How to gather context from colleagues before a client project starts

Most of what a delivery team needs to know about a new client project already sits inside the firm. The person who sold the work heard what the client is worried about. Someone who ran a similar project two years ago knows where it went wrong. An account lead knows which stakeholder really decides. Very little of this is written down, and the usual kickoff meeting gets a fraction of it, from whoever happened to be free that hour.

This guide sets out a way to collect that knowledge on purpose, in the week before a project starts, without booking a meeting with everyone. It works with a shared document and a few phone calls. It also works with a tool that interviews people for you; the method is the same either way.

Decide what you need to find out before you ask anyone

Write down the decisions the first two weeks of the project depend on. Not topics, decisions. "How should we staff the discovery phase" is a decision. "Client background" is a topic, and a topic invites people to tell you everything or nothing.

Three to five questions is enough. Good ones for a project start:

• What did the client say they want, and what do we think they need? Where do the two differ?
• What was promised during the sale that is not in the statement of work?
• Who on the client side decides, who can block, and who was missing from the conversations so far?
• What went wrong the last time we did work like this, for this client or another one?
• What would make this client consider the project a failure by the end of month one?

List who knows, including the people who are usually skipped

The obvious people are whoever sold the work and whoever will lead it. The ones who get left out tend to know the most specific things:

• Anyone who worked with this client before, even on an unrelated project or at a previous employer.
• Whoever delivered the most similar project, for any client.
• The people who wrote the proposal and the estimate, who know which numbers were guesses.
• Finance or operations, who know how the client pays and how they behave when scope changes.
• Junior team members and freelancers who were on the last engagement and saw the day-to-day.

The list is usually longer than expected, and that is the reason a single kickoff meeting does not work: an hour between twelve people is five minutes each, and the most senior voice uses most of it.

Ask each person separately, and ask about a real event

People give better answers alone than in a group. In a meeting they agree with the first plausible account. Alone, they tell you what they saw.

Ask about something that happened, not for an opinion. "Tell me about the last conversation you had with the client about timelines" gets a fact. "What do you think about the timeline" gets a position.

When an answer stays general, follow up once and make the question narrower. "You said the client is difficult about sign-off. What happened the last time something needed signing off?" If the second answer is still vague, move on. Pressing a third time produces guesses.

Do not suggest answers. "Was it a budget issue?" gets a yes. "What was the issue?" gets the issue.

Ten minutes per person is enough for three to five questions. Let people do it when it suits them. A written form is the cheapest way, but written answers are short and tidy, and the useful detail is usually in what someone says out loud while thinking. A short call, a voice note, or an AI interviewer that asks and follows up will all get more than a form.

Keep the words, then look for the pattern

Write down what each person said in their own words before you summarise. Summaries written from memory lose the detail and take on the views of the person summarising.

Then group across people. For each of your questions, look for three things:

• What several people said independently. This is what you can plan on.
• Where two people contradict each other. This is what to resolve before kickoff, usually with one short conversation between those two people.
• What only one person said but which would change the plan if true. Check it, do not average it away.

Keep each finding attached to who said it. A finding with no source cannot be checked, and three weeks later nobody will remember where it came from.

Turn it into one page the team reads before kickoff

The output is a page, not a deck. A structure that works:

• What we know: the points several people agree on, one line each.
• What we disagree on: each contradiction, and who is resolving it.
• What was promised: commitments made during the sale that sit outside the written scope.
• Risks from last time: what went wrong on similar work and what we are doing differently.
• Open questions for the client: what nobody inside the firm could answer.

Send it a day ahead. The kickoff meeting can then be spent on the disagreements and the open questions, which is the part that needs everyone in a room.

Common ways this goes wrong

• Asking too late. After kickoff, people defend the plan that exists. Ask in the week before.
• Asking only senior people. They know the commercial side. The people who did the work know why it took twice as long.
• Asking too much. Twenty questions gets skimmed answers to all of them. Five gets real answers.
• Letting the summary replace the source. Keep the original words where the team can read them.
• Doing it once. The same questions asked again at the midpoint show what the team has learned and what is still unknown.

Where AI Scouts fits

AI Scouts does the asking step of this method. You write what you need to find out and who from; a scout holds a short voice or text conversation with each person in their own time, follows up when an answer stays vague, and hands back what everyone said grouped into findings, each linked to the words behind it. There is a free plan. The method above does not depend on it: a document, a list of names and ten minutes with each person will get most of the way.