Preparing a Client Meeting with AI: Research, Questions and Objections
By Anis Hammouche·October 29, 2026
To prepare a client meeting with AI, you give it three tasks: summarise what is public about the prospect, suggest questions to ask, and write the likely objections. You stay in charge of the source material, of checking the facts, and of anything confidential. The AI drafts, you prepare.
What AI does well in a preparation
Preparing a meeting mostly takes sorting and formatting. AI handles both quickly. It can turn ten pages of website into three paragraphs, list the topics covered in an article, propose a dozen questions grouped by theme, or play a sceptical prospect so you can practise.
It does less well in two areas. It does not know the company's current situation unless you give it the source. And it can invent a plausible detail, a job title or a date, with the same tone it uses for a correct one. The whole method rests on one simple rule: you provide the material, it structures it.
Step 1: gather the material, without personal data
Start with what you can find yourself in a few minutes: the prospect's home page, their product or services page, a recent press release or article, a published job offer (it often reveals current priorities). Note the date of each source. A news item from two years ago no longer describes the company.
Public information remains personal data as soon as it concerns a natural person. The CNIL recommends defining the authorised and forbidden uses of a generative AI tool, and keeping personal data out of prompts (CNIL, questions and answers on generative AI). In practice, you do not paste a full LinkedIn profile, a list of contacts or another client's data. You paste the company's public text, not information about its employees.
Here is an example of an instruction once the material is gathered:
You are preparing a discovery meeting for a freelance web developer.
Below is the public text of the prospect's home page and services page.
1. Summarise in five bullet points what the company does, who it serves and what it sells.
2. List what seems to be a priority for them, quoting the sentence from the text that shows it.
3. Flag anything that is not in the text and would need to be confirmed.
Do not complete anything from your own knowledge.
The last instruction matters. It forces the tool to separate what it reads from what it assumes.
Step 2: prepare the questions to ask
A good meeting question gets the prospect talking about their problem, not about your solution. It covers the context, the current situation, the cause of the problem, who makes the decision, the timeline, and what has already been tried.
Ask the AI for a list grouped by theme, then keep only the questions that fit how you actually work. Twenty questions are useless in a meeting. You need five or six, well chosen, each with a follow-up question.
| Theme | Example question | What it reveals |
|---|---|---|
| Context | "What made you decide to talk about this now?" | The real trigger |
| Current situation | "How does it work today, step by step?" | The process and its friction points |
| Decision | "Who else needs to approve this project?" | The real decision makers |
| Timeline | "Is there a deadline that matters to you?" | Any urgency |
| Past attempts | "What has already been tried, and why didn't it hold?" | Blockers and expectations |
Step 3: anticipate the objections
Objections are predictable. Price, lack of time, an existing provider, doubt about the result, fear that the team will not adopt the tool. Ask the AI to list them for your type of service, then write the answer to each one yourself. This is where credibility is built, and an answer written by a machine often sounds hollow.
To practise, ask the tool to play the prospect and raise these objections one by one. Answer aloud or in writing, then ask for a critique of your answer: is it precise, does it address the real worry, does it promise something you cannot deliver?
What to check before you walk in
Everything the AI writes must be verified. The table below sums up what to check and how.
| Item | How to check it | Risk if you skip it |
|---|---|---|
| Contact's name and job title | The company's official page, not an aggregator | Wrong person, an embarrassing mistake |
| Figures and dates | Official, dated source | Stating outdated information |
| News items cited | Open the article and read the date | Talking about a topic that has moved on |
| Features of a tool you mention | The provider's official documentation | Promising an option that does not exist |
| Legal or GDPR obligations | Official text or service-public.fr | Giving wrong legal advice |
One rule of caution applies too. If the AI suggests a promise of results, a deadline or a quantified gain, delete it unless you can back it with your own references. The prospect will judge you on your precision, not on your optimism.
Frequently asked questions
Can I ask the AI to search the internet for the prospect? Some tools do this directly, but the results must be checked. What it finds may be old, belong to a namesake or be misattributed. The most reliable approach is to read the main sources yourself and only give the tool text you have verified.
Should I tell the prospect that I used AI to prepare the meeting? It is not required, but what matters is that everything you say is accurate and that you can justify it. A prospect is less wary of a tool than of a claim you cannot support.
How long does this preparation take? With the material already gathered, the AI part takes a few minutes. The real time goes into checking and into thinking through your own answers. Serious preparation remains human work, even when it is sped up.
Can I paste notes from a previous exchange into the tool? Only if you know where they go and they contain no unnecessary personal data. Before pasting anything, read what you should never hand to a public AI.
Stuck on a specific case? Write to me via the contact page.
