Before you build AI, find out where it actually pays off.
This is the honest entry point. We look at your work, your data, your tooling, and your risk, then tell you plainly where AI would earn its cost and where the answer is "not yet." You leave with a plan you own and a pilot-first rollout that measures value on one real task before anyone scales it. No hype, and a candid "not yet" when that is the right call.
> A plan you can act on,
> with or without us.
The work that decides whether an AI project succeeds.
Most AI projects do not fail because the technology was wrong. They fail because the wrong task was chosen, the data the system relied on was not ready, or nobody agreed in advance what success would look like. Those are not problems you discover halfway through a build. They are problems you can see at the start, if you look honestly, and looking honestly is what this engagement is for.
So before any code, we sit with your work and ask the unglamorous questions. Where would AI genuinely help, and where would it just be a demo that impresses once? Is the content and the data it would lean on clean, current, and reachable, or scattered across systems and quietly out of date? What guardrails would keep it trustworthy? And how should a rollout be sequenced so your team can absorb it rather than be handed a black box? The output is a short, honest plan: a few places worth pursuing, the things to fix first, and a sensible order to do them in.
This is the counterpart to the build work, and it can stand entirely on its own. The deliverable is a plan you own and can act on, with us or without us. We would rather tell you that your data is not ready, or that a packaged tool already covers your need, than talk you into a program that will not hold up. That candor is the value here, not a polite add-on to it.
What this engagement is
A consultative assessment that ends in a plan, not a sales pitch for a build.
- A standalone engagement, or a lead-in to a build
- An honest read on where AI fits and where it does not yet
- A clear-eyed look at whether your data is ready
- Guardrails, governance, and risk decided up front
- A pilot-first plan that measures before it scales
- A plan you own and can run either way
Four lenses, one honest plan.
The assessment runs along four lines: the work, the data, the tooling and architecture, and governance and risk. Each one is a place AI projects commonly come apart, and each one is something you can read honestly before committing a budget.
We keep it tool-agnostic on purpose. The model is a swappable part behind a clean interface, something you can change later without rebuilding everything around it, so the plan does not chain you to a single vendor or a name that may not be the right one in a year.
What the assessment covers
- Where AI pays off across your processes, and where it does not yet
- A data-readiness review: clean, current, and reachable content
- Tooling and architecture matched to the systems you already run
- Build-versus-buy guidance, with a candid recommendation
- Governance, privacy, and risk decided up front, not bolted on
- The guardrails the system must run inside, written down
- A pilot-first rollout sequenced so your team can absorb it
- A plan you own and can act on with or without us
The least glamorous part, and the one that decides the outcome.
A model is only as good as what it can read. An assistant grounded in your content cannot answer well from material that is scattered, stale, locked in formats nothing can reach, or simply wrong in places. It will inherit every gap and answer confidently from bad material, which is worse than not answering at all. This is why so much AI work quietly comes apart at the data layer rather than the model.
So a large part of this engagement is an unromantic look at whether your data is actually ready: whether the content the system would rely on is clean, current, and reachable, whether it is trapped in silos that make it hard to discover, and what preparation it needs before anything is built on top of it. Often that preparation is the honest first phase of the whole project, and naming it early saves you from paying to discover it after a build is already underway.
None of this requires a perfect data estate before you start. It requires knowing where you stand, so the plan fixes the few things that matter and does not pretend the rest away. That is the difference between a system that holds up and one that looked impressive in a demo and fell over on real work.
What we look at in your data
- Whether the content is clean, current, and correct
- Whether it is reachable, or trapped in silos
- Formats and structure the system can actually use
- Gaps and stale material that would mislead an answer
- What preparation has to happen before any build
// clean the record first, then teach the scribe to read it
Read the situation, then pilot before you scale.
The assessment moves from understanding the work, to checking the data, to defining how AI would be governed, to a rollout that proves value on one real task before it is widened. Each step can end in "not yet," and that is a legitimate result, not a failure of the engagement.
// measure on real work first, scale only what earns it
- Understand the workWe start with your processes, not with AI. Which tasks eat time, follow a pattern, and might genuinely benefit, and which are about judgment that should stay with people. Out of this comes a short list of candidates worth examining, and an early, honest read on the ones that are not a fit.
- Check whether the data is readyFor each candidate, we look at the content and records the system would lean on: clean, current, reachable, and in a usable shape, or not. This is where many projects quietly fail, so we surface it now. Sometimes preparing the data is the first phase, and we will say so plainly.
- Decide tooling, governance, and riskWe outline what would sit where, how it fits the systems you already run, and which model-class fits the job, kept tool-agnostic so nothing chains you to one vendor. Then the guardrails: what the system may do, what it must refuse, where data is processed, and what gets logged.
- Set the bar before anything is builtWe define what good looks like in advance and tie it to business outcomes rather than model accuracy alone. That bar is what a pilot will be measured against, so quality is something you can watch on real work rather than take on faith later.
- Plan a pilot-first rolloutWe sequence a small pilot on one real task, then a deliberate widening only if it clears the bar. You get a plan you own, with a candid recommendation on build-versus-buy and a clear "not yet" wherever the timing, the data, or the fit is wrong.
What this engagement will and will not do.
We would rather set the expectation honestly than oversell the assessment. It is planning work, not a build, and a few things are worth saying plainly so you know what you are buying.
Done well, it saves you from spending a lot on the wrong thing. It will not, on its own, make AI suddenly the right answer where the work or the data says otherwise.
Worth being clear about
- The deliverable is a plan, not a working system; building is a separate, optional next step.
- A real possible outcome is "not yet," or "buy this instead of building." That is a result, not a failure.
- Grounding a future system in your own content reduces confident wrong answers, but no AI removes that risk entirely, so anything consequential keeps a human in the loop.
- AI fits some tasks well and others poorly. We will name which is which rather than stretch a fit that is not there.
- The honest read on your data may be that preparation has to come first, before any AI is worth building.
Frequently asked questions
Related services
Strategy and Readiness is where the AI work begins. When the plan points to a build, these are the siblings it leads into, alongside the hub that ties them together.
Not sure whether AI is right for you yet?
That is exactly who this is for. Tell us the work you are wondering about. We will give you a straight read on where AI would pay off, whether your data is ready, and what a small first step looks like, including an honest "not yet" if that is the answer.