Free sample — nothing withheld

This is the entire report. Read it before you decide whether to buy one about your own company.

Same structure, same charts, same plan — only the company is fictional and the answers are illustrative. No email required to read it, and no one from Unconsultancy will contact you because you did.

AI Readiness & Buying Diagnostic — Sample

Meridian Components (illustrative)

Ahead of most peers; focus on scaling and closing governance and buying gaps.

Industrial manufacturing & distribution · ~$180M revenue, 640 employees

AI-Forward

Strongest dimension

Strategic Priority

78 / 100

Largest gap

Production Risk

0 / 100

High-priority actions

sequenced below

Peer percentile

68th

mid-market sample

Executive summary

What the diagnostic found

You are ahead of mid-market peers on appetite and strategic intent, and behind them on buying discipline. That combination is the single most expensive pattern we see: leadership wants to move, nobody owns evaluation, and vendors fill the vacuum. Two of the four tools already purchased are effectively shelf-ware. Fix the buying process in the next 90 days and the rest of this plan gets cheaper and faster.

Exploring

0–39

Ready

40–59

AI-Forward

60–79

Leading

80–100

Readiness profile

Six dimensions, scored and ranked

The radar shows shape; the ranking shows where the work is. Both are generated from the same 24 answers.

Data & Systems

Process Maturity

Team Readiness

Regulatory & Risk Exposure

AI Buying & Procurement Discipline

Strategic Priority

Production Risk

Peer benchmark

How this compares with similar mid-market companies

Peer medians are drawn from organizations of comparable size and sector.

Points above or below the peer median

Sequencing

Priority against effort, and where the work lands

Nothing advanced is recommended before the foundation it depends on is fixed.

Remediation plan

The actions, in order

A real report contains 10 to 15 of these. Seven are shown here.

AI Buying & Procurement Discipline

Install a one-page AI purchase test before any new spend

Five questions every proposed tool must answer in writing: which named process it replaces, the measured baseline today, who owns adoption, what the 90-day kill criteria are, and what the exit looks like. Anything that cannot answer all five does not get a budget line.

High priority · low effort

AI Buying & Procurement Discipline

Run a shelf-ware audit on existing licences

Pull seat-level usage for every AI or automation tool bought in the last 24 months. Cancel or renegotiate anything under 20% weekly active use at renewal. This typically funds the rest of the first year of work outright.

High priority · low effort

Regulatory & Risk Exposure

Publish an acceptable-use policy people can actually follow

One page: what data may never be pasted into a third-party model, which tools are approved, and who to ask. Shadow usage is already happening; the policy converts it from unmanaged risk into managed usage.

High priority · low effort

Data & Systems

Consolidate the two systems of record behind one lookup layer

Most retrieval failures here will trace back to the same records existing in both the ERP and the CRM with different keys. Establish one authoritative source per entity before any assistant is pointed at them.

High priority · medium effort

Process Maturity

Document the top three repetitive workflows end to end

Automation cannot outrun undocumented process. Capture the current steps, exceptions, and handoffs for the three highest-volume workflows — this doubles as the specification for later automation.

Medium priority · medium effort

Team Readiness

Name a single accountable AI owner with a budget line

Not a committee. One named executive who can approve, kill, and report on AI spend quarterly. Ownership is the strongest single predictor of whether a pilot ever reaches production.

High priority · low effort

Strategic Priority

Convert the mandate into two measurable outcomes

Replace 'adopt AI this year' with two numbers — for example, quote turnaround time and support first-response time. Every subsequent purchase is judged against those two numbers only.

Medium priority · low effort

Do not buy

What not to buy in the next twelve months

The section nobody with something to sell you will write. Each item is a purchase the answers indicate would not survive contact with this company's operating reality.

Do not buy 01
Next 12 months
An enterprise agentic AI platform (UiPath Autopilot, Salesforce Agentforce or equivalent)
Agentic platforms only pay back where the underlying process is documented and the exceptions are understood. Process Maturity scores 48 here and no workflow is written down end to end, so the agents would be automating an undocumented process nobody can audit. Expect $180K to $400K in first-year platform and integration spend with no measurable baseline to prove against.
Do this instead
Spend roughly $0 and six weeks documenting the top three workflows. That document is also the specification you would have paid a systems integrator to produce.
Revisit when
Process Maturity clears 65 and the three highest-volume workflows have a written baseline with cycle time measured.
Do not buy 02
Next 12 months
Additional Microsoft 365 Copilot seats beyond the current pilot
Seat-level usage on the existing pilot is under 20% weekly active. Buying more seats does not fix adoption; it multiplies the unused licence line and makes the renewal conversation harder, not easier.
Do this instead
Hold the current seat count, publish the acceptable-use policy, and instrument weekly active use by department for one quarter.
Revisit when
Existing pilot seats reach 50% weekly active use for two consecutive months.
Do not buy 03
Next 12 months
A bespoke model fine-tune or custom LLM build
There is no proprietary, cleanly labelled dataset here that a general model does not already handle. With duplicate records across the ERP and CRM, a fine-tune would learn your data quality problems and make them permanent and expensive.
Do this instead
Use retrieval against the existing documentation. It costs a fraction, ships in weeks, and stays current when the source changes.
Revisit when
A single authoritative source per entity exists and retrieval has been in production for two quarters with a measured accuracy ceiling that retrieval cannot lift.
Do not buy 04
Next 12 months
A vendor-run or consultancy-run AI readiness workshop
You are holding the output of one. Typical quotes run $75K to $250K and end in a maturity grid plus a shortlist of the sponsoring vendor's own products. Nothing in that engagement addresses the governance and ownership gaps identified in this report.
Do this instead
Name the accountable AI owner and put the one-page purchase test in front of the next proposal. Both are free and both are prerequisites the workshop would not have delivered.
Revisit when
Never, in this form. Buy implementation capacity against a defined scope instead of buying an assessment.

Post the Do Not Buy list on LinkedIn

The formatted text below is sized for a LinkedIn post. Copy it, or download it, and put it in front of the people who approve AI spend.

Four AI purchases a mid-market company should refuse this year.

From a real diagnostic structure — the section no vendor, analyst or consultancy will ever write for you, because every one of them is paid on the other side of the decision.

1. An enterprise agentic AI platform (UiPath Autopilot, Salesforce Agentforce or equivalent)
   Why not: Agentic platforms only pay back where the underlying process is documented and the exceptions are understood. Process Maturity scores 48 here and no workflow is written down end to end, so the agents would be automating an undocumented process nobody can audit. Expect $180K to $400K in first-year platform and integration spend with no measurable baseline to prove against.
   Instead: Spend roughly $0 and six weeks documenting the top three workflows. That document is also the specification you would have paid a systems integrator to produce.
   Revisit when: Process Maturity clears 65 and the three highest-volume workflows have a written baseline with cycle time measured.

2. Additional Microsoft 365 Copilot seats beyond the current pilot
   Why not: Seat-level usage on the existing pilot is under 20% weekly active. Buying more seats does not fix adoption; it multiplies the unused licence line and makes the renewal conversation harder, not easier.
   Instead: Hold the current seat count, publish the acceptable-use policy, and instrument weekly active use by department for one quarter.
   Revisit when: Existing pilot seats reach 50% weekly active use for two consecutive months.

3. A bespoke model fine-tune or custom LLM build
   Why not: There is no proprietary, cleanly labelled dataset here that a general model does not already handle. With duplicate records across the ERP and CRM, a fine-tune would learn your data quality problems and make them permanent and expensive.
   Instead: Use retrieval against the existing documentation. It costs a fraction, ships in weeks, and stays current when the source changes.
   Revisit when: A single authoritative source per entity exists and retrieval has been in production for two quarters with a measured accuracy ceiling that retrieval cannot lift.

4. A vendor-run or consultancy-run AI readiness workshop
   Why not: You are holding the output of one. Typical quotes run $75K to $250K and end in a maturity grid plus a shortlist of the sponsoring vendor's own products. Nothing in that engagement addresses the governance and ownership gaps identified in this report.
   Instead: Name the accountable AI owner and put the one-page purchase test in front of the next proposal. Both are free and both are prerequisites the workshop would not have delivered.
   Revisit when: Never, in this form. Buy implementation capacity against a defined scope instead of buying an assessment.

The full sample report is published free, no email gate: https://unconsultancy.app/sample

Matched AI capabilities

Which type of AI, which products, build or buy

Matched to the answers given — not a generic vendor list.

Now — 0 to 90 days

Retrieval-augmented assistant (RAG)

Quote and spec lookup across the product catalogue

Your documentation exists and is reasonably structured; the loss is time spent finding it, not creating it.

Products to evaluate

Microsoft 365 Copilot · Glean · Danswer (self-hosted)

Build or buy: Buy. There is no defensible advantage in building retrieval yourself.
Watch out: Retrieval quality collapses if duplicate records remain across the ERP and CRM — sequence this after the consolidation item.

Document extraction

Supplier invoice and packing-slip intake

High-volume, semi-structured documents with an existing manual keying step and a measurable error rate.

Products to evaluate

Azure Document Intelligence · Rossum · Klippa

Build or buy: Buy, with a human review queue for the first two months.
Watch out: Do not measure success on extraction accuracy; measure it on hours removed from AP.

Next — 3 to 9 months

Conversational support with escalation

Tier-one customer service triage

Repetitive inbound question mix with a documented escalation path already in place.

Products to evaluate

Intercom Fin · Zendesk AI · Ada

Build or buy: Buy. Deflection rate is the only metric that matters at renewal.
Watch out: Deploy behind a live-agent handoff from day one or CSAT will drop faster than cost.

Later — 9 to 18 months

Agentic workflow automation

Order-to-cash exception handling

Rules-heavy multi-system process where most exceptions follow a small number of patterns.

Products to evaluate

UiPath Autopilot · n8n + model API · Workato

Build or buy: Buy the orchestration, build the decision rules — the rules are where your advantage actually lives.
Watch out: Blocked until process documentation exists. Attempting this first is the classic mid-market failure.

Now run it on your own company.

Five minutes, 24 questions, your score free. The full report — everything above, about you — is $149, one time. No account, no call, no follow-up sequence.

Start free assessment