AI DEVELOPMENT
An AI readiness assessment that ends in a decision, not a deck. We map your workflows, data and systems, score every candidate process on value, feasibility and risk, and name the first one worth handing to an agent - including when the answer is none of them.
What an AI readiness assessment is
An AI readiness assessment is a structured review of your workflows, data, systems and constraints that scores candidate processes on how much value automating them would return, how feasible it actually is, and what risk it carries. It ends with a ranked backlog and a recommendation for the first build.
It exists because the expensive mistake in this field is not a failed build. It is a successful build of the wrong process - one that was rare, or judgment-heavy, or sitting on data nobody can reach. Those constraints are knowable in days. Finding them in month four costs the whole engagement.
This is fixed-scope and deliberately cheap, and it is designed to be able to tell you not to proceed. An assessment that can only ever recommend a build is a sales call with a document attached.
You probably need this if you recognize these
- Leadership has asked “what’s our AI plan” and the honest answer is a list of tools.
- Several teams are running small experiments and none has reached production.
- You know something should be automated but not which thing.
- A vendor has quoted for a build and you have no way to judge whether it is the right one.
- You have data everywhere and no clear view of what an agent could actually reach.
What you get
A workflow map
The processes in scope, described as they actually run, including the informal steps that never made it into the documentation.
A scored backlog
Every candidate ranked on value, feasibility and risk, with the score visible so you can disagree with it.
A data and systems read
What an agent could reach, what it could not, and what integration work each option implies.
A recommended first agent
One process, with the reason it beat the others.
An architecture recommendation
What the build would involve, roughly what it would cost to run, and where the governance gates would sit.
A “not yet” list
The processes to leave alone, and what would need to change first.
How the assessment runs
AI readiness assessment and AI strategy
Frame
We agree the scope: which teams, which processes, what counts as success.
Self-healed - retried with fallback tool. Human not required.
Observe
Sessions with the people who do the work, not only the people who own it.
Trace
Systems, data, permissions and volumes are checked against what the interviews claimed.
Score
Each candidate is scored on value, feasibility and risk using a published rubric.
Model
The top candidates get a rough cost-per-task and time-to-value estimate.
Report
Findings, backlog, recommendation and the honest list of what not to do.
What the first agent often turns out to be
Analytics Digest Agent
Often first, because the inputs are clean and nothing irreversible happens.
CRM Hygiene Agent
Often first in sales-led businesses, where bad data is quietly costing more than anyone has measured.
Order & Logistics Agent
Often first in operations, where exceptions are frequent, rule-bound and currently handled by whoever is free.
What we look at
We are assessing reachability, not selling connectors. The output includes the systems where integration would be the hard part.
- CRM
- ERP and finance
- Helpdesk and ticketing
- Data warehouse
- Document stores
- Email and calendar
- Identity and access
- Whatever the spreadsheet is doing that nobody will admit to
Platform names are shown as examples of the categories agents connect to. They are not partnerships or endorsements.
How we keep the assessment honest
Autonomy boundary
The audit recommends. It never commits you to a build, and it does not create a dependency on us to act on it.
Approval gates
Scope, interview list and the scoring rubric are agreed with you before we start.
What stays human
The decision. We supply the scores and the reasoning; the priority call is yours.
Logging
Every score carries its reasoning, so a recommendation you disagree with can be traced to the assumption behind it.
How the work runs
Typical ranges from our engagement model (doc 04 §5), not a quote.
| Stage | Typical | What happens |
|---|---|---|
| Assessment | 3-10 days | Framing, interviews, systems tracing, scoring and the report. Fixed fee, fixed scope. |
| Decision | your pace | You take the backlog and decide. Nothing expires and nothing auto-renews. |
| Build | 3-8 weeks, optional | If you proceed, the blueprint starts from the audit rather than repeating it. |
| Managed | ongoing, optional | Re-scoring as new candidates surface, usually quarterly. |
What we measure
- 3-10 days
- Framing to report (Typical)
- Every candidate
- Scored on value, feasibility and risk against a published rubric (Target)
- A "not yet" list
- Delivered every time, because it is the part that saves money (Target)
- Your backlog
- Ranked in your priorities, with the scores visible (Yours)
“ Until then these are design targets and typical ranges, not results.
What this looks like in practice
Logistics exception hive
Reference scenario · Logistics. Shipment exceptions detected and actioned before the customer notices.
Reference scenario - a composite build illustrating our method. Figures are modeled and the model is shown.
Frequently asked questions
What do we actually receive at the end?
A readiness report, a scored automation backlog, a recommended first agent with the reasoning behind the choice, an architecture recommendation, and a list of what to leave alone for now. It is written to be read by your CTO, your operations lead and your CFO, because all three have to agree before anything gets built.
Can the audit tell us not to build anything?
Yes, and it does. If the processes in scope are too rare, too judgment-heavy or sitting on data an agent cannot reach, that is the finding, and you have spent days rather than months learning it. We would rather lose a build than deliver one you regret.
Do we have to build with you afterward?
No. The report stands on its own and is written so another team could act on it. Most clients continue with us, but the assessment is deliberately priced and scoped so that it is not a commitment device.
How much of our team's time does it take?
Interviews with the people who do the work, plus read access to the relevant systems. We work from what exists rather than asking you to prepare documentation, because the gap between documented process and actual process is one of the things we are looking for.
Is this the same as AI consulting?
It overlaps, with one difference we care about: this ends in a scored backlog and a named first build, not a strategy document. The test is whether an engineering team could pick it up and start. If it cannot, it was a deck.
What if we already know which process we want automated?
Then the audit is shorter and mostly checks feasibility: can an agent reach the data, what does the exception volume look like, and what would it cost per task. Sometimes it confirms your instinct. Sometimes it finds the process next to it is the better first move.
Related services
AI Agent Development
Purpose-built agents that own one job.
Custom AI & Machine Learning Development
Models built for your data.
Business Process Automation
End-to-end processes that run without a queue.