Blog › AI
Building after ChatGPT: what I would focus on as a founder
My priorities as a founder: understand a customer problem, prove a useful result, and build a business that can keep delivering it.
As a founder, I want to build something customers find useful enough to keep using and paying for. That starts with understanding their work, solving a problem that matters and learning what it takes to deliver the result reliably.
Research on Y Combinator’s startup mix gives me places to explore. Its public directory shows more activity in industrial markets and the systems that other software needs to run. These are observations from a dated company list, not proof of customer demand or business success. Read the findings and their limits.
My next step would be to speak with customers in those markets. What is difficult today? What have they already tried? What improvement would make a new product worth adopting?
The priorities below are my practical takeaways as a founder. They are suggestions to test with customers, not outcomes proved by the research.
Learn from other founders without turning their journey into your deadline
A successful fundraising round deserves recognition. It can give a team the resources to develop technology, reach customers and take on difficult problems.
But a public announcement gives us only part of the picture. It rarely explains every customer conversation, failed experiment, change of direction or year of preparation behind the result.
A useful response is curiosity. Which customer problem did the team understand? How did it reach buyers? What made customers trust it? Which parts of that approach could apply to our own market?
These questions help us learn from another company without assuming we need the same valuation or timetable.
Start with one customer problem you can prove you solve
Choose a clear customer and a problem that matters enough for someone to spend time or money fixing it.
For example, “we help distributors check supplier invoices” gives a team a clearer starting point than “we use AI to improve business.” It identifies a user, a task and a result to test.
Then agree with the customer on what improvement would look like. It might be less checking time, fewer missed errors or faster approval. Measure the current process before testing the product, and include the time people spend correcting its output.
Early pilots are useful learning opportunities. Be clear about what each one proves: interest, successful use, willingness to pay, or repeat demand. Those are different steps, and each can guide the next decision.
Make the first customers easier to serve
Winning a customer is the start of learning how the business works in practice.
Watch where setup slows down. Notice which questions keep returning. Find out whether people use the product after the initial trial and what brings them back.
For an AI product, include model usage, hosting, support and human review when estimating the cost of serving a customer. A product can save time in one part of a workflow while creating extra work elsewhere.
These observations help decide what to improve next. Sometimes the most valuable work is a clearer setup process, a more reliable integration or a faster response to a customer problem.
Give growth a clear purpose
Different businesses need different amounts of capital. A hardware team, a research company and a small software team may have very different spending needs. Raising funding and growing from customer revenue can each suit particular circumstances.
When considering new spending, connect it to a concrete milestone. What should this hire, product release or sales effort help the company demonstrate? How will we know whether it worked?
Keep a clear view of available cash and expected spending. Leave room for delays and learning. A plan is more useful when the team can adjust it as new evidence arrives.
A practical plan for the next 30 days
Use this as a starting point and adapt it to your stage:
- Speak with five customers or likely buyers. Ask how they handle the problem today, what it costs them and who decides whether to buy.
- Choose one result to improve. Agree on a measure such as time saved, fewer errors or a higher completion rate.
- Run a small, agreed test. Compare it with the current process. Record what worked and where people needed to step in.
- Review the full cost. Include setup, software, support and manual work. Identify what must improve before serving more customers.
- Choose the next milestone. It could be a paid pilot, a renewal, a shorter setup time or a reliable deployment. Make it specific enough to guide the team's work.
A research-heavy company may need a technical milestone before a customer trial. An established product may need stronger renewals before a wider sales push. The next step should fit the business.
A caution about valuations
The valuation figures come from a September collection, not a fresh November survey. Some records are missing or uncertain, and collection methods differ. Some amounts may describe money raised rather than the value of the company. A small number of very large reported values can also move the average substantially. Repeating the calculations does not independently verify every record. I would not use this comparison to set a target for my own business.
A valuation is also different from sales, profit or money returned to investors. This research cannot tell us which company will succeed, or give any founder a target they should expect to reach.
Other founders' progress can be a source of ideas and encouragement. Our own progress becomes clearer when we can show who we help, what improves for them and how we can keep delivering it. That is useful evidence for the next customer, the next hire and the next investor conversation.