Pitching an AI Project to Leadership: Method and Template
To pitch an AI project to your leadership, start from a quantified problem, propose a 30 to 60 day pilot with a capped budget and written stop criteria, and let the decision to scale depend on evidence. A two-page brief is enough, as long as it answers the questions leadership actually asks: how much, for what gain, with which risks, and what happens if it does not work.
This guide is for HR, training and operations managers, and for startup founders who need a green light. It complements the complete guide to AI in business for SMEs, which sets out the overall strategy and adoption framework.
Why AI projects get refused or left hanging
In most cases, AI is not what gets refused. Vagueness is. A leadership team that answers "let's revisit this later" is not rejecting the technology. It is noting that it does not know what it is buying, what the total cost is, and who is accountable if something goes wrong.
Three patterns keep showing up in pitches that fail:
- The tool-led pitch. The presenter demonstrates ChatGPT, Claude or Copilot and their features, without saying which specific company problem they solve.
- The enthusiasm-led pitch. Productivity gains are announced with no baseline, so nobody can verify them afterwards.
- The pitch with no exit. Nothing says when to stop. The executive concludes they are committing without a safety net.
If your latest draft looks like this list, the fix is mechanical and takes a few lines per block.
Start from the problem, not the tool
A good pitch opens with a sentence describing a current cost, not a technology. For example: "Our admin team re-enters the same order data into three tools every week, which represents about 24 hours of work." Leadership can confirm or challenge that sentence in two minutes, which is exactly the point.
To build that sentence, measure before you pitch:
- Pick a single repetitive process with a volume you can count (invoices, emails, case files, meeting notes).
- Record the time actually spent, per person, over one to two weeks, without extrapolating.
- Note what happens when something goes wrong: late invoicing, customer complaints, manual fixes.
This baseline is what will make the pilot result credible. Without it you can only tell a story, not demonstrate a gain. It is the same principle as AI ROI for SMEs: a return can only be calculated against a measured starting point. To choose the right process, a short AI audit, even an internal one, helps rank the candidates (see also our article on the AI audit method for SMEs).
The 7 blocks of a pitch brief
A two-page, seven-block brief covers what leadership expects before signing. The table below sums up each block and the question it answers.
| Block | Expected content | Leadership's question |
|---|---|---|
| 1. Problem | One quantified sentence with a measured baseline | Why now? |
| 2. Scope | One process, one team, one duration (30 to 60 days) | How far does this go? |
| 3. Solution | The approach in plain language, tool named last | What are we buying? |
| 4. Total cost | Licences, internal time, training, outside support if any | How much, all in? |
| 5. Expected gain | Cautious range, measurement method, indicators | How will we know it works? |
| 6. Risks | Data, compliance, dependency, adoption, each with a response | What could go wrong? |
| 7. Decision | Continue and stop criteria, a named owner | What do we decide, and when? |
Two remarks on this table. Block 3 deliberately comes after problem and scope: if the tool is presented first, the discussion drifts into software comparison. And block 7 is the one that makes the difference: a pilot with a written stop criterion is a capped expense, whereas a project without one is an open-ended commitment.
Costing without overselling: a detailed example
A credible estimate shows the full cost and a range of gain, never a single optimistic figure. The example below is fictional and only serves as a calculation model: replace each value with your own data [To complete with your real figures].
Situation: a team of 6 people each spends 4 hours a week re-entering and checking data between a spreadsheet and a management system. The average loaded hourly cost is 30 EUR, over 45 working weeks.
- Current cost of the process: 6 people × 4 h × 45 weeks × 30 EUR = 32,400 EUR per year.
- Cautious gain assumption: a 30 to 50% reduction in time spent, or roughly 9,700 to 16,200 EUR per year. Do not present 100%: part of the checking work stays human.
- Pilot cost (60 days): licences for 6 people, one training day, about 3 days of internal time to configure and follow up. Add them line by line, including your own teams' time, which is often forgotten.
The result then reads simply: "The pilot costs X, aims for an annual gain between 9,700 and 16,200 EUR, and we stop it if time saved is still below 20% after 60 days." You can now answer "what if it doesn't work?" before anyone asks.
For more on cost items, see how much AI costs an SME and how to size your AI budget. For tracking indicators, see the AI KPIs to measure in 2026 and the definition of a KPI.
The objections to expect
Every objection from leadership is legitimate and deserves a written answer in the pitch, not an improvised one in the meeting. Here are the most frequent ones and how to handle them.
- "What about our data?" State which data goes into the tool, which is excluded, and whether a professional plan with contractual commitments is planned. The GDPR applies as soon as personal data is involved. The risk of unmanaged usage is covered in our article on shadow AI.
- "Is it compliant?" Mention Article 4 of the EU AI Act (Regulation (EU) 2024/1689), which requires organisations deploying AI systems to ensure a sufficient level of AI literacy among their staff. A project that includes training partly addresses this requirement [To verify before sending: application timeline and enforcement arrangements in force at the date of the pitch].
- "Who owns it?" Name an owner. Depending on company size, this may be a part-time AI lead or the head of the department concerned. Lightweight governance avoids adding a layer of complexity.
- "Will our teams use it?" This is the most underestimated risk. Plan training time and a weekly check-in. The field notes on AI change management detail the obstacles seen in practice.
- "Why not do it in-house?" The answer depends on your real skills. The article on agency versus internal team offers selection criteria.
Presenting on the day
The most effective format is ten minutes of presentation and twenty minutes of discussion, with the brief sent the day before. Open with the problem sentence, go straight to scope and total cost, then spend the remaining time on risks and decision criteria.
A few simple practices:
- State the decision you are asking for in the first minutes: a 60-day pilot, not "an AI project".
- Bring a short, concrete example (a five-minute demo on your own documents beats a product presentation).
- Acknowledge what you do not know yet. Saying "this figure will be confirmed by the pilot" is stronger than a promise.
- Leave with a fixed review date, not a "let's talk again".
If the pitch is accepted, the logical next step is a short roadmap: see the AI strategy for SMEs over 90 days. If the pilot succeeds, avoid the classic deployment mistakes when scaling up. To place the company before proposing anything, assessing AI maturity gives an objective starting point (detailed method here).
Frequently asked questions
How long does it take to prepare an AI project pitch?
Allow two to three weeks if you really measure the baseline: one to two weeks of time tracking, then a few days to cost and write the brief. Skipping the measurement saves time at first but weakens everything that follows.
Should I present a quantified ROI from the start?
Yes, but as a range with explicit assumptions, not a single figure. Leadership knows the first number is an estimate; what it expects is the measurement method and the threshold below which you stop.
What budget should I request for a first pilot?
It depends on scope, and there is no universal amount. The principle is to propose a capped envelope covering licences, training and internal time. A pilot limited to one process and one team usually stays modest compared with the cost of the current process [To verify with your real quotes].
How do I convince a leadership team that is sceptical about AI?
Do not try to convince them of AI in general. Offer to test a precise hypothesis on a small scope with a stop criterion. Scepticism then becomes an asset: leadership sets the success threshold itself.
What if the pilot does not deliver the expected results?
Apply the planned stop criterion, document what you learned (wrong process chosen, poor data quality, insufficient training) and report it. A pilot stopped cleanly builds your credibility for the next project.
Can an employee without a management role make the pitch?
Yes, provided there is an internal sponsor, often the head of the department concerned, who carries the budget request. The brief does not change; what changes is who presents the decision.
How GrowthPerf helps
GrowthPerf is a Qualiopi-certified training organisation specialising in AI, no-code and automation for SMEs and nonprofits. We help prepare this kind of file in two ways: a short audit to identify the process to pitch and measure its baseline, then training for the teams involved so the pilot is actually used. See the AI training for business page for our programmes.
If you are preparing a pitch and want a brief adapted to your context, request the GrowthPerf pitch template: we review it with you on a real case.