← Back to Blog

How to Decide What AI Product to Build

The Questions to Answer Before Any Code

By Amin Rabinia · Founder, Glissando AI

From Idea to Application: Building an AI product that benefits your business, a talk by Amin Rabinia, with the barbershop menu app built during it

You know your market. You see what your customers struggle with, and you see what the tools on offer get wrong. For years, turning that into a product of your own needed a funded team and a year of work. AI has changed that. A first real version of a product now takes weeks, and building one is within reach of an established business, not only a startup with investors.

That shift moves the expensive decision. Building is no longer the hard part. Deciding what to build is. Anyone can now build the wrong thing quickly, and pay for it twice.

I gave a talk on this at Startup Folsom's AI for Business series in Roseville this November, and the part people asked about most afterwards was the sequence below. These are the questions I work through with founders before a line of code is written. You don't need a technical spec to answer them. You need answers only you have.

Amin Rabinia walking founders and business owners through the 6 Ps of a product brief at the Roseville talk
Walking the room through the 6 Ps of a product brief, Roseville, November 2026.

1. Build what's missing; buy what already works

Start by sorting the idea into one of two piles.

Buy what already works. Phone answering, email, scheduling, lead sorting: these are solved problems with good, cheap tools that everyone uses. Building your own version is money spent on something you could have had next week.

Build what's missing. The tool doesn't exist, it isn't made for the way you work, or the price doesn't fit what it does. You can see what it should be, and nobody sells it. That gap is where a product of your own starts.

The test I use is one question: if a competitor bought the same tool tomorrow, would you lose anything? If not, buy it. If yes, it's worth building, because what you build becomes something they can't buy.

One more filter before moving on. AI fits a process that is clear and repeatable. It does not fix a process that is broken or lives only in someone's head; it just runs the mess faster. If the process isn't settled yet, settle it first.


2. Is it needed? Look for signals, strongest first

Good product ideas are rooted in a market you know: something you see over and over, from customers who keep asking. The question is whether that need is real enough that people will pay. These are the signals I look for, in order of strength:

  1. People already paying for a worse alternative. The strongest signal there is.
  2. People doing it the hard way: spreadsheets, manual work, hiring someone to do it by hand.
  3. The same request, unprompted, from several people.
  4. Search demand and active communities around the problem.
  5. Competitors with real customers. Competition is evidence of a market, not a reason to stop.

Compliments, likes and "I'd use that" are opinions, not signals. The best proof is money. If someone will pay for it, or already pays you to do it by hand, the need is real. I go deeper on testing this in the Core, Form, Market framework.


3. Say it in one sentence, without the word "and"

Write down what the product does and for whom, in one sentence. Then look for the word "and".

"A practice tool that lets sales students rehearse real conversations with feedback from their trainer's own method" is one product. "A practice tool and a course platform and a CRM" is three products wearing one name, and three products cost three times as much to build and are harder to sell than one.

If you can't write the sentence yet, that is the first thing to work on, and it costs nothing. Every later decision, from the screens to the price, gets easier once it exists.


4. Separate the function from the features

A car's function is getting you from A to B. Heated seats are a feature. Nobody buys a car without the first, and plenty of people buy one without the second.

Your product works the same way. There is one function the customer is paying for, and a long list of features that would be nice. The first version delivers the function and nothing else. Features wait until real users ask for them. This is the single biggest lever on cost and timeline: most budgets that run over do so on features nobody needed yet. I wrote about this in more depth in function versus features in an AI product.


5. Write down how your business actually works

AI can build a system. It can't build a vague description. "Something that helps my customers choose" produces a generic app that could belong to anyone. A clear picture of how the work is actually done produces a product that fits.

So before the build, I capture the system with the founder:

  • The steps the work goes through, in order
  • The choices made along the way, and what they depend on
  • The rules you already follow, often without noticing
  • The exceptions you know by heart and nobody else does

Most business owners carry all of this in their head. Getting it onto paper is the most valuable hour of the whole project, and it is also your asset. That knowledge is what makes the product yours and hard to copy. A design studio I built for couldn't find a tool that designs the way they do, so we built their platform around their own design process. The process was the product. More on that in why your domain knowledge matters more than the model.


6. Turn the answers into a brief: the 6 Ps

Everything above goes into one short document before anything is built. I use six headings:

  1. Profile: the business and its market
  2. People: who uses the product, and when
  3. Purpose: the one job it must do
  4. Priorities: everything else, ranked as must, should, could, and won't
  5. Process: the steps, choices, rules and exceptions from section 5
  6. Payoff: what a happy user gets out of it

Here is the 6 Ps part of the talk, in under three minutes:

You bring the answers only you have: your market, your customers, how the work is done. I bring the design: which of those answers the product depends on, what goes in the first version, and what waits. The brief is where the two meet, and it is short enough to read in five minutes.

In the talk I showed what difference this makes with a simple example: a barbershop that wants customers to pick a cut, a fade and a beard trim and see the time and price before they book. Asked for "a barbershop pricing app", an AI builder produced invented services, made-up prices and no real logic for combinations. Given the shop's actual service list, the time and price of each option, and the rules about which combinations are allowed, the same AI built this shop's menu with correct totals. Same tool, same afternoon. The difference was the thinking that went in.

Fade Street Barbers menu app built from the brief: choose a main service, add extras, and see the live total price and time
The version built from the brief: the shop's real services, prices and times, with a live total.

The six questions to answer before building an AI product, and what a good answer looks like for each

Are you ready to bring in a builder?

You are ready when you have:

  • the one sentence, with no "and"
  • the function separated from the features
  • your system written down: steps, choices, rules, examples
  • signals from your market, ideally someone willing to pay
  • a date that matters: a launch, a client, a season, an investor meeting

You don't need all five finished. You need each one started, so the first conversation is about your business and not about guesses. A good builder will work the rest through with you, and will ask you questions before quoting a price. If someone quotes without asking, they are pricing a generic version of your idea. These are the questions worth asking them on the first call.

For reference, I build the first real version of a product for $2,000 to $10,000, with the scope fixed before work starts. Here is how that budget breaks down. Whoever you hire, make sure you own the code, the data and the accounts at the end.

If you want to see the core of your idea working before you commit to a build, that is what my free working demo is for.

This post is part of the Building with AI Guide, covering everything from the first idea to a product people use.


Related reading

Want to See Your AI Product Working Before You Commit?

Tell me the idea. I'll build a working sample of the hardest part on your real data in about a week, so you can judge the build on something that runs.

Got Questions?

Send Us a Message

We'll reply within one business day.

+1 916 936 1544
Sacramento, CA
What do you want built?
What are you building it for?
Budget for this project?
Waking up the assistant…