Lesson 3 · 18 min
The client can't judge whether your technical solution is good: they've never rolled out AI. What they judge is whether you look like you know what you're doing, and they read that in your method. This lesson shows the method (survey, measure, pilot, scale) and the pricing in steps.
Start the lesson
Goal: map the current state, measure the baseline, and put together the one-page proposal with 3 options in steps.
Trust grows with each step delivered on time, before any technical result.
Models get discontinued. Monitor what you delivered in a monthly advisory, or give the client the list of what's in use and what to do when it changes.
Dollar ranges as an order of magnitude for the international market. Adapt to your market.
A shared space where you answer questions once a week.
It holds the client who had one or two conversations and would disappear without it.
Always with prep work beforehand: you show up knowing the industry and the likely stack.
Bundle it when it makes sense: one session plus five questions by email in the month.
Weekly meeting, email support in an agreed format, team training, and priority on your calendar.
This is the step that turns consulting into predictable work.
Several departments, a long timeline, and specialists subcontracted for what isn't yours.
Here you answer for the result. The build, when there is one, is just one part of it.
Write the steps on a single page, for yourself. Improvising a price in the middle of the conversation is how the discount is born.
Let the client build the plate instead of choosing between packages they don't understand.
Starters (beginner), quick dishes (practical wins), intermediate, advanced, desserts (executive level) and banquets (closed packages).
Each item has just four fields: name, duration (90 to 150 min), level and a short description of the result, never of the tool.
The client stops asking "how much does it cost" and starts saying "I want number 3, number 4 and the light version of number 8."
Start with the pain, in their own words, and close with the goal repeated out loud.
Then: scale, stack and where the data lives.
Play back what you understood and present 2 or 3 hypotheses, stated as hypotheses.
In the recommendation, say why it may not be realistic for their business and show cheaper or faster alternatives.
Offer the steps and leave with one task: list the tools with the number of users, choose the pilot department or name an internal owner.
The twenty minutes left are for the question they saved for the end.
The proposal fits on one page. The baseline comes before any target number.
PROPOSAL · <company> · <date> 1. WHAT WE HEARD <the pain in the client's own words, in 2 lines> 2. BASELINE (measured on <date>, by <how it was measured>) - <indicator>: <current value> → target <value> in <timeframe> 3. THREE WAYS TO WORK TOGETHER A) Access · <amount/month> Weekly Q&A group. For those who want to start on their own with support. B) Pilot · <amount> · <30 to 60 days> A group of <n> people in <area>, one metric, report at the end. C) Project · <amount> · <timeframe> Pilot + scale-up to <department> + <n> months of follow-up. 4. WHAT IS NOT INCLUDED <e.g.: integration with system X; running campaigns> 5. RISK AND TRADE-OFF <what may not work and what the company gives up with this choice> 6. NEXT STEP <client's task> by <date>. Next conversation: <date>.
Indicator: <e.g.: time to answer a quote request> Current value: <number and unit> Target: <number> in <timeframe> How it is measured: <who measures it, where the number comes from, how it is counted> Review frequency: <weekly or monthly; never daily>
You are a senior AI consultant building the service menu for your own business. Write in Brazilian Portuguese, direct tone, no sales adjectives. My profile: <what you can really do today> My clients: <industry, size and job title of the people who hire you> What I do NOT do: <list it, so it stays off the menu> Build a menu with 6 categories: starters (beginner), quick dishes (practical wins), intermediate, advanced, desserts (executive level) and banquets (closed packages). At least 30 items in total. For EACH item, exactly 4 fields: - name (short, no jargon) - delivery duration, between 90 and 150 minutes - level (beginner / intermediate / advanced / executive) - description in at most 2 lines, stating the result, not the tool Prioritize diagnosis, teaching, review and follow-up; building only as part of one of these. End with: "Menu updated in <month and year>".
Use a real client or the clinic case from lesson 2.
Pick one of the client's processes (or the simulated clinic's) and fill in the baseline template for 2 indicators.
Done when: both indicators have a current value as a number, a target with a deadline, and the way of measuring written down.
Fill in the proposal template with your baseline and the 3 options in steps. Write down what is not included and the trade-off of the recommended option.
Done when: the proposal fits on one page, has the baseline as a number and 3 options (access, pilot, project) with price and timeframe.
4 questions. The explanation appears after your attempt.
Question 1 of 4
You are going to propose a project and you don't know the client's budget yet. What do you do?
Question 2 of 4
Why measure the baseline before the pilot?
Question 3 of 4
Which measurement frequency avoids distortion?
Question 4 of 4
Where should you start the pilot?
Pick an answer. The explanation shows up after you try.