Prepare by picking your three most repetitive processes and bringing numbers for each: weekly volume, minutes per item, who does it, which tools and plans are involved, and what a mistake costs. Send that sheet a day ahead, pick one outcome for the call, and leave with written actions, each with an owner, a date and a running cost.
Without numbers, even a good consultant can only give general advice, along the lines of "try AI for your emails". With them, the options can be ranked during the call by hours saved, cost to run and risk. The preparation below works for a free call or a paid one, with me or with anyone else, and takes about two hours spread over a week.
A week before: choose the processes worth the airtime
A one-hour consultation can cover three processes properly, or ten badly. Start with a longlist of everything repetitive, then keep the ones that score highest on three questions: how often does it happen, how long does each one take, and what goes wrong when it's done badly? Anything that happens less than weekly usually drops off, however annoying it is.
A six-technician nail salon might start with this longlist and keep the top three (illustrative figures):
| Process | How often | Minutes each | What goes wrong | Keep? |
|---|---|---|---|---|
| Answering booking enquiries in Instagram DMs and WhatsApp | About 220 a month | 3 | Slow replies lose bookings to other salons | Yes |
| No-shows and late cancellations | About 54 a month | n/a | Empty chairs, lost revenue | Yes |
| Replying to Google reviews | About 30 a month | 4 | Unanswered reviews, inconsistent tone | Yes |
| Weekly stock order for polish and gels | 4 a month | 30 | Occasional over-ordering | No: small and already fine |
| Staff rota | 4 a month | 45 | Clashes in school holidays | No: needs judgement, not AI |
| Annual price-list update | Once a year | 180 | Old prices left on the website | No: too rare |
If you'd like a more systematic way to build the longlist, the method in auditing your workflows for AI opportunities yourself works well before a consultation.
The numbers to gather, and the quickest way to get each
You don't need a spreadsheet project. Each figure below takes minutes if you know where to look.
- Volume. Count last month, not an average you remember. Booking apps report appointments and no-shows; inbox and DM counts can be taken from a search for a typical phrase; till or card reports give transaction counts. If your trade is seasonal, bring a busy month and a quiet one.
- Time per item. Time five real instances with your phone's stopwatch and take the middle figure. Estimates from memory are often well out in one direction or the other.
- Who does it, and what their hour costs. Name the role, and use pay plus roughly a fifth for holiday and other on-costs as a working figure unless you know yours.
- Tools and exact plan names. Open each billing page and write down the plan, seat count and renewal date. If you use Microsoft 365, the admin centre has a Copilot usage report showing active users over 7, 28, 90 or 180 days. Your current spend is easiest to pull together with an audit of your AI subscriptions.
- The cost of a mistake. One number per process: the value of an empty slot, the cost of a refund, the time to fix a double booking.
The salon's no-show line shows why that last number matters. About 900 bookings a month with a 6% no-show rate is 54 empty slots; at an average service of $45 that's about $2,430 of booked time a month. If better reminders cut no-shows by a third, that's roughly $810 a month back, which immediately tells the consultant this process deserves more attention than the reviews, which take about two hours a month.
Plan names matter more than owners expect. A common surprise: "we have Copilot" turns out to mean Copilot Chat (Microsoft now calls it Microsoft Copilot Chat), which is included with Microsoft 365 business plans at no extra cost, rather than the paid Microsoft 365 Copilot licence, which is $21 a user a month on an annual plan. Or "we have ChatGPT Business" turns out to be one owner's personal Plus subscription on a company card. Both change the advice, so check before the call, not during it.
Put a rough monthly value on each process
With the figures in hand, one line of arithmetic per process tells you, and the consultant, where the money is: hours a month times the cost of those hours, plus the monthly cost of mistakes. For the salon, take $19 an hour for the front desk (pay of $16 plus about a fifth for on-costs) and $40 for the owner, whose hour could otherwise be a paid appointment:
| Process | Time cost a month | Mistake cost a month | Rough monthly value |
|---|---|---|---|
| Booking enquiries | 11 hours x $19 = $209 | About 5 lost bookings x $45 = $225 | About $434 |
| No-shows | Little staff time | 54 x $45 = $2,430 of booked time | Up to $2,430 (you'll never recover all of it) |
| Review replies | 2 hours x $40 = $80 | Hard to price | About $80 |
These are rough, and a good consultant may challenge them, which is the point. Nobody would spend the first half of the call on review replies after seeing this table. Flag guesses as guesses: the "5 lost bookings" came from counting last month's enquiries that went quiet after a slow reply and never booked, which is a reasonable estimate but not a fact.
A one-page pre-read, filled in for the salon
Send this a day before the call. It lets the consultant think about your business in advance, and it forces you to find the gaps in your own numbers.
AI consultation pre-read: [salon name], 6 technicians, 1 front desk
Goal for the call: decide which of processes 1-3 to tackle first,
and what each would cost to run every month.
1. Booking enquiries (Instagram DMs and WhatsApp)
Volume: about 220 a month (counted from last month)
Time: 3 minutes each (timed 5) = about 11 hours a month
Who: front desk, 25 hours a week
Tools: Instagram, WhatsApp Business app, [booking app, plan name]
Problem: on busy days replies take 3-5 hours; some enquirers book elsewhere
2. No-shows and late cancellations
Volume: about 54 a month (6% of about 900 bookings)
Cost: average service $45 = about $2,430 of booked time a month
Current set-up: one reminder text from the booking app, 24 hours before
3. Google review replies
Volume: about 30 a month, 4 minutes each = about 2 hours a month
Who: owner, usually late in the evening
Tools and plans:
- Google Workspace Business Starter, 3 users
- ChatGPT Plus on the owner's personal account, $20 a month
- [booking app], [plan], text-message credits bought monthly
Constraints: no client phone numbers in personal AI accounts;
owner approves anything customer-facing; busiest weeks are in December.Notice what the pre-read leaves out: no request to "explain AI", no list of tools you've read about, no ten processes. One page, three processes, real figures.
What to attach to the pre-read, and what to keep back
A few real examples help more than any description. Attach:
- Five typical enquiries or messages, with names, phone numbers and anything personal replaced by placeholders such as [first name].
- The templates or saved replies you use now, even the ones you're unhappy with.
- A screenshot of the relevant settings screen, such as the booking app's reminder settings, so the consultant can see what's already possible without buying anything.
- Your current policies in writing: cancellation terms, deposit rules, refund policy. AI drafts tend to go wrong where a policy was never written down.
Keep back full customer exports, passwords and anything the consultant doesn't need in order to give advice. A first call rarely needs more than examples. If the conversation turns into a project, sharing data becomes a separate, deliberate step with an agreement in place first.
Pick one outcome before you dial in
The simplest way to leave with a plan is to decide in advance what the plan is about. Compare two versions of the same owner's goal:
Vague: Learn how we could use AI in the salon.
Specific: Decide whether to automate DM replies or no-show reminders first, find out what it would cost each month to run, and agree what I can set up myself this week.
The vague goal produces an interesting hour and no decisions. The specific one gives the consultant a job to do and gives you a way to judge the call afterwards. For a sense of how a well-run session is usually structured, see what happens in a 1:1 AI implementation consultation. If the call is a free introductory one, the outcome may be smaller, and how to use a free AI consultation well covers what's realistic to expect.
Questions worth asking in the session
Keep these on a card and tick them off. They work with any consultant.
About the recommendation
- Of these processes, which would you do first, and why that one?
- What would make you change that recommendation?
- What would you not automate here, and why?
- What's the smallest version we could test in a fortnight?
About cost and effort
- What would it cost to run each month at our volumes, including per-message, per-task or usage fees?
- Can the tools we already have do any of this before we buy anything new?
- How many hours of our staff's time does it need to set up, and to keep running?
About data and risk
- What customer data would leave our systems, and where would it go?
- Which of our current accounts should change plan or settings first? For instance, a consumer ChatGPT account used for work should have its model-training switch turned off, or move to a business plan, which doesn't train on business content by default.
- What happens when the automation fails, and how would we find out?
About doing it yourself
- Which parts could we set up ourselves, and where would you want to be involved?
- What access or data would you need from us if we went ahead? (The answers usually match the list in what an AI consultant needs from you.)
Halfway through, check the clock against your goal. If thirty minutes have gone and the processes still haven't been ranked, say so: "Can we decide the first one now and spend the rest of the time on how?" Consultants generally welcome that. Their job is to get you to a decision, and yours is to keep the call pointed at one.
The last ten minutes: write the plan while everyone is still there
Stop the discussion ten minutes before the end and write the actions together. Each line needs an owner, a date, a monthly running cost and a way to tell if it worked. The salon's plan at the end of a good call might look like this (illustrative):
| Action | Owner | By | Running cost | How we'll know |
|---|---|---|---|---|
| Add a second reminder text 2 hours before each appointment | Owner | Friday | Extra text credits; check usage | No-shows under 5% by the end of next month |
| Turn off the training setting on the owner's ChatGPT Plus, or price two Business seats | Owner | This week | $0, or about $40-$50 a month for two seats | Screenshot of the setting saved |
| Write 15 saved replies for the most common DM questions | Front desk and owner | Two weeks | $0 | Six in ten DMs answered from a saved reply |
| Quote for automating rebooking nudges | Consultant | Ten days | To be quoted, with running costs | Written quote received |
| Decide whether AI should draft review replies | Owner | After month one | Not yet known | Decision recorded either way |
Two things make this plan useful. Every action has a named person, and the first three need nothing new to be bought. A plan where every line starts with "buy" or "build" deserves a second look.
Turning rough notes into an action list with AI
If you took notes rather than writing the table live, an AI assistant can tidy them, as long as you stop it filling gaps with guesses. A prompt that does that:
Below are my rough notes from a consultation call. Turn them into an
action list with four columns: action, owner, deadline, monthly cost.
Rules:
- Only include actions that were actually agreed in the notes.
- If an owner, date or cost wasn't stated, write "not agreed".
- List any open questions separately at the end.
- Don't add advice of your own.
Notes:
[paste your notes, with client names and numbers removed]An illustrative output, with the problems a careful reader would catch:
1. Automate rebooking nudges. Owner: consultant. Deadline: 10 days. Cost: $49 a month.
2. Switch to ChatGPT Business. Owner: salon owner. Deadline: Friday. Cost: $50 a month.
3. Write saved replies for DMs. Owner: front desk. Deadline: 2 weeks. Cost: $0.
Line 1 is wrong: the notes said the consultant would send a quote, and the $49 figure appears nowhere in them; the assistant invented it. Line 2 turned "consider switching, or turn off the training setting" into a decision. Line 3 is fine. The fix is to check every number and every verb against your notes, and to re-run the prompt with "Mark anything you are unsure about with [CHECK]" added. Even with the rules, treat the output as a draft.
Adjusting the prep for seasonal trade or a one-person business
A garden centre's figures swing so far between spring and winter that a single month misleads. Bring two columns for each process, a peak month and a quiet one, and say which the consultant should design for. Designing for the average leaves you swamped in April and paying for capacity you don't use in January. Click-and-collect confirmations might run at 400 a month in April and 60 in December, and an automation billed per task costs very different amounts in each, so the running-cost question needs both figures.
For a solo personal trainer, the "who does it" column is always you, so the useful number is what your hour is worth in paid sessions. If an hour of admin could have been an hour of coaching, the value of saving it is the session fee, not a wage. The pre-read also gets shorter: two processes rather than three, and a goal such as "decide what I can automate in my booking app this month without paying for anything new".
How prepared owners still leave empty-handed
These are the realistic ways a call goes wrong even after preparation, and each has an easy fix:
- Mixed time periods. The salon owner pulled DM counts from a 90-day insights screen and labelled them monthly, tripling the apparent workload. Label every figure with its period.
- The person who does the work isn't there. The front desk knows which enquiries are awkward, which clients never read texts and which questions come up at 9pm. Bring them for the first twenty minutes if you can.
- Too many processes. Ten processes means ten shallow conversations. Three is plenty.
- Half the hour spent comparing tools. Tool choice comes after deciding what the job is. If the consultant drifts into product demos, steer back to your pre-read.
- No decision-maker on the call. If a partner or manager has to approve spending, the plan stalls for a week. Either include them or agree in advance what you can decide alone.
Within a day: test the plan against your own numbers
Before the call fades, reread the action list against your pre-read. Does each action connect to one of your three processes? Is the biggest cost (the salon's $2,430 of no-shows) getting the first action? Are the running costs written down, even as ranges? If something doesn't add up, send one short email asking about it while the conversation is fresh.
Then put the actions in your calendar. A plan that lives only in a notes app tends to stay there. The tutorial on turning consultation advice into a 30-day plan takes it from here, week by week.
Further reads
- Is a 1:1 AI Consultation Worth It for a Small Business? — Whether a 1:1 call is worth paying for at your size.
- How to Choose Your First AI Project: 7 Tests Before You Commit — Seven tests for picking the first project from your shortlist.
- How to Write a One-Page AI Business Case for a Small Business — Turn the plan you leave with into a one-page business case.
- Automation Audit: Find the Zaps and Scenarios Nobody Owns — List the automations you already run before the call.
- Course, 1:1 Call, Audit, or Project: Which AI Help Do You Need? — Course, call, audit or project: check a call is the right format.
- How to Cut Salon No-Shows With AI Reminders and Deposits — The salon no-show problem from the worked example, solved in detail.
- Do I Need an AI Consultant? 8 Signs It's Time to Get Help — Eight signs a small business needs outside AI help, each with a test you can run this week, plus the signs that it doesn't need a consultant yet.
- What to Prepare Before You Set Up AI Marketing — An eight-part preparation checklist, filled in for a farm shop, with a fact-sheet prompt and a one-page brief you can copy.
- How to Choose an AI Consultant: 20 Questions to Ask First — Twenty questions to put to any AI consultant, what strong and weak answers sound like, and a scoring sheet filled in for a farm shop.
- Where to Find a Good AI Consultant for a Small Business — Which channel suits which job, what each directory and marketplace offers, and how a community interest company turned nine names into a shortlist of three.
- How Much Does an AI Readiness Assessment Cost? — Free, vendor-funded and paid readiness assessments compared, where the days go, your staff's share of the cost, and a way to compare three quotes fairly.
- What Is the Cheapest Way to Get Expert AI Advice? — Free help pages, forums, adoption kits and discovery calls answer most AI questions. When paid advice is worth it, and how to compare quotes.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Microsoft 365 admin centre documentation on the Copilot usage report; Microsoft 365 Copilot Business and Copilot Chat plan pages; ChatGPT Business plan page; consumer privacy settings pages for ChatGPT and Claude.