A small consultancy can turn its method into a fixed-scope, AI-assisted diagnostic by choosing one problem, listing the questions its method always asks, collecting the client's evidence through a fixed intake, using AI for first-pass analysis, and spending the saved hours on senior review. Price it on the value of the answer, with a fixed number of review hours.
The reason this sells is not the AI. It is the fixed scope: a buyer who cannot judge a consultant's hours can judge a defined deliverable, a start date and a price. AI makes the defined version profitable at a price a small client accepts. Compare two lines in a proposal: "eight to twelve days at our day rate" and "a profitability report in three weeks for $3,500". The second is easier to approve, because the buyer knows the ceiling before saying yes.
The example: a consultancy that audits job profitability
An illustration, with figures made up for the example and meant to be swapped for your own. A three-person consultancy helps trade businesses understand which jobs make money. Today it sells open-ended engagements at a day rate: a discovery call, a few weeks of spreadsheet work, a report. Projects run long, clients hesitate over an unknown bill, and each project is built from scratch.
The owner decides to package the first stage as a "Job Profitability Diagnostic" for electrical contractors and plumbing firms with 5 to 25 staff. The offer: in three weeks, the client receives a report showing profit by job type, the five jobs or customers that lose money, and three ranked fixes, plus one 60-minute walkthrough. Price: a single fixed fee.
Step 1: choose a question the client already feels
A diagnostic must answer a question the buyer would lose sleep over. "Digital transformation readiness" fails that test; "which of our jobs lose money?" passes. Good diagnostic questions share three traits: the client knows the symptom (cash is tight despite a full diary), the evidence exists in records they already hold, and your method can produce a ranked answer without months of interviews.
Test a candidate question with this short check: can you name the three files you would ask for, and could a competent assistant read them in an afternoon? If the answer needs interviews with twenty staff, the diagnostic is too big; narrow it to one function or one process.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowStep 2: write the scope as a list of included and excluded items
Scope creep in fixed-fee work comes from vague deliverables. Write them as plain statements the client can tick off.
| Included | Excluded (quoted separately) |
|---|---|
| Review of up to 12 months of job and invoice data, from one system | Cleaning up or re-entering data |
| Profit by job type and by customer segment | Pricing changes across the full price list |
| Two 45-minute interviews: owner and office manager | Interviews with every engineer or fitter |
| Written report of up to 12 pages with three ranked fixes | Implementing the fixes |
| One 60-minute walkthrough call within two weeks of delivery | Further revisions after the walkthrough |
The right-hand column protects you and reassures the client; they can see exactly where the next paid step begins. It is also where your follow-on work comes from, so list the likely implementation options without pricing them.
Step 3: design the intake so the evidence arrives ready to use
Intake is the unglamorous centre of a profitable diagnostic. Half the hours in an open-ended engagement go on chasing files in odd formats. A fixed intake removes that.
- A one-page request list naming each file, the system it comes from, the date range and the columns you need: job reference, job type, customer, labour hours, materials cost, amount invoiced, date completed.
- A secure upload route agreed in the proposal, not email attachments.
- A deadline. The three weeks start when the complete intake arrives, not on signing. State that in the offer.
- A short questionnaire for the owner: how jobs are quoted, who sets prices, what happens to overruns. Ten questions, answered in writing before the interview.
The questionnaire answers feed the AI step as context. Say the owner writes: "We quote on a rough day rate and add 10% if it looks tricky." That sentence tells the model what pricing rule to test against the data.
A filled-in intake request list
The blank version is easy to write; the filled-in version is what clients can act on. An illustrative request list for the trade-business diagnostic:
| File | Source | Range | Columns needed |
|---|---|---|---|
| Job list | Job management system export (CSV) | Last 12 months | Job ID, job type, customer, date completed |
| Time records | Timesheet export | Same | Job ID, hours, staff grade |
| Materials | Supplier invoices or purchase ledger | Same | Job ID (if recorded), cost |
| Invoices | Accounts package export | Same | Job ID, amount invoiced, date paid |
| Price list | Current quoting sheet | Current | Day rate, call-out fee, mark-up rules |
The materials row will usually be the weak one; many small firms do not record materials against job IDs. Say so in the offer: where materials are not tied to jobs, the diagnostic estimates them from a sample of 30 jobs and labels the result as an estimate. A stated limitation is better than a surprise on day six.
Step 4: use AI for first-pass analysis, with rules
This is where the hours come back. Four jobs suit AI well in a diagnostic, and each has a human check.
Cleaning and structuring the data
Client exports are messy: job types spelt six ways, blank cost fields, duplicates. Ask an assistant with file analysis to propose a mapping of job-type spellings to a clean list of categories, then check the mapping yourself before applying it. In the illustrative example, 640 jobs carry 38 distinct job-type labels that collapse to 9 categories. The assistant proposes the groupings in minutes; the consultant spends 20 minutes checking the ten ambiguous labels.
Calculating and ranking
Profit per job type is arithmetic, and a spreadsheet does it better than a chat assistant. Use the assistant to write the formulas or a pivot-table recipe, then let the spreadsheet calculate. Never take figures from a model's prose summary of a file; take them from the spreadsheet. For the formula route, see how to use ChatGPT with Excel.
Drafting findings from the numbers
Once the figures are checked, paste the summary table into the assistant and ask for plain-language findings.
Below is a verified profit-by-job-type table from a client's last 12
months. Draft three findings for a business owner, each with: the
finding in one sentence, the evidence (quote figures from the table
only), and one likely cause to investigate. Do not add any number
that is not in the table. Flag anything that needs the owner's
explanation.
[TABLE]A realistic output: "Finding 1. Emergency call-outs make up 14% of revenue but 31% of the jobs that lost money. Evidence: 41 of 132 losing jobs. Likely cause: the call-out fee does not cover travel and parts handling." You verify "41 of 132" against the table and replace "likely cause" with what the interview told you. A model's cause is a hypothesis, never a conclusion.
Testing the owner's pricing rule
Take the rule from the questionnaire ("day rate plus 10%") and ask for a check against the data: how often did actual hours exceed quoted hours, and by how much? Again, you compute in the spreadsheet and use the assistant for the narrative. In the example, jobs quoted at under one day overrun by 28% on average; jobs over three days overrun by 9%. That is the sort of result a client remembers.
Keep client data safe throughout: use a plan whose terms exclude training on your content, remove customer names before upload where the analysis does not need them, and say in the proposal which steps use AI. The guidance in how to stop AI tools training on your business data covers the settings, and how to mention AI use in client contracts covers the wording.
Step 5: set the review hours before you set the price
The saved hours should not vanish into margin alone; they buy the review that makes the diagnostic trustworthy. Plan the hours in the offer design, not by feel.
| Stage | Without AI (hours) | With AI (hours) | What the saved time buys |
|---|---|---|---|
| Intake and data cleaning | 6 | 3 | Checking category mapping line by line |
| Analysis in the spreadsheet | 6 | 4 | A second method to cross-check totals |
| Interviews (two) | 3 | 3 | No change; this stays human |
| Drafting the report | 8 | 4 | A full read-through for unsupported claims |
| Senior review and walkthrough | 3 | 5 | The most valuable hours in the offer |
| Total | 26 | 19 |
All the numbers are illustrative. Notice what happens to the hours: total effort falls from 26 to 19, but senior review rises from 3 to 5. The AI step does not shave quality off; it moves hours from typing to judgement. That is the story to tell a client when asked what AI changes.
Step 6: price on the answer, then sanity-check against hours
Price logic has two sides. The value side: what is the client's problem worth? If the business turns over $1.2 million and the diagnostic finds that 14% of jobs lose money, even recovering a small part of that dwarfs the fee. You do not promise recovery, but you do price relative to the stakes rather than to your hours. The cost side: your hours, plus a margin for risk.
A simple method:
- Take your internal cost: 19 hours at your blended internal rate. At an illustrative $90 an hour that is $1,710.
- Add the tool cost, which is tiny: an individual plan from a major assistant is listed at $20 a month, and the cost per diagnostic is a few dollars.
- Add a risk and non-billable allowance of 30%, for scope questions, rescheduling and the sales effort: about $510.
- That gives a floor of about $2,220. Now look at the value side and decide a price above the floor, say $3,500 for a client with $1.2 million turnover.
Tiering by client size is easier than a custom price. A "Starter" for up to 10 staff, a "Standard" for 11 to 25, each with a fixed fee, means you quote in a minute. Do not discount for AI speed; if you need a reason to hold the price, point to the review hours and the fixed deliverable. The wider question is covered in whether to charge clients less when AI speeds up your work.
Step 7: write the offer page and the proposal
A diagnostic sells from a one-page description. Include: the question it answers, who it is for, what arrives and when, what is excluded, the fee, and how AI is used and checked. Keep the AI paragraph short and factual: "We use AI tools to speed up data preparation and first drafts. Your data goes only into business-grade tools that do not train on it, and every finding is checked by a senior consultant before you see it."
The proposal can be drafted from that page in minutes; see how consultants use AI to write proposals in under an hour for the drafting side. Add a one-line next step: "Book a 20-minute scoping call; if the diagnostic is a fit, we start on receipt of the intake."
The scoping call: what to ask in 20 minutes
A free 20-minute call decides whether the diagnostic fits. Run it from a fixed list, so every prospect is judged the same way and you never improvise scope on the phone.
- What made you look at this now? (The symptom, in their words.)
- Which system holds your job and invoice records, and can you export to CSV?
- Who would collate the files, and by when?
- Has anyone looked at this before? What did they conclude?
- If we found the answer, who decides what to change?
The last question filters out prospects who want a report to file rather than act on. Decline politely if the files do not exist, if nobody owns the decision, or if the data sits on paper. A diagnostic built on guesses damages your reputation more than a lost sale does.
Realistic pushback and replies: "Can you just start and send an invoice later?" No: the fee and the start of intake are agreed together. "Can you also interview all the engineers?" That is a separate quote; the diagnostic includes two interviews. "Why not hourly?" Because you cannot judge my hours, but you can judge a report with a date and a price.
Before and after: rewriting a vague offer line
Many consultants describe the offer the way they describe their method. The change that helps most is to describe the result and the boundaries.
Before: "We take a broad look across your operations and identify opportunities for improvement using data and AI."
After: "In three weeks you receive a 12-page report showing profit by job type, the five customers or job types that lose money, and three ranked fixes. Fee: $3,500. You supply five exports; we supply the analysis and a 60-minute walkthrough."
The second version names a deliverable, a time, a price and each side's job. A buyer can compare it with an hourly quote and see it is cheaper to say yes. It also gives you something to measure: did the report arrive on day 21 with the five items?
Tiers and follow-on work
A single fee is simplest; two or three tiers avoid underpricing larger clients. Illustrative tiers for the example:
| Tier | Client size | Data scope | Interviews | Fee |
|---|---|---|---|---|
| Starter | Up to 10 staff | One system, 12 months | 2 | $2,800 |
| Standard | 11 to 25 staff | One system, 12 months | 2 | $3,500 |
| Extended | 26 to 50 staff | Up to two systems, 24 months | 4 | $5,200 |
The offer built above sits in the Standard tier; Starter is cheaper for firms of 10 or fewer because their data takes fewer hours to clean, and Extended adds a second system, a longer period and two more interviews. Follow-on work sits behind the diagnostic: a pricing-rule redesign, a quoting template with AI-assisted estimates, a monthly profitability dashboard. List them as "typical next steps" in the report with a rough effort band, so the client knows what comes next without feeling sold to. Price the diagnostic so it earns your target margin on its own; the follow-on is where the relationship grows, and a first stage that delivered exactly what it promised makes the second easier to sell.
A second example: a cleaning company with a quality problem
To show the pattern transfers, imagine the same consultancy builds a second offer for cleaning companies: a "Repeat-Complaint Diagnostic". The question: why do some sites generate repeat complaints? The intake is 12 months of complaint logs, rota records and site checklists. AI clusters the complaint text into themes (missed areas, late arrival, supplies), the consultant checks 30 of the clustered complaints by hand to confirm the groups make sense, and the report ranks three causes. The structure is identical: one question, a short intake, AI for first pass, human review, fixed fee. Only the files and the vocabulary change.
How things go wrong in these offers
Three failures recur. First, the intake arrives incomplete and the clock runs anyway; fix this by starting the three weeks on complete intake. Second, the AI summary contains a figure that is not in the data; the rule that every number comes from the spreadsheet, never from model prose, prevents this. Third, the client wants the diagnostic to also fix the problem; the excluded list and the follow-on options answer that politely.
How to check the offer is working
After three diagnostics, review: average actual hours against the planned 19, the share of scoping calls that become paid diagnostics, and how many clients buy the follow-on work. If hours overrun by more than 20%, the intake is leaking time. If under a third of scoping calls convert, the question or the price is wrong. A diagnostic that sells and delivers on time three times running is a product; record the method as a checklist so a junior can run the data stage while you keep the judgement.
Questions consultants ask when packaging a diagnostic
How long should a fixed-scope diagnostic take to deliver?
Most small-consultancy diagnostics fit into two to three weeks of elapsed time and 12 to 25 hours of your own work. Shorter than that and the client does not feel the depth; longer and the scope starts to creep. Fix the elapsed time in the offer, not only the hours.
Should I tell the client that AI is used in the analysis?
Yes, in plain words in the proposal: which steps use AI, that nothing client-confidential goes into a consumer tool, and that you review every finding. Saying it up front avoids an awkward conversation if a client finds out later and wonders what else was not mentioned.
Can I charge the same price if AI makes the work faster?
Price on the value of the answer and the risk you carry, not on your hours. A sensible default is to keep the fee and put part of the saved hours into a deeper review. If a client asks, explain what the fee covers rather than apologising for efficiency.
What stops a diagnostic becoming free consulting?
A written list of what is out of scope, a fixed number of review calls, and a clear end point: the report and one walkthrough. Anything beyond that, such as implementation or extra interviews, is a separate quote.
Further reads
- How to Write Business Proposals Faster With AI — Faster proposals once the offer is defined.
- AI Readiness Checklist for Accountants, Solicitors, Consultants — A readiness checklist you could adapt as an intake form.
- What an AI Consultant Does for a Law Firm, and What It Costs — How another sector prices AI consulting work.
- AI Proposal Software vs a General AI Assistant: Which to Pay For — Proposal software against a general assistant.
- How to Evaluate an AI Implementation Proposal or Quote — How a client reads your offer from the other side.
- AI Hallucinations Explained for Business Owners: Causes and Fixes — Why every AI-assisted finding needs a human check.
- Best AI Model for Business Consulting: Pick by Task, Not by Hype — Three consulting tasks to run through ChatGPT, Claude and Gemini, with long-document limits, client confidentiality and plan costs checked on vendor pages.
- Consultant SEO for ChatGPT Search: Get Your Practice Recommended — Make a consulting practice easy for AI search to identify and cite: entity pages, case studies, reviews, schema and a fixed prompt test you rerun monthly.
- Consulting Engagement Letter: Scope, Fees and AI-Use Wording — A clause-by-clause engagement letter for independent consultants: scope, fees, change control, confidentiality and plain AI-use wording you can adapt.
- Consulting Utilisation Rate: Track It Weekly and Price Projects With AI — Billable versus available hours, bench time, a weekly AI roll-up from timesheets and a worked project price checked against your rate card.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: ChatGPT and Claude pricing pages (ChatGPT Plus and Claude Pro at $20 a month list price) and their business-plan training terms, checked 4 November 2026; the example figures are illustrative, not benchmarks.