Diagnose first. Build only what survives the diagnosis.
Five phases, from how your work really runs to keeping what we build running. The second phase, the audit, decides whether the other three happen at all.
- Understand. We map how the work actually happens, with the people who do it, and cost it: volume, time per unit, loaded cost, and where the handoffs fail.
- Assess. The audit. Every opportunity is scored on impact, feasibility, cost and risk, and ends in one of four outcomes. Nothing gets designed before this exists.
- Build. A proof of concept against your real data, with a pass threshold agreed before it starts, then production. Agreed test cases decide when it is ready, not a demo.
- Integrate. Into the tools your team already uses: ERP, CRM, file shares, single sign-on. A system people have to open a separate tab for is a system people stop using.
- Maintain. Accuracy, latency, cost per request and actual usage, monitored. Systems that handle language degrade in ways ordinary software does not, and nobody notices without instrumentation.
The audit: find out what is worth building, before you build it.
Two to four weeks, a fixed scope agreed in advance, and one deliverable your team can act on whether or not you hire us for the build.
The audit runs in six steps: discovery, process mapping, scoring, architecture, ROI and roadmap. At step four, architecture, each opportunity lands on one of four endings: a process change, deterministic automation, an AI system, or building nothing.
That is not a failed audit. Chasing technology instead of an outcome is the most cited cause of AI project failure, and finding that out in week three costs a fraction of finding it out in month nine.
What you get from the audit.
- A scored list of every opportunity we found, including the ones we recommend dropping.
- The scoring model itself, so the assessment is repeatable without us.
- A data readiness assessment per opportunity, because that is what usually decides feasibility.
- Architecture outlines for the opportunities worth pursuing.
- An ROI model with its assumptions exposed, in ranges rather than false precision.
- A prioritised roadmap: quick wins, medium term, structural, each with a cost range.
- A walkthrough session with your team, recorded.
What it costs
It depends on how many processes are in scope and how scattered your data is, and we would rather quote accurately than publish a number that is wrong for you.
What is fixed and stated before anything starts: the scope, the duration, the deliverable and the price. You will not receive a variable invoice, and you will know the number before you commit to anything.
The first call is free and it is where the number comes from. If the honest answer is that you do not need an audit, you will hear it on that call.
How the team is put together.
We are a small core that works with a network of specialists. That is a deliberate structure rather than an apology for one.
A retrieval problem, a data pipeline and a workflow automation are three different disciplines. A firm that staffs all three from the same bench solves all three the same way, which is how you end up with a retrieval system built by someone whose actual expertise is pipelines.
The core does the diagnosis and owns the architecture and the delivery. Specialists come in where the work calls for them. You always know who is doing what, and the people who did the diagnosis stay on the project.
A headcount we do not have. If a project needs capability we cannot assemble, we say so and point you elsewhere.