SaaS Idea Generator

Ask any model for SaaS ideas and you will get the same list everyone else gets. A CRM for freelancers. An AI tool for social media scheduling. A project management app for remote teams. The ideas are not wrong, they are just already built, by companies with funding and a five year head start.

The problem is not the model. The problem is the prompt. "Give me SaaS ideas" is a request with no constraints, so it returns the statistical average of every SaaS article ever written. A saas idea generator is only useful if it produces ideas that are specific enough to act on.

This guide gives you prompts that add the constraints back, a rubric to throw most of the output away, and a way to test whether anyone wants the survivor before you build it.

Why generic prompts give generic ideas

A language model answers with what is most likely. Ask a broad question and you get the most common answer, which by definition is the crowded market.

Constraints change the arithmetic. When you say "for independent dance studio owners in cities under 200,000 people who currently run scheduling through group chats," the model can no longer fall back on the average. It has to reason about a specific situation, and specific situations are where unbuilt software lives.

So the job is not to find a better generator. The job is to feed it better inputs. Three inputs matter most:

• A person. Not "small businesses." A role you can picture and, ideally, reach.
• A trigger. The moment the pain shows up. End of month, a new hire, a client handoff.
• A current workaround. What they do today. Usually a spreadsheet, a group chat, or unpaid manual work.

Every prompt below is a different way of supplying those three.

Prompt 1: Start from constraints, not categories

This is the workhorse. Fill in the brackets and run it.

You are helping me find an unbuilt SaaS product.

Constraints:
- Buyer: [specific role, e.g. independent bookkeepers with 5 to 30 clients]
- Trigger: [the moment the pain appears, e.g. month-end close]
- Current workaround: [what they do today, e.g. a shared spreadsheet plus email]
- My unfair advantage: [what you know or can access that most people cannot]

Give me 10 product ideas that fit these constraints.
For each one, write:
1. The job it does, in one sentence a buyer would say out loud
2. What the buyer does today instead
3. Why an incumbent has not bothered to build it
4. The smallest version that would still be worth paying for

Skip anything that requires more than one integration to be useful.

That last line matters more than it looks. Ideas that need five integrations to deliver value are not v1 products, they are two-year roadmaps.

Prompt 2: Mine your own friction log

The best ideas usually come from work you already do. Most people forget their own annoyances within a day, so write them down for a week first: every time you do something manual, repetitive, or clearly stupid, add a line.

Then run this:

Here is a list of tasks I did manually or dreaded this week:
[paste your list]

For each item, tell me:
- Is this a personal quirk or a pattern others in my role share?
- If it is a pattern, roughly how many people share it and how would I reach them?
- What would the smallest tool that removes this task look like?

Rank the results by how easy the audience is to reach, not by market size.

Reachability beats market size for a first product. A market of two million people you cannot find is worth less than a market of three thousand who all read the same newsletter.

Prompt 3: Read the complaints about tools that already exist

Complaints are pre-validated demand. Somebody already pays for a product and is still unhappy, which means budget exists and the problem is not solved.

Take the product category [category, e.g. invoicing tools for agencies].

List the 10 most common complaints users have about the leading products,
based on what people typically say in reviews and forums.

For each complaint, tell me:
- Whether it is a missing feature, a pricing problem, or a wrong-audience problem
- Whether a smaller focused product could fix it without rebuilding the whole category
- Who exactly would switch, and what would make switching worth the hassle

Ignore complaints that only enterprise customers have.

Verify what the model gives you. It is generating plausible complaints, not quoting real ones. Use its output as a list of things to go check, then read actual reviews and forum threads yourself.

Prompt 4: Find the workflows still running on spreadsheets

A spreadsheet that several people edit is a product waiting to happen. It means the workflow matters enough to maintain by hand and nobody has packaged it.

For the industry [industry], list workflows that are commonly still run in
spreadsheets or shared documents rather than dedicated software.

For each one:
- Who maintains the spreadsheet and who else depends on it
- What breaks when it gets big or when the maintainer leaves
- Why no vendor has replaced it yet
- What a focused tool would need to do on day one to be worth switching to

Prefer workflows where the spreadsheet is shared, not personal.

Shared is the key word. A personal spreadsheet is a habit. A shared one is a process, and processes have budgets.

Prompt 5: Follow a change that just happened

New rules, new platforms, and new costs create problems nobody has tooling for yet. This is the closest thing to a fair race, because incumbents start from zero too.

In the last 18 months, what has changed for [role or industry]?
Consider new regulations, platform or API changes, cost shifts, and new tools
that became standard.

For each change, tell me:
- What new work it created, and for whom
- How people are handling that work today
- What a small tool built specifically for this change would do

Focus on changes that created recurring work, not one-time migrations.

Recurring is what makes it a subscription. One-time work makes a service, not a SaaS.

Filter hard: the five-question rubric

A generator that returns thirty ideas has not helped you until you have discarded twenty nine. Score each idea from 0 to 2 on these five questions.

Can you reach the buyer? Do you know where these people gather, or someone who does? Score 0 if your plan is "run ads and hope."

Do they already pay for something adjacent? Existing spend proves budget exists. Convincing someone to spend for the first time in a category is a much harder sell than convincing them to switch.

Does the pain repeat? Monthly beats yearly. Weekly beats monthly. Recurring pain is what makes people tolerate a recurring charge.

Can a useful v1 be small? If you cannot describe a first version you could build in a few weeks, the idea is not wrong but it is not your first product.

Do you know something about this domain? Domain knowledge is the one advantage that does not require money or a team.

Anything scoring below 7 out of 10 goes in a note file, not on your roadmap. Ideas that score 9 or 10 are rare and usually come from your own friction log, which is why prompt 2 is worth the week it costs.

Validate before you build, not after

The cheapest mistake to fix is the one you catch before writing code. Three steps, in order, none of which require a product.

Talk to five people who have the problem. Not about your idea, about their week. Ask what they did last time the problem came up, what it cost them, and what they tried already. If they cannot remember the last time it happened, the pain is not recurring enough.

Ask for a commitment that costs something. Not "would you use this," which everyone answers yes to. Ask for a pre-order, a pilot slot, an intro to their boss, or a scheduled call to review a prototype. Interest that survives a small cost is real.

Put up one page and send traffic to it. A single page describing the outcome, with an email field and a price on it. If people who match your buyer profile read it and do not sign up, the message is wrong or the problem is not painful. Both are useful to learn in a week rather than a quarter.

Only after those three does building make sense.

Validated the idea? Do not start from an empty repo. LaunchKit gives your AI coding assistant a complete SaaS to start from, with sign-in, billing, and a dashboard already working on day one.

From validated idea to shipped product

Once an idea survives the rubric and the validation, the bottleneck moves. It is no longer "what should I build," it is the two or three weeks of setup work that stands between you and a product: authentication, billing, a dashboard shell, email, a landing page that does not look like a template.

This is the part where most validated ideas quietly die. The idea was good, the founder spent three weekends wiring up auth and Stripe, and the momentum was gone before the first customer saw anything.

Two things help here.

Start from a codebase that already has the boring parts wired in rather than an empty folder. Our Website Starter Kit ships with authentication, payments, AI chat, email, and analytics already connected, so the first thing you build is the thing that makes your product different.

And give your AI coding tool a design system before you ask it to build screens, otherwise every SaaS you generate will also look generated. A DESIGN.md file carries the visual decisions so your product looks intentional from the first screen.

Frequently asked questions

Is an AI SaaS idea generator actually useful?

Yes, but not as an oracle. It is useful as a way to explore a space quickly once you have supplied real constraints. It is not useful for deciding what to build, because it has no idea what you can reach, what you know, or what you will still care about in six months.

How many ideas should I generate before choosing?

Enough that discarding one does not hurt. Thirty to fifty is a reasonable pool from a few prompt runs. The point of volume is not to find a gem, it is to make you willing to reject ideas you have grown attached to.

Are micro SaaS ideas better for a first product?

Usually, for one reason: a smaller scope means you reach a real customer sooner. A narrow tool for a specific role can be built and sold by one person. A broad platform generally cannot.

Should I worry that someone already built it?

Existing competitors are a good sign, not a bad one. They prove people pay. Worry when a market is empty, because empty usually means nobody wants it, not that nobody thought of it.

Can I trust what the model says about market size?

No. Treat every number it produces as a guess that needs checking. Models are good at generating hypotheses and bad at knowing facts about markets. Use the output as a research list, not as research.

What if none of my ideas score above 7?

That usually means you are generating for an audience you cannot reach. Change the buyer, not the idea. Run the prompts again with a group you actually have access to, even if that group seems small or unglamorous.

Start with the constraints you already have

The useful version of a saas idea generator is not a button that returns ideas. It is a loop: supply real constraints, generate widely, filter hard, then test the survivor on five real people before writing a line of code.

The constraints are the part only you can provide. Everything else is mechanical, and your AI tool is good at mechanical.

When an idea survives all of that, the next step is not another prompt. It is getting a working product in front of the people who told you the problem was real, quickly enough that they still remember the conversation.