How to find your first customers after vibe coding

Conceptual illustration, not a customer photograph.

Start looking for customers by finding a recurring problem and talking to people who already face it. Offer a small, useful result that you can reliably deliver. Faster AI-assisted building gives you room for more tests, but those tests still need to establish who will pay and how you can reach them.

Start with a situation, not a broad audience

Give your first outreach message a specific situation to address. In a hypothetical small agency, someone reconciles contractor invoices every Friday and spends time chasing missing details. You now have a workflow to investigate, a likely contact and a result to demonstrate. The category “small businesses” gives you none of that detail.

Map the task from its trigger to the consequences of delay. Identify who does the work and who authorises the spending; they may be different people. Enthusiasm means little as demand evidence if it comes from someone who does not face the problem or lacks a budget.

Approach a group you can reach openly. Say that you are building a product and researching their process instead of presenting yourself as a neutral observer. Respect community rules. You can learn from conversations without mass messaging people or concealing a promotional interest.

Ask about the last time the problem happened

Ask for a recent example of the problem. Have the person walk through what they did, where time went and which mistakes caused trouble. Find out what they have tried and what workaround they now use. Those details can expose constraints that a general opinion of your idea will miss.

Follow their account with questions about the work, rather than opening with a feature list. If someone offers screenshots or documents, obtain permission to use them. Remove private information before passing the material to a tool or collaborator.

In his essay on doing things that do not scale, Paul Graham advocates actively recruiting and serving the earliest users. That advice draws on startup experience; it does not promise that manual outreach will produce customers. Here, the proposed exercise applies it to a small task whose outcome you can observe.

Offer a pilot with a clear outcome

In the fictional agency case, offer to process a single invoice batch and return a reconciled list of exceptions. Specify the required inputs, delivered result, price and turnaround. Tell the customer which steps you will perform manually and which parts of the product remain unfinished.

Promise only what you can deliver. While resolving an inability to handle customer records safely, demonstrate the flow using synthetic data. Work through the production checklist before introducing real accounts or taking payments.

Payment for this pilot is evidence of willingness to buy this particular offer. Retention and market size remain open questions. State delivery and refund terms clearly, and keep refundable reservations distinct from recurring revenue when describing the result.

Pick one acquisition route and measure the steps

Pick a way to reach buyers that fits their circumstances. That could be an introduction, a useful tutorial, a relevant industry group or direct contact with organisations you can credibly help. A public launch may attract attention, but reactions do not count qualified buyers.

As an illustrative test, approach a small, deliberately chosen group. Track delivery of messages, replies and qualified conversations, then pilots begun, outcomes completed and repeat orders. This guide supplies no benchmark sample size or conversion threshold. Decide what result would justify continuing in light of your costs and capacity before you run the test.

Write down the reason behind each refusal. An infrequent task may call for different pricing; a user without purchasing authority may lead you to an approver. If people do not understand the promised result, make the demonstration clearer before trying to reach a larger audience.

Price the work you actually deliver

Cost the whole delivery, including manual work and support. A modest software subscription bill can coexist with an expensive service operation. Use the cost worksheet to separate subscriptions, usage and labour, then judge whether you can repeat the pilot as a viable offer.

If a discount compensates for unfinished features or substantial feedback, explain it and set an end date. Otherwise you may learn little about demand at the normal price. Ask which result would prompt another purchase, and follow up to see whether the customer actually buys again.

Discuss the completed pilot with the customer. Find out what worked, what they continued doing elsewhere and whether another use would still need your involvement. The demand-testing and pre-sales guide helps define the commitment. Explore funding options when a specific constraint, confirmed through that work, requires resources.

A worksheet you can reuse

OUTREACH
Hi [name], I am the founder of [product], which I am developing for [specific recurring task]. I would like to understand how your team dealt with it most recently. Could we have a short conversation? I am researching whether the product is worth building.

PILOT OFFER
Customer and task: [ ]
Input needed: [ ]
Delivered outcome: [ ]
Manual or unfinished steps: [ ]
Delivery date and responsibilities: [ ]
Price and payment timing: [ ]
Cancellation and refund terms: [ ]
Data handling and deletion: [ ]
Agreed success check: [ ]
Follow-up: repeat / revise / stop