Why the description matters more than the technology
Every way of getting software built starts from the same place: a description of what it should do. Where the description is clear, the builder builds what you meant. Where it is vague, someone fills the gap with a guess, and a guess is the most common reason a finished app is not the app you had in mind.
You do not need technical words for any of this. A good description is about people and what they do, not about databases or frameworks. Choosing the technology is the builder's job.
The five things to write down
- Who it is for. The people who will use it, and if there is more than one kind, each kind: customers and staff, for example, or players and an organiser.
- The problem it solves for them. What they do today without it, and what is slow, costly or annoying about that.
- What it must do on day one. Written as things a person can do: “a guest can book a table”, “I can see today's bookings”. A handful is enough for a first version.
- What it is not. The things you are deliberately leaving out. This line saves more time than any other, because it stops the scope growing quietly.
- How you will know it works. What you would check on the first day, or what you would measure after a month.
- Who it is for
- Guests booking a table, and the restaurant staff who take the bookings.
- The problem
- Bookings arrive by phone while the kitchen is busy, and some are lost.
- Day one
- A guest books a table for a date, a time and a party size. Staff see today’s bookings.
- Not yet
- Deposits, a waiting list, and more than one restaurant.
- It works when
- Most bookings arrive online within a month, and none are lost.
Vague and clear, side by side
The same idea, described two ways. The examples are illustrative.
| Vague | Clear |
|---|---|
| An app for my restaurant. | A booking page for my restaurant. Guests choose a date, a time and a table size up to 8, and get an email confirming it. I see the day's bookings on one screen. |
| Users can manage their account. | A guest can cancel a booking from the link in the confirmation email. There are no guest accounts. |
| It should be easy to use. | Most guests book from a phone, so the booking takes one screen on a phone. |
The rules and edge cases people forget
Most surprises in a finished app come from situations the description never mentioned. For each thing the app does, ask what should happen when:
- it is full, sold out or closed;
- someone changes their mind, cancels or asks for a refund;
- someone types something wrong, or leaves a field empty;
- two people try to do the same thing at the same moment;
- you, the owner, need to correct something by hand.
You do not need every answer up front. You do need to know these are questions, so they get asked.
Decisions that change the size of the build
A few early decisions shape the whole app. Say what you know about each:
- Sign-in. Does anyone need an account, or can people use it without one?
- Data. What does the app keep, and who can see it?
- Payments. Does it take money, and for what?
- Other services. Does it need to talk to anything you already use, such as a calendar or an email list?
- Phone or computer. Where will people mostly use it? A web app can work well on both, and can be installed on a phone from the browser.
Each of these adds work. The guide on what an app costs to build explains how.
Keep the first version small
The best first version does one job well for one kind of user. It is quicker to build, cheaper, and it tells you sooner whether people want it. Put everything else on a list for later: an app is easier to improve once real people are using it than to perfect before anyone has.
What happens to your description with 19 Agents
You type the idea in your own words, and Sofia, the AI project coordinator, asks the questions needed to settle the scope, the requirements and the important decisions. The agents then write a specification of what will be built, and what will not, which you can read and correct.
- YouDescribe the ideain your own words
- SofiaQuestionsuntil the scope is clear
- SpecWritten scopewhat is in, and what is not
- BudgetOne fixed pricefor that scope
- Your decisionAccept, or change itthen the build starts
From that scope, the agents plan the work and prepare one fixed price for it. You see the scope and the price before anything is built, and nothing starts until you accept.
The beta page explains how to join. During the beta every product starts from a new idea, which is exactly what this guide is for.
Questions
- How detailed should my app idea be before I start?
- Detailed enough to say who it is for, the problem it solves, the few things it must do on day one, and what it is not. You do not need technical detail. With 19 Agents, the agents ask about anything the description leaves open before the scope is written.
- Do I need to write a formal specification or requirements document?
- No. A plain description in your own words is enough to start. With 19 Agents, the agents write the specification from your description and your answers, and you read and correct it before anything is built.
- What is the most common mistake when describing an app?
- Leaving out what the app is not. Without that, the scope grows quietly, and the first version takes longer and costs more than it needs to.
- Can I change my mind after the scope is approved?
- Yes. With 19 Agents a change of scope gets its own quote, and nothing changes until you approve it.