“Can you turn this into a presentation?”
You have interview notes, customer comments, and a meeting tomorrow. You paste everything into an AI tool and request six slides.
The first draft looks ready. Then someone asks: “What exactly are you asking us to approve?”
That question belongs in the brief you give the tool.
This third guide in the Product Builder’s AI Arsenal explores Communicate and present. We’ll turn customer feedback into a small proposal, adapt it for different colleagues, and prepare for the questions it raises.
The central idea: Decide what your audience needs to understand and act on before asking AI to shape the presentation.
1. Start with the decision you need
In Part 2, we followed a hypothetical meal-subscription business investigating cancellations. One customer couldn’t pause while travelling. Another had successfully skipped a week.
We’ll continue that fictional example. These are teaching inputs, not findings from a real company.
SourceCustomer’s wordsWhat the comment establishesC01“I liked the meals, but I couldn’t pause my subscription while travelling.”This customer reported difficulty pausing.C04“I skipped a week through the app. That worked fine.”This customer completed a related task.C05“After the introductory offer ended, it cost more than I could keep spending.”This customer described an affordability concern.
Imagine the team now needs time from customer support and design to investigate the pause experience.
A request to “improve retention” leaves too much open. It could mean changing prices, menus, delivery arrangements, or subscription controls.
A more useful request is:
Agree to a small investigation of the pause and skip experience, with named support and design owners and a review date.
That sentence defines the purpose of the discussion. It also sets a boundary: the team is considering an investigation. There is no evidence here to promise a reduction in cancellations.
Before generating content, write down:
Audience: The people who must decide, contribute, or carry out the next step.
Decision: The specific agreement you need from them.
Evidence: What supports the request, including conflicting accounts.
Unknowns: What you still need to establish.
Constraint: The available time, capacity, or other relevant limit.
Use an approved workspace for real customer material, and remove unnecessary personal details.
2. Give AI an argument it can help you examine
An AI assistant can help organise the brief and identify gaps. Start with the notes and a clear instruction:
Using only the source notes below, draft a short proposal for a support lead, designer, engineer, and business owner. The decision requested is whether to investigate the pause and skip experience. Separate customer evidence, interpretation, unknowns, and proposed action. Keep source IDs beside evidence. Identify missing information. Do not invent customer counts, costs, deadlines, or expected results.
The output should preserve a distinction like this:
Evidence: C01 reported difficulty pausing. C04 successfully skipped a week.
Interpretation: Customers may encounter different journeys, rules, or circumstances.
Unknown: We haven’t established whether pausing and skipping use the same process.
Proposed action: Observe affected customers attempting the task and check the relevant rules.
Watch for wording that makes the evidence sound stronger than it is.
“Customers struggle to manage subscriptions” turns one reported experience into a broader claim. “A confusing interface causes cancellations” adds an unverified cause.
A defensible heading would be: “One customer reported difficulty pausing; the obstacle is still unknown.”
Keep C05’s price concern visible as a separate issue. Even if the pause experience improves, affordability could still influence cancellations.
Your judgment determines which claims the presentation can carry. Read each headline as if a colleague might repeat it without the supporting notes.
3. Adapt the explanation to the audience
Different colleagues need different details to assess the same request.
AudienceQuestion they need answeredWhat to emphasiseCustomer supportWhat do we need to learn from affected customers?Recruitment criteria, neutral follow-up questions, and where to record findings.Design and engineeringWhat task failed, and under what conditions?The attempted journey, device, timing, rules, and any available error evidence.Business ownerWhy allocate capacity to this investigation?The uncertainty it will resolve, proposed scope, staff effort, and review point.
Ask AI to adapt the emphasis while preserving the facts:
Rewrite this proposal as three brief openings: one for customer support, one for design and engineering, and one for the business owner. Keep the evidence and uncertainty identical. Change the emphasis to address each audience’s questions. Flag any missing information needed for each version.
For the business owner, the opening might be:
We propose investigating why one customer couldn’t pause while another could skip a week. This would help us distinguish an interface, technical, or policy issue before committing to a change. We need to agree on ownership, effort, and a review date.
For support, it could begin:
We need help finding customers who encountered difficulty pausing and inviting them to show us what happened. Please preserve their wording and capture what they tried, when they tried it, and what happened next.
Both versions describe the same situation. Neither claims the problem is widespread.
This is where perspective-taking matters. A support lead may be concerned about adding work to a busy queue. An engineer may need enough detail to reproduce an error. Address those concerns directly.
4. Turn the proposal into four useful slides
Once the argument is clear, choose the format. A short email may be sufficient for a straightforward agreement. Slides help when the group needs to inspect evidence together.
For our example, a four-slide outline could look like this:
Slide 1: Identify the obstacle before choosing a change
Main message: We need to understand what prevented the customer from pausing.
Show the decision requested: agree on a small investigation, its owners, and a review date.
Slide 2: Two customers describe different experiences
Place C01 and C04’s exact quotes beside each other, with their source IDs.
Add the unresolved question: Do pausing and skipping follow the same process and rules?
Include a short note that affordability is a separate concern in C05. This prevents the discussion from treating pause controls as an explanation for every cancellation.
Slide 3: Observe the task and check the rules
Show the proposed investigation:
Invite affected customers to demonstrate what they attempted.
Record the steps, timing, device, and any error.
Check those observations against the relevant subscription rules.
Call this a proposed plan. Do not draw a detailed customer journey where the intermediate steps are still unknown.
Slide 4: Agree on ownership, effort, and the review point
State the required contributions: support helps recruit; design observes; engineering examines technical evidence if relevant.
Leave unconfirmed effort and dates visibly open for agreement. At the review, the team should assess the observed obstacle and decide what to test next—or what remains unresolved.
The presentation should make the request easy to assess, including the work it creates.
5. Use presentation tools after checking the outline
Gamma supports generating presentations from prompts, outlines, or existing content. It also offers editing and export options.
Canva offers AI-assisted presentation drafting and tools for editing the design and applying branding.
For this exercise, give either tool the checked outline. These are suggested workflows, not a head-to-head product test.
Create a four-slide presentation using the approved outline below. Preserve customer quotes and source IDs exactly. Keep unknowns visible. Use one main message per slide, readable text, and simple visuals. Do not add statistics, projected benefits, customer photographs, or invented interface screens. Keep detail that does not fit in speaker notes, where supported.
Inspect what comes back. A generated chart can imply measurement where you only have comments. A polished screen mockup can look like an existing feature. A shorter headline can remove a necessary qualification.
Choose visuals that help people inspect the argument:
Comparing experiences: Two quote cards with sources.
Explaining a known process: A simple journey with verified steps.
Requesting an agreement: A clearly labelled decision box.
Check the exported version too. Text must remain readable, and evidence labels must stay beside the claims they support.
6. Rehearse the questions the proposal raises
Use AI to prepare for scrutiny:
Review this proposal from the perspectives of a support lead, engineer, and business owner. Identify the strongest unanswered question from each. Explain why it matters. Suggest an answer only where the source notes support one; otherwise name the evidence we need. Check for claims stronger than the evidence.
Three likely questions are:
“How common is this?” We cannot tell from these comments. We need a broader review to estimate frequency.
“Why not move the pause button now?” We haven’t established that finding the button was the obstacle.
“What will this do to cancellations?” The investigation can clarify a task failure. Any retention effect would require separate assessment.
AI-generated objections are rehearsal material. Colleagues may have operational knowledge the tool lacks. Invite them to correct the proposal and change it when their evidence warrants it.
After the discussion, record what was agreed, who owns the next action, and when the team will review it. Check that record with the people responsible.
Try it with one proposal this week
Choose a real piece of work: a research request, partnership proposal, executive update, or change to a customer journey.
Write the decision you need in one sentence. Give AI the supporting material and ask it to separate evidence from interpretation. Then build the smallest useful communication around that request.
Before sharing, ask a colleague to read it and tell you:
What are we being asked to decide?
What evidence supports that request?
What is still unknown?
If their answers differ from yours, revise the explanation.
At Product Pulse Africa, we’re exploring how builders use AI in the work they actually need to complete. For this part of the arsenal, the practical output is a proposal your audience can question and act on.
Try this with one upcoming meeting, then share which question made your proposal clearer.
Further reading and references
The Product Builder’s AI Arsenal — series overview: The five categories and the product-building jobs they support.
Think and research — Part 1: The first deep dive in the series.
Capture and understand customers — Part 2: The preceding guide introduces the fictional customer feedback used here.
Gamma — official product overview: Presentation generation, editing, and sharing capabilities.
Canva — AI presentation maker: AI-assisted presentation creation and design.
The meal-subscription scenario, proposal, and slide outline are illustrative. Tool descriptions were checked in September 2026; features and availability may change.