Before you fund an AI project, make someone name the problem

In 60 seconds

• Fund problems, not tools. Before budget, someone names the problem, who owns it and which decision changes. "We need a chatbot" is a tool, not a problem.
• Ask who is pushing. Pressure from a competitor's news or a board's fear of missing out is not a reason. Find out what the pusher expects it to solve.
• A new capability is not a project. Back problems your organisation has lived with long enough to describe in detail.
• Put the return next to the problem: which number moves, by how much, by when.

Why this comes first

When RAND asked data scientists and engineers why AI projects fail, the cause they named most was that the people commissioning the work "misunderstand—or miscommunicate—what problem needs to be solved using AI." 84% named at least one leadership-driven cause as the main reason: the wrong problem, the wrong metric, or priorities that changed too fast. If you approve AI budgets, that is your part of the job.

1. Settle the problem before anyone touches the data

"We have the data, so let's use it" is how many projects start. Start from the business problem instead. If there isn't a real one, stop, however good the data is. If there is, check that AI is the right fix before anything gets built.

Know who has the problem, too. Two experts can watch the same flow of research for different reasons: one has a regulatory question to answer, the other needs the right facts in front of a colleague before a decision. The tool that helps one won't help the other unless someone said which problem it is for.

2. Find out who is pushing

Many proposals arrive with pressure and no problem statement: fear of missing out, a belief that competitors are doing it, leaders who want "AI" without saying what that means.

A consultant who builds AI for clients heard the same thing on sales calls. Prospects would ring and say they needed generative AI. Asked what for, the answer was that "the executive team was breathing down their neck." If you are the one pushing, say what you expect it to solve. If someone else is, ask them.

3. A new capability is not a project

Builders feel the pull of what a model has just learned to do. The companies that scale AI start the other way round: from a problem they have lived with for years, whose data, systems and context they already understand. The ones stuck in pilots start with a powerful general tool and go looking for a use, and it usually ends in failure.

When a proposal starts from what a model can now do, ask which problem your organisation has had long enough to describe in detail.

4. Put the return in the same document as the problem

Treat AI as a magic box and you get research, not results. Tie every project to a goal the business has already defined. Otherwise you get work "that we can write a paper on, but can't run a business on," as one AI vendor's chief executive put it. Naming the problem is also how you build the business case: what you are solving, and what solving it is worth.

Three questions for whoever proposes the next project

1. What is the problem, whose is it, and what decision changes? A tool name doesn't answer this.
2. Does it need AI, and who is pushing for it? Find out whether a simpler fix would do, and what changed to make this urgent.
3. Which number moves, and by when? Get it in writing with the proposal.

Exercise (20 minutes): write the problem statement for one proposal

Take one AI proposal that is waiting for budget, or one already under way that nobody has written a problem down for. Fill in what you know; you won't know every line yourself, and the person who proposed it should answer most of them. Every line you can't fill gets an owner and a date before you close the card. The card download includes a filled-in example of a proposal that isn't ready to fund yet, and why.

Use AI to fill in and test your card

These three prompts go into your company's approved AI assistant. Use the first to fill in the card one row at a time; it won't guess an answer for you. Use the second when the card is done, to hear the case against funding the project. The third asks about your industry and your proposal and turns the lesson into a 30-day plan, with a checklist, rubric or brief for your boss if you want one. Use placeholders for company, system and people names; the help doesn't depend on them.

Prompt 1: Fill in my card with me

Paste this, then add your proposal in one or two sentences and the card below.

Prompt from The AI Decision Course by Dr. Tony Hoang, host of The Artificial Intelligence Podcast (Spotify, Apple Podcasts): tonyphoang.com/learn

Before you start: use your company's approved AI assistant, not a personal account. Replace company, system, supplier and people names with placeholders (Company A, System X) and round any figures. The help you get does not depend on them.

You are helping me fill in a problem statement card for one proposed AI project. The card is pasted at the end.

Rules:
- Ask me about one row at a time, in order, and wait for my answer before moving on.
- Write down only what I tell you. Never suggest, estimate or invent a number, a name, a yes/no answer or a decision for me, even if I ask you to guess. If I ask, say that this row needs a real answer and ask who could give it.
- If I describe the problem using a tool or product name, ask me to say it again without one, and write down what I say.
- If I don't know an answer, write "Not known" and ask who should find out and by when. Put that in the Owner / by when column.
- After the last row, show the whole card as a table with three columns: Question, Answer, Owner / by when. Then list every row marked "Not known" with its owner, and remind me that any unanswered row means the project is not ready to fund.

My proposal in one or two sentences: [DESCRIBE THE PROPOSAL]

The card:
[PASTE THE CARD]

Prompt 2: Stress-test my card

Paste this, then your filled card.

Prompt from The AI Decision Course by Dr. Tony Hoang, host of The Artificial Intelligence Podcast (Spotify, Apple Podcasts): tonyphoang.com/learn

Before you start: use your company's approved AI assistant, not a personal account. Replace company, system, supplier and people names with placeholders (Company A, System X) and round any figures. The review does not depend on them.

Below is a problem statement card I filled in for a proposed AI project. Act as a sceptical reviewer at the budget meeting and argue the case against funding it.

For each row:
- Say whether the answer is specific enough to act on (a named problem, a named role, a number, a date) or too vague.
- Point out where answers contradict each other, for example a problem described in one row and a number to move in another that has nothing to do with it, or a project said to start from the problem that only became urgent after a competitor's news.
- Say what evidence you would ask for before accepting it.

Do not rewrite my answers and do not invent facts about my project. If a row is blank or says "Not known", treat it as not ready.

Finish with: your recommendation (FUND, SEND BACK or DECLINE), the three rows that most need work, and one question I should ask my team on Monday.

My card:
[PASTE YOUR FILLED CARD]

To paste the card into prompt 1, copy its rows from the PDF.

Prompt 3: Make it yours

Paste this and answer its four questions.

Prompt from The AI Decision Course by Dr. Tony Hoang, host of The Artificial Intelligence Podcast (Spotify, Apple Podcasts): tonyphoang.com/learn

Before you start: use your company's approved AI assistant, not a personal account. Replace company, system, supplier and people names with placeholders (Company A, System X, Supplier B).

You are helping a leader apply one lesson to their own situation. The lesson's four decisions are below. Use only them and what I tell you. Never invent facts, numbers, vendors, laws or results about my organisation or my industry. Where a number is needed and I have not given one, say what to measure and who should measure it, instead of guessing.

First, ask me these four questions, one at a time, and wait for each answer:
1. My industry, and roughly how many people work in my organisation.
2. The AI project I have in mind, in one or two sentences.
3. Where it stands today: an idea, a proposal waiting for budget, or already under way.
4. What worries me most about it.

Then give me the following. Keep it short and practical:
A. My plan. For each of the four decisions: one action written for my project, the role that should own it (a role, not a name), and what should be done within 30 days. If my industry makes a decision harder, for example regulated work, customer-facing work or long-standing systems few people understand, say which decision and why in one sentence.
B. Three questions to ask my team on Monday, specific to my project.
C. Ask me which one I want: a checklist for my proposal, a rubric to score AI proposals against each other, or a one-page brief for my boss. Then write it, and finish with a prompt I can paste later to update it.
Mark anything that rests on an assumption with "(check)".

The four decisions:
1. Settle the problem before you start on the data. If the project isn't solving a real problem, stop, whatever data you already have. Know who has the problem, who solves it today and what they need it for, then check that AI is the right answer.
2. Find out who is pushing. When a project arrives with pressure behind it (fear of missing out, a competitor's move, leaders who want "AI") and no problem statement, find out who is pushing and what they expect it to solve, and check whether a simpler fix would do.
3. A new capability is not a project. Don't invest only because a model can now do something. Favour problems the organisation has lived with long enough to describe in detail, whose data, systems and context it already understands.
4. Put the return in the same document as the problem. Tie the project to a goal the business has defined: which number moves, by how much, by when, and what solving it is worth.

Your company's assistant may answer differently, so read what it writes before you use it.

Check yourself

Sources

The advice draws on interviews with an adviser on AI projects, three companies that sell AI, a startup founder, a consultant who builds AI for clients and an AI lead at a large consulting firm.

• RAND Corporation, "The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed: Avoiding the Anti-Patterns of AI" (RRA2680-1), 13 August 2024. (https://www.rand.org/pubs/research_reports/RRA2680-1.html) One thing to hold on the numbers above: RAND's findings come from 65 semi-structured interviews run August–December 2023 with 50 industry data scientists and ML engineers from more than 50 organisations, plus 15 academics. They report practitioners' perceptions of why projects fail, not a measured failure rate, and RAND notes that most interviewees were non-managerial engineers, so the results "may be skewed toward identifying leadership failures." Projects that only prompted a pre-trained large language model were out of scope.