Before you build,
find out where.
The highest-value automation is rarely the one people arrive asking for. We measure where the hours actually go, whether your data can support a model yet, and what your compliance team will need to see.
Quick answer
How do you decide where AI is worth using?
The decision rests on three measurable things rather than on enthusiasm. First, volume and rule-boundedness: work that happens frequently and has a clear correct answer returns far more than work that is rare and requires judgement. Second, data readiness: a model can only act on information that exists in a structured, accessible and reasonably accurate form, so processes running on scattered documents usually need digitisation before automation. Third, governance capacity: some processes carry regulatory or reputational consequences that require approval thresholds, audit logging and override capability before deployment is defensible. A readiness assessment measures all three and ranks candidate processes by recoverable hours, which is frequently not the ranking an organisation expects.
Enthusiasm picks the wrong
process almost every time.
The process people want automated first is rarely the one that returns the most. Measurement finds the one that does.
Failure 01
Automating the loudest complaint
The most-complained-about process is usually painful because it requires judgement. Automating it produces a worse version of something that was working for a reason.
Failure 02
Building on data that cannot support it
The model is capable, the information is scattered across documents nobody has structured. The project fails and the technology takes the blame.
Failure 03
Governance discovered at the end
The system works, then procurement asks for evidence of oversight, audit logging and residency. Retrofitting governance costs more than building it in.
Four deliverables, not a feature list.
A document you can hand to your board, act on yourselves, or bring to another firm.
01
Opportunity ranking
Candidate processes measured by volume, rule-boundedness and recoverable hours, ranked by return rather than by visibility.
02
Data readiness assessment
Whether the information a process depends on exists in a form a system can act on, and what it would take if not.
03
Governance gap analysis
Your current position against NIST AI RMF and ISO 42001, with the specific gaps a buyer or regulator would ask about.
04
Sequenced roadmap
What to build first, second and third, with the reasoning stated so the sequence survives a change of personnel.
Five steps, in this
order, every time.
Measurement before commitment. The ranking is produced from evidence, not preference.
- 1
Inventory
List candidate processes across the organisation, including the ones nobody has proposed.
- 2
Measure
Volume, frequency, hours consumed and how rule-bound each one genuinely is.
- 3
Assess
Whether the underlying data exists in a form a system can act on, and at what quality.
- 4
Rank
Order by recoverable hours against implementation cost. Judgement-heavy work is marked to stay human.
- 5
Sequence
Produce a build order where each step makes the next cheaper, with reasoning attached.
Intelligence rules assessed during the governance review.
Frameworks the assessment reports against: NIST AI RMF and ISO 42001.
Deliverable you keep regardless of whether you build with us.
Why governance is cheaper before deployment than after.
An organisation deploying autonomous systems will eventually be asked to evidence oversight — by a procurement team, an auditor or a regulator. The controls involved are approval thresholds, audit logging, capability bounds, override capability and declared data residency. Each of these is straightforward to include while a system is being designed and disruptive to add once it is running, because several require changes to how decisions are recorded rather than additions to what the system does. The assessment reports the gap early, when it is still a design decision rather than a rebuild.
Source: The Qawex Standard v1.0, rules INT-01 to INT-10 · /standard/intelligence/Answers, not brochures.
What do we actually receive?
A ranked list of candidate processes with recoverable hours attached, a data readiness assessment for each, a governance gap analysis against NIST AI RMF and ISO 42001, and a sequenced roadmap with reasoning. It is a document you can hand to your board or to another firm, and it remains yours regardless of what you do next.
How long does the assessment take?
It depends on how many processes are in scope and how accessible the people who run them are. Most of the effort is interviews and measurement rather than analysis. We scope it as a fixed piece of work with a defined deliverable rather than an open engagement.
What if the answer is that we should not use AI yet?
Then that is what the assessment says. Frequently the correct first step is digitisation under the Paperless pillar, because a process running on unstructured documents cannot be automated until the data exists in a usable form. Telling you that early is worth more than a project that fails in month four.
Do we need this if we already know what to build?
Not necessarily, and we will say so. If you have a well-defined process, measurable volume and accessible data, the assessment adds little and you should go straight to scoping. It earns its place when there are multiple candidates and no agreed basis for choosing between them.
Will you recommend your own services?
Where they fit, yes, and you should read the recommendations with that in mind. What we will not do is recommend building something the measurement does not support. The ranking is produced from volume and hours rather than from what would make the largest engagement.
Measure before you commit.
The readiness assessment ranks candidate processes by recoverable hours and reports your governance gap. Yours whether or not you build with us.