How to get your first startup customers: from conversation to a scoped offer
A first-customer learning cycle
Choose a reachable buyer, learn how they handle a recent problem, then propose an honest next step you can deliver. Record a payment separately from polite interest.
For
Australian early-stage founders preparing their first customer conversations—not a contractor recruitment guide.
Take away
Six supplied fictional answers, a dated follow-up decision and a matching editable twelve-field record.
No guarantee
A month is a planning window, not a promise of customers. Refusals and reasons not to build are useful results.

You have an idea or prototype and need to learn whether someone will buy it. Start with a specific person’s recent problem, not a campaign promising that AI will transform everything. This guide helps an early-stage founder conduct and document the conversation, then offer an appropriate demonstration or paid pilot.
For channel selection, spending limits and funnel arithmetic, use the GTM channel-test guide. Here the task is narrower: what to say, what evidence to record and how to avoid reporting more traction than actually happened.
1. Identify a person who experiences the problem
NYU Entrepreneurship’s initial-customer guidance recommends testing a focused segment and considering problem intensity, access and decision-making ability. It is practical guidance, not evidence of demand for your startup.
Fictional example used throughout: a Melbourne founder is exploring a review-only quote-follow-up assistant for small commercial maintenance firms. The proposed buyer is the owner who approves follow-ups. The founder suspects manual drafting is a problem but has not established its frequency, cost or priority. An office administrator may use the workflow without being able to buy software.
Write a fit rule before recruiting: “handles or approves quote follow-up and can describe a recent instance.” Ask how purchases are evaluated instead of assuming a job title means spending authority. Include people who prefer their existing process; they can reveal why your offer is unnecessary.
2. Ask for a conversation without disguising a sale
Use an appropriate, permitted route, such as an introduction the recipient has agreed to or an invitation made under a group’s rules. The example below is for a person who has agreed to hear from you; it does not establish permission for unsolicited email or other marketing. Check the requirements for your actual channel and jurisdiction before contacting people.
Thanks for agreeing to an introduction. I am exploring a tool for drafting quote follow-ups and trying to understand how people handle this today. Would you be open to a 15-minute conversation about a recent example? There is no need to share customer details or buy anything. If relevant, I can separately outline a synthetic demonstration. No problem if this is not a priority.
Replace the description with what you are actually building. Do not imply a mutual connection, customer result or finished product you do not have. Do not bulk-message event attendees, scrape contact lists or describe community membership as consent.
Agree any follow-up during the conversation: “Would you like the demonstration outline on Thursday, or should we leave it here?” Record the answer. A refusal ends that proposed follow-up; silence is not agreement to an escalating sequence.
3. Ask about the last occurrence before showing your solution
- Tell me about the last quote you followed up. What happened from quote to reply?
- Who did each step, and what did you use?
- Where did it take extra effort or go wrong? What happened as a result?
- What have you tried changing, and why did you keep or stop it?
- How would a change to this process be approved? Who else would need to assess it?
- Is there a next step you would find useful, or is the current process good enough?
Avoid leading questions such as “Would saving hours with AI be valuable?” They introduce an unproven benefit and invite agreement. If the participant estimates time or lost revenue, label it as their estimate; do not turn it into a measured saving. Ask to see a redacted or synthetic walkthrough only where appropriate, without requesting unnecessary confidential records.
Explain note-taking and intended use; obtain appropriate permission before recording or using a transcription service. Keep the minimum information needed in approved systems. Review AI summaries against the actual notes and retain uncertainty rather than having the tool invent a coherent success story.
Read the answers before deciding what they mean
COMPLETED FICTIONAL EXAMPLE — NOT CUSTOMER RESEARCH. All dialogue, dates and statuses are invented teaching material. No interview, message, demonstration or pilot actually ran. The future dates describe a story, not a forecast. SYNTHETIC-01 is a separate example, not one of the companion GTM guide’s 28 records.
Inspect all six scripted interview answers
These are supplied fictional answers and editorial interpretations, not a transcript or an anonymised real interview. N1–N6 are source labels within this exercise.
- N1: Tell me about the last quote you followed up. What happened from quote to reply?
Scripted answer: Last week I drafted six follow-ups from our quote spreadsheet. I think drafting took about an hour altogether. I did not keep a time log.
Interpretation: Participant-reported count and rough total time, not six separately timed tasks. No spreadsheet was inspected.
Still unknown: Frequency in other weeks, verified time, customer replies and financial consequences.
- N2: Who did each step, and what did you use?
Scripted answer: I prepare the drafts in email from the spreadsheet. The owner checks the wording before I send them.
Interpretation: The administrator describes the workflow and a separate approver. This is not verified authority to buy a new tool.
Still unknown: The actual approval policy, access permissions and implementation requirements.
- N3: Where did it take extra effort or go wrong? What happened as a result?
Scripted answer: The awkward part is finding whether the owner has already agreed something by phone. Sometimes I wait for that answer. I cannot tell you what it costs us.
Interpretation: Missing context may be the problem; a faster draft would not resolve it. No lost-revenue figure is supplied.
Still unknown: How often missing context occurs, time spent waiting and whether a different process would solve it.
- N4: What have you tried changing, and why did you keep or stop it?
Scripted answer: We made an email template. It covers most straightforward follow-ups. I would probably tidy the spreadsheet before adding another tool.
Interpretation: Contrary evidence: the existing template may be adequate. Include the simpler process option.
Still unknown: Whether the remaining exceptions justify a new product or a paid pilot.
- N5: How would a change to this process be approved? Who else would need to assess it?
Scripted answer: The owner decides on software. We have not discussed a budget for this. I am not arranging an introduction to them.
Interpretation: A user conversation is not buyer approval. No permission to approach the owner is given.
Still unknown: Owner interest, budget, procurement conditions and willingness to pay.
- N6: Is there a next step you would find useful, or is the current process good enough?
Scripted answer: You can send one outline on Thursday, but I am not booking a demo. Private written notes are fine for this exercise until 30 September; no recording, AI upload or public quote.
Interpretation: One dated outline is permitted in the story, not a sequence, booking or publication release. The later reply can end the proposed next step.
Still unknown: Whether the participant will want a demonstration after reading the outline.
N3 changes the problem: missing information about a phone agreement will not be fixed merely by drafting faster. N4 supplies contrary evidence—the existing template may already be enough. A useful interview can lead away from building AI.
4. Keep an evidence record, not a persuasion score
Copy these fields, or save the editable record and fictional completed example. The file includes the same twelve blank and completed fields, all six answers, the five-event sequence, stage checks and four-week plan. This is a working aid, not a validated interview instrument.
Internal record ID and conversation date: Role, segment fit and permission to follow up: Recent problem described by the participant: Current workaround and consequence: Evidence observed versus participant report: Buying process and unresolved approval: Offer actually shown, price and exclusions: Response / demo / trial / accepted terms / payment / delivery: Objection, uncertainty or reason to decline: Next action, owner and agreed date: Retention and public quotation permissions: What would change our decision:
Follow one permission change through the record
This fictional case ends before a demonstration. Read the actual outline and refusal in E3–E4, rather than inferring what they said from a status label.
E1 — : Recipient-agreed introduction and invitation for one conversation.
Source material: Supplied invitation above; the contact route is an assumption of the story, not permission for a real recipient.
State: Conversation accepted for 16 September. No demo booked or paid offer shown.
E2 — : Fictional conversation produces notes N1–N6.
Source material: The six supplied answers; private paraphrases only. No customer document, timing log or recording.
State: One outline agreed for 17 September. Buying authority unresolved; no demo booking.
E3 — : Founder sends the single agreed outline.
Source material: As agreed, here is the outline. A proposed walkthrough would compare two fictional quote records with draft follow-ups, checking wording before any send. There is no live integration and nothing will be sent to customers. You do not need to send records or book anything. This is an outline, not a completed test or paid proposal; I will not send reminders without a new agreement.
State: Outline sent, not a demonstration attended. No price or paid terms shown.
E4 — : Participant declines further contact.
Source material: Thanks, please leave it here. Our template is enough for now. Please do not chase this or contact the owner.
State: No demo, trial or purchase. No further outreach or owner introduction agreed.
E5 — : Founder reviews the record without contacting the participant.
Source material: N3, N4, N5 and E4: missing context, a useful existing template, unresolved buyer approval and an explicit refusal.
State: Stop contact with this participant. The later refusal ends the proposed demonstration path; the earlier permission for one outline cannot justify reminders or an owner approach. Keep the recorded decline, do not invoice, and do not claim a customer win.
Completed twelve-field record after the refusal
- Internal record ID and conversation date
- SYNTHETIC-01; conversation 16 September 2026, record updated 21 September after a refusal on 18 September. These are future fictional dates, not bookings or observations. This record is separate from the companion GTM dataset.
- Role, segment fit and permission to follow up
- Office administrator at a fictional Melbourne commercial-maintenance firm. Fits the workflow-user question; not established as a buyer. One outline was agreed and sent. E4 ends further contact; no owner approach is permitted by this story.
- Recent problem described by the participant
- N1 reports six manually drafted follow-ups last week, roughly one hour total with no time log. N3 describes missing context about promises made by phone. Neither frequency nor loss is verified.
- Current workaround and consequence
- Spreadsheet, email template, manual drafting and owner review. N4 says the template covers most cases and spreadsheet tidying may be preferable. The consequence of waiting for missing context is not quantified.
- Evidence observed versus participant report
- In the story, the founder records the conversation, sent outline and refusal. Workflow counts, effort and approval descriptions are participant reports only. No real document, product test or independent timing is supplied. Outside the story, every item is synthetic.
- Buying process and unresolved approval
- N5 reports that the owner decides on software. Budget and owner interest remain unknown. The owner was not approached; the administrator did not approve a purchase or offer an introduction.
- Offer actually shown, price and exclusions
- Only the E3 unpriced demonstration outline was sent: two fictional records, draft review, no live integration or customer sends. The optional AUD 600 ten-record review-only pilot is a possible future proposal, not a price shown, rejected or accepted here.
- Response / demo / trial / accepted terms / payment / delivery
- Conversation held and outline sent; later refusal recorded. No demo booked or attended; no trial, accepted paid terms, payment or completed delivery. One participant is not a market or channel-conversion sample.
- Objection, uncertainty or reason to decline
- E4 explicitly says the current template is enough for now and asks for no further contact. Do not infer a price objection: no price was shown. Buyer demand, urgency beyond this participant, actual cost and delivery feasibility remain unknown.
- Next action, owner and agreed date
- Founder completes the internal review on 21 September; no outgoing message or owner approach. The four-week learning plan does not override the 18 September refusal. Any other discovery requires independently appropriate access and its own record.
- Retention and public quotation permissions
- N6 permits private written notes until 30 September in this fictional exercise; no recording, AI upload or permission to publish names, quotes, logos or results. Review/remove working notes then. Appropriate retention and minimal do-not-contact handling must be approved for a real workflow; this invented date is not legal guidance.
- What would change our decision
- Stop this contact now. The broader product hypothesis could change with independent evidence of unresolved exceptions, an authorised interested buyer and a deliverable scope; it is not rescued by inventing savings or revisiting a refusal. Prefer a simpler template/process if the evidence supports it.
What can this record support?
Wide tables scroll sideways; focus a table region and use arrow keys, or read the same rows in the text download.
| Stage | Recorded state | Evidence boundary |
|---|---|---|
| Conversation response and completion | One conversation accepted and held in the story. | E1–E2; agreement to speak is not agreement to buy. |
| Outline and later reply | One requested outline sent; further contact declined. | E3–E4; the negative reply is a response, not a conversion. |
| Demonstration | No booking and no attendance. | N6 and E4; an outline is not a demo or use of a product. |
| Trial and use | No trial, usage or completed pilot review. | Nothing was provided for independent use; no use result exists. |
| Price and accepted terms | No priced offer shown; no accepted paid terms. | The possible AUD 600 pilot below is a founder planning assumption, not an offer made in this conversation. |
| Payment and delivery | No payment or completed delivery. | No order, invoice, paid pilot or delivered workflow is supplied. |
Decision: Stop contact with this participant. The later refusal ends the proposed demonstration path; the earlier permission for one outline cannot justify reminders or an owner approach. Keep the recorded decline, do not invoice, and do not claim a customer win. The refusal is not a failed pricing experiment: no price was shown. It is also not proof that no other buyer has the problem.
5. Make a small offer only when the next step is appropriate
Stone & Chalk’s first-customer guide describes early acquisition as learning whether a specific group needs what you are building. Its advice does not guarantee a sale or justify making a benefit claim before testing it.
For a different, continuing case—not a new message to this participant—the founder could prepare a fixed AUD 600 total pilot proposal for one review-only workflow, ten synthetic records and one review session. Exclude live integrations, sending messages, ongoing support and sales guarantees. This invented price matches the companion GTM example; it is not an MLAI offer, market benchmark or proof of profitability. It was not shown in SYNTHETIC-01, so willingness to pay remains untested.
Before anyone commits, agree the actual deliverable, price and payment conditions, scope exclusions, review criteria, responsibilities, completion date and what happens if the test fails. Clarify access, data handling and handover. If you cannot safely deliver the requested workflow, narrow it or decline rather than selling an undefined “agent”. Use the scoped pilot guide for the delivery decision.
Keep the stages separate: a reply is a response; attending a demonstration is engagement; using a free trial is usage; accepting paid terms is an agreement; receiving payment is a payment; completing the promised work is delivery. Continued use and repeat purchasing are further evidence, not implied by any earlier stage.
6. Use a four-week plan without forcing a sale
The weeks organise work, not promise a customer by day 30. Choose your own time/cash limits and review date before starting. A refusal or lack of safe delivery capacity stops the relevant next step immediately; spare time or a calendar deadline cannot override it.
| Window | Do | Keep as evidence | Decision |
|---|---|---|---|
| Week 1 | Define the user/buyer question, establish an appropriate route, invite and hold a manageable first conversation. | Fit rule, invitation, supplied questions, actual answers and permission boundaries. | In the fictional case, 14–18 September includes a conversation, outline and refusal. Stop contact when refused; do not wait for week four. |
| Week 2 | Review the notes and any expressly agreed follow-up; distinguish user reports from buyer evidence. | What was sent, replies or missing replies, approvals and contradictory evidence. | On 21 September the fictional founder reviews internally. The template may be adequate; no further contact is made. |
| Week 3 | Only for a continuing, appropriate case: demonstrate synthetic material and separately discuss bounded terms. | Actual booking/attendance, price shown, who can approve and feasible scope. | Not run in this example. No permission, no useful unresolved problem or unsafe delivery means no demo or offer. |
| Week 4 | Review the cycle and decide whether another, independently appropriate test is justified. | Limits and actual costs; accepted terms, payments, use and delivery separately, if any. | Planned, not run here. Do not fabricate a sale to finish the schedule or restart declined follow-up. |
No replies may indicate an unsuitable access route rather than no market demand. No urgency in relevant conversations may indicate a weak problem. Accepted terms followed by non-payment require clarification, not a fabricated conversion. Write alternative explanations before choosing the next experiment.
The download has separate actual-case slots for the current permission or refusal, dated evidence, owner, next allowed action, costs/limits and retention review. Keep private notes out of public URLs, unapproved AI tools and analytics. Do not copy fictional answers into a real evidence log.
Discuss the question you still cannot answer
Bring one anonymised offer and one unresolved customer question to an appropriate MLAI event. Read its topic, format and experience level first. Participate to learn and contribute; attendance is not a promise of leads, buyers or an individual sales review.
Carry your preference to MLAI events
Only your event-format preference travels to the calendar. Your interview notes and worksheet are not uploaded or attached. Check the actual listing; this does not book a place or arrange a customer interview or sales review.
Applied to the event CTA: All formats.
Select All formats and apply to clear it. Your selection is kept in this page's URL, not saved as a private shortlist.
Bring the question your conversations changed
Find a relevant MLAI founder or AI event. Bring one anonymised or synthetic question about your buyer, offer or decision to stop. Check the topic, experience level and online or in-person format. Your notes are not attached; attendance does not include leads, buyers or an individual sales review.
Explore founder learning eventsAfter actual delivery, request permission for any public quote or case study and retain the context, baseline, measurement period and limits. Do not turn an illustrative pilot into “our customers saved” copy. If you need to organise actual company evidence for an update or content plan, Founder Tools is a supporting option—not a replacement for gathering evidence. Opening a workspace requires an account; your record is not automatically imported.
Sources checked 11 September 2026. The invitation, questions, worksheet and fictional scenario are editorial aids, not original interviews or endorsements by the source organisations. AI-assisted editorial example; independent founder/source review pending. The stable article URL is retained; the advice does not depend on an unsupported claim that acquisition changed in 2026.
Frequently Asked Questions
How many conversations are enough?
Is a free trial a customer?
Should I offer a discount?
Can I publish a prospect’s feedback as proof?
Disclaimer: This article provides general information and is not legal or technical advice. For official guidelines on the safe and responsible use of AI, please refer to the Australian Government’s Guidance for AI Adoption →
Bring the question your conversations changed
Find a relevant MLAI founder or AI event. Bring one anonymised or synthetic question about your buyer, offer or decision to stop. Check the topic, experience level and online or in-person format. Your notes are not attached; attendance does not include leads, buyers or an individual sales review.
Explore founder learning events