AI tools for founders
Most founders need four AI tools, not forty: one that writes and edits code, one that writes and edits words, one general-purpose assistant for research and analysis, and one that decides what to work on next. Everything else is a feature of one of those four. The trap is collecting tools in the first category, where the results are visible and satisfying, while leaving the fourth empty, which is where side projects actually die.
Sort tools by the job, not by the category page
Every AI tool list is organised by product type, which is useless when you are choosing. Organise by the job instead, because you only need one tool per job and the second one you buy in the same category is money you have set on fire.
There are four jobs a founder actually has, and they are worth naming in the order they get neglected.
- Build it. Cursor, Claude Code, Lovable, Bolt, v0, Replit. Pick one and get fast at it. Switching costs more than the difference between any two of them.
- Say it. Claude or ChatGPT for landing copy, emails, and posts. The differentiator is not the model, it is whether you have given it real detail about your product and your customer.
- Understand it. A general assistant for research, competitor teardowns, pricing analysis, reading your own numbers. Connect it to your real data if you can; a model reasoning about your actual Stripe figures is a different tool from one guessing.
- Decide what is next. This is the empty slot for most people. See below.
The fourth job is the one nobody buys a tool for
Building has never been easier and it has never been less of a bottleneck. A founder with AI coding tools can ship a working product in a weekend. The thing that has not got easier is knowing which weekend to spend on which product, and what to do on the Monday after you launch it.
That is the gap. You can ask a general assistant what to do next, and it will give you a competent, generic answer, because it has no memory of what you already tried, no view of your numbers, and no expectation that you will do any of it. Ask the same question next week and you get a different competent, generic answer.
What closes the gap is a tool that holds state: it knows what you sell, what you charge, what you shipped and when, and what you said you would do last week. That turns "what should I do next" from a question you ask into a thing that is already waiting for you.
What is worth paying for
A rough test: pay for a tool if it saves you an hour a week and you would notice within a fortnight if it disappeared. Cancel anything you have to remember to open.
For most side projects the honest bill is one coding tool, one assistant subscription, and one thing that runs the plan. Under a hundred a month, total. If you are paying for more than that before you have revenue, the tools have become the project.
- Worth it early: a coding tool you are fast in, a general assistant, and something that keeps the plan and the accountability.
- Worth it later: analytics, email, a CRM, once you have enough customers for those to have anything in them.
- Rarely worth it: a second tool in a category you already have covered, and anything sold on the promise that it replaces having a plan.
A note on AI that does the work rather than describing it
The useful shift in 2026 is not better answers, it is tools that take the action. An assistant that tells you how to write a cold email is a search engine with better manners. One that drafts the twenty emails, against your actual customer list, in your voice, and puts them in front of you to send, has done the job.
When you are comparing tools, ask what it produces rather than what it knows. A tool that ends its turn with a paragraph of advice has handed the work back to you.
Frequently asked questions
What AI tools do I need to start a business?
One AI coding tool if you are building software, one general assistant for writing and research, and one thing that decides what to work on next and keeps you to it. That covers the jobs. Add analytics, email and a CRM once you have customers for them to be about.
Can AI replace a business coach or a co-founder?
Not as a substitute for judgement or for someone with their own stake in the outcome. It replaces a large part of what a coach actually does day to day, which is helping you work out the next right thing and expecting you to have done it. The part it cannot replace is caring whether you did.
Is ChatGPT enough on its own?
For research, drafting and analysis, largely yes. What it does not do is remember your business between conversations, watch your numbers, or notice when you have gone quiet for two weeks. Those are the jobs that decide whether a side project reaches revenue, and they need something that holds state.
How much should a founder spend on AI tools?
Before revenue, under a hundred a month across everything. A coding tool, an assistant, and something that runs the plan. If the bill is bigger than that pre-revenue, the tools have quietly become the project.