Your business partner is not against new tools. They are against spending money on things that sounded clever in a sales demo and then quietly became another monthly charge nobody checks. If you have been partners for a while, they have probably watched you do exactly that at least once.
So walking into the conversation with "everyone is using AI now" will not work, and it should not. What works is a short, specific document that treats their scepticism as reasonable and answers it with your own numbers.
Below is that document: a one-page template in five parts, with an example filled in and notes on what makes each part persuasive. It is the kind of business case a chatbot or any other small tool should have to pass.
Why one page
Long proposals signal that the idea needs a lot of defending. One page signals that you have thought about it clearly enough to be brief. It also forces you to drop the parts that are really enthusiasm, not evidence.
The five parts, in order:
- The problem, in a sentence your partner would agree with.
- The evidence, from your data.
- The cost, all of it.
- The risks, and how you will notice them.
- The review, with a date and a stop condition.
Part one: a problem they already feel
Do not start with the solution. Start with something your partner has complained about.
Weak: "We should get an AI assistant for the website."
Stronger: "Enquiries that come in after 5pm wait until the next morning, and we have both noticed some of those people have already gone elsewhere."
If you cannot name a problem your partner already recognises, stop here. You may be solving something only you care about, and that is worth knowing before the conversation rather than during it.
Part two: evidence from your own data
This is where most cases fall apart, because they borrow statistics from vendor blogs. Your partner will rightly ignore those. Use your inbox instead. An evening with your sent folder and job records produces three numbers that carry far more weight:
- Out-of-hours share. Of the last 80 website and email enquiries, how many arrived after 5pm or at weekends?
- Reply delay. For those, what was the typical gap before first reply?
- Repeat questions. How many of the last 80 asked something with a standard answer: prices, areas covered, availability, how the process works?
Here is an example with invented numbers, which you should replace. Of 80 enquiries, 29 arrived out of hours, with a typical first reply 14 hours later. Of all 80, 47 asked at least one question with a standard answer. Of the 29 out-of-hours enquiries, 6 went cold after our reply.
Those are four sentences your partner can check. That is the point. Modelling the cost of slow replies shows how to push this analysis further if your partner wants more.
Part three: the cost, including time
Leaving out your own time is the fastest way to lose credibility with the partner who does the books. Put it all in:
| Item | Year one |
|---|---|
| Subscription (for example Starter: $20 a month for the 3-month trial, then $17 a month billed annually) | $213 |
| Setup and first content, 10 hours at our internal rate of $40 | $400 |
| Weekly review, 30 minutes for 50 weeks at $40 | $1,000 |
| Total | $1,613 |
Then compare it with the break-even. If an average job brings $900 of margin, the assistant needs to add roughly two jobs across the whole year to cover itself. That framing is often more persuasive than any ROI percentage, because it is small and concrete. Current plan details are on the pricing page, and the flat price matters here: the bill does not rise in a busy month.
Part four: risks, stated plainly
Naming the risks yourself is what earns trust. Your partner is thinking of them anyway.
| Risk | How we would notice | What we do about it |
|---|---|---|
| It gives a wrong answer | Weekly read of conversations | Correct the content; answers come only from what we give it |
| It annoys customers | Ratings and conversation tone | Adjust the welcome message, or remove it from certain pages |
| Leads pile up unanswered | Leads count vs callbacks made | Daily 10-minute check, owner named |
| Nobody reviews it after month two | Review slot missed twice | Named owner; stop the trial |
| Private details typed into chat | Weekly read | Short retention window; tell visitors not to share sensitive details |
If your business is in a regulated field, add a line making clear the assistant gives information, not advice, and that anything clinical, legal or urgent goes to a qualified person. Guardrails for regulated industries covers the wording.
Part five: the review date that makes it safe
This is the section that usually wins the argument. You are not asking for a permanent commitment. You are asking for a measured trial.
Write something like this: "We run it for 90 days on monthly billing. On 15 December we look at leads captured out of hours, how many became paid work, and how many answers were wrong. If it has not produced at least two paid jobs, or if we have found wrong answers we could not fix, we cancel."
A stop condition turns a debate about whether AI is good into an agreement about what counts as evidence. Sceptical partners tend to say yes to that.
Two details make the review work in practice. First, name who gathers the numbers before the meeting, so the review is not spent searching the inbox. Second, agree now what "paid job from the assistant" means: a lead first captured in chat, not someone who chatted after already ringing you. Settling that definition in September saves an argument in December.
Handing it over
Hand the page over rather than presenting it. Give your partner a day with it. Then ask one question: which line would you change? That invites them to improve the proposal instead of defending a position against it, and the edits they make usually improve the review section, which is exactly where you want their scepticism applied.
Put it on paper tonight
Open a blank document and write the five headings. Fill part two first, from your own inbox, because the evidence decides whether the rest is worth writing. Keep it to one page. If you want the cost section grounded in a proper estimate, run your figures through the chatbot ROI calculator and paste the result in. Then hand it over with a proposed review date already filled in, and let your partner change it.