules: agent permissions, approval for consequential actions, logging and how to roll back.
Set a regular review: cost, quality, latency and the number of human interventions. That is the data the next decision rests on.
The lowest of your four scores. Start with one process that repeats every week: write down its steps and measure the baseline. Without it, every implementation is guesswork.
The lowest of your four scores. Pick one dataset the solution will use, decide which system is the source of truth, and check who may use it. That is usually cheaper than another round of model tuning.
The lowest of your four scores. Before starting with a model, decide who owns systems and integrations and what exactly needs to be built or connected. A partner can carry part of that work.
The lowest of your four scores. What's missing is an owner, time and shared rules — not tools. Start with one team and one task instead of a company-wide AI programme; a result in one team persuades the rest faster than a presentation.
Your answers don't point to a bottleneck. That's rare, and it doesn't mean there's nothing to improve. The next step is keeping the measurement in place across two or three use cases and making sure quality doesn't drop as their number grows.
The engagement type that fits
Software development and integrations
We design and build what's missing, connect systems and clean up the data flow — from discovery through deployment.
What's missing is the technical foundation and ownership, not an AI idea. We start with architecture, integrations and the question of who maintains the solution after launch.
You have no engineering team to take on AI work, so we build the tool and the process around it first. Maintenance ownership is agreed in the contract.
See how we workAI implementation and workflow automation
We start with one process and its data, pilot one use case and measure the result — from discovery to deployment and maintenance.
Process is your bottleneck, not technology. We start with an assessment of one process and its baseline, then decide whether automation makes sense.
Your answers point to data as the bottleneck. We start by connecting sources and checking quality, then run the pilot on data you can trust.
See how we workAI coding adoption for your team
Working-practice review, tool setup, workshops and a supported pilot on your own tasks. Your team keeps the practices.
Your engineering team exists and has the skills, but working with AI isn't daily practice yet. We start with setup and workshops on real tasks, not prompt theory.
See how we workNothing is sent or stored. Scoring runs in your browser.
This is a structured self-assessment, not an audit. It reflects the answers you gave; a call covers what a form cannot.
Method
Four dimensions decide whether an AI implementation survives contact with daily work. We test each in two or three questions.
Do you know how the work you want to improve actually runs today — and can you measure it?
AI improves a process that is described and measured. If nobody knows how long a task takes or what an error costs, no implementation can show whether anything changed.
Does the data AI needs exist in one place, stay current and have clear access rules?
A model will not fix data you don't have. Scattered spreadsheets, duplicates and no data owner are the most common reason a pilot looks good in a demo and fails in daily work.
Does someone own your systems, integrations and deployments — with the time and skill to maintain what gets built?
Every AI implementation has to connect to your systems and have an owner after launch. Missing that person doesn't kill the project; it changes the plan — what's missing gets built first.
Does the team use AI in daily work, or does it live in a few private experiments?
A tool doesn't change a company on its own. If nobody owns adoption and no time is set aside to learn, the project stops at the demo stage regardless of the technology.
Each question is worth 0 to 3 points, 30 in total. The result is a percentage of the maximum: 0–33% foundations, 34–66% ready for a pilot, 67–100% scale what works. We also show the lowest of the four dimensions — that sets the first step, not the total.
0–33%
Foundations
34–66%
Ready for a pilot
67–100%
Scale what works
We don't rank you against other companies. We have no dataset of other organisations' scores and won't invent a position for you. The assessment describes your own profile and the next step that follows from it.
They stay in your browser. We don't send them, store them or track you. You only share an email address when you decide to contact us.
In a free consultation we'll go through your answers and agree the first step: discovery, a pilot or a concrete build. If AI isn't what you need right now, we'll say so.