TL;DR: Before you hire AI app developers, get eight things clear: the problem and user, the core workflow, where your data lives, apps you like, must-haves, who decides, your deadline, and how you'll know it worked. You don't need a full spec or a tech stack.
Why does prep matter before you hire AI app developers?
A fuzzy first call gets you a fuzzy quote. The developer guesses what you mean. You guess what they heard. Then you both build on the guess.
RAND interviewed 65 experienced data scientists and engineers about why AI projects fail. The top root cause was people misunderstanding or miscommunicating the problem the AI needs to solve. RAND also cites estimates that more than 80% of AI projects fail. That is twice the rate of IT projects without AI. It is an estimate RAND quotes, not something it measured. Read it in RAND's report on why AI projects fail.
Google's DORA research on AI-assisted development found AI works as an amplifier. It magnifies the strengths and weaknesses a team already has. A brief works the same way. A clear one gets sharper. A fuzzy one gets fuzzier, faster.
So do the prep. Set aside an afternoon. It helps if you want a small tool, and it helps more if you want full AI app development with several user roles. Write each answer below in one or two plain sentences.
What is the one problem and the one user?
Start here. Everything else hangs off it.
Write one sentence for the problem. Write one sentence for the person who has it. Not "our customers". One type of person, in one situation.
Here is a template:
- Who: a [type of person] who [situation].
- Problem: they can't [do the thing] without [pain].
- Today: they cope by [spreadsheet, email, call, guesswork].
The third line matters most. If you cannot say how people cope today, you may not understand the problem yet. Talk to three real users before you talk to developers.
If you have two users, such as a coach and a client, pick the one who feels the pain first. The other comes later.
How do you write your core workflow as steps?
The core workflow is the one path that makes the app worth using. Write it as numbered steps, from the user's first action to the result they get.
Keep it to 5 to 10 steps. For example:
- The user signs up and answers three questions.
- The app reads their answers.
- The AI drafts a plan from your method.
- The user edits and saves the plan.
- You get a note that a new plan is ready.
Mark each step: does a person do it, or does the AI? Mark where a human should check the AI's output before the user sees it. A good developer will want to talk about that spot.
Don't draw screens yet. A list of steps is enough to quote from.
Where do your content and data live?
AI apps run on your material. Your course, your notes, your client records, your templates. List where each piece sits and who can get to it.
A simple table does the job:
| What | Where it lives | Who owns it | Has personal info? |
| Course videos | Course platform | You | No |
| Client notes | Google Docs | You | Yes |
| Past reports | Shared drive | You and clients | Yes |
The last column matters. If the app will hold personal information, say so on the first call. Might any of it go to an overseas provider? The OAIC says an Australian business must take reasonable steps to make sure the recipient follows the privacy principles. It can be held accountable for what the recipient does with the data. That is APP 8 on cross-border disclosure. It is general information, not legal advice.
For a plain primer, read AI and your customer data: security basics.
Which apps do you like, and why?
Bring two or three apps you like and one you don't. They don't need to be in your industry.
For each one, write what you like in a line. "The sign-up takes 30 seconds." "It shows my progress." "It never asks me for things twice." Examples beat adjectives. "Clean and modern" means something different to everyone.
Also name what you don't want copied. If an app has 40 features and you want 3, say so. This helps a developer see what size of thing you have in mind.
Send screenshots before the call if you can. A screenshot is quicker than a description.
What are your must-haves, and what can wait?
Split your ideas into two lists. Must-have for version one. Later.
A must-have passes this test: without it, the core workflow breaks. Everything else goes in Later. Login is a must-have. A leaderboard or a referral system is usually Later.
Be strict. Every item on the first list adds time to the build. A short first list gets something real in front of users sooner. What they do with it tells you what to build next.
If this list is hard to write, you are not alone. It is what a good scoping session is for. See scoping an MVP for an AI product for how that cut works. Cost depends on scope, and scoping comes first.
Who decides?
Name the person who says yes. Then name everyone else who gets a say.
Projects can stall here, and the code has nothing to do with it. Three people each want changes. The developer waits. Nothing moves.
Write down:
- The decider: one name. They approve scope and sign off on changes.
- The input group: people who give feedback but don't decide.
- The user tester: one real user who will try each version.
If you are a solo founder, you are the decider. Say it out loud anyway. It tells the developer how fast they can move.
What is your deadline, and why?
Give a date and the reason behind it. "Launch before our September intake" is useful. "As soon as possible" is not.
The reason tells a developer what can bend. A launch tied to a live event can't move. A soft target can. That changes what they cut, and how.
If you don't have a real date, say that. A made-up date can force a rush you don't need.
For a feel of how long builds run, read how long it takes to build an AI app.
How will you know it worked?
Pick one number that tells you the app did its job. Write it before you build.
Examples: how many clients finish the first plan. How many hours you save each week. How many users come back in week two. Pick the one that ties to your real goal.
Gartner made a forecast in 2024. At least 30% of generative AI projects, it said, would be abandoned after proof of concept by the end of 2025. It blamed poor data quality, weak risk controls, rising costs and unclear business value (Gartner press release). It is a forecast, not a count. But "unclear business value" is the one you can fix before you start.
A clear success number also keeps the team honest. You can both see whether it worked.
What do you not need to bring?
You might be putting off the call because you think you need more. You don't.
| You need | You do not need |
| One problem and one user | A full spec |
| The core workflow as steps | A tech stack |
| Where your data lives | A pitch deck |
| Apps you like | Wireframes or designs |
| Must-haves and a later list | A final name or logo |
A good developer picks the tech. They should explain the choice in plain words. If someone pushes a stack on you before they understand your problem, that is a warning sign. See red flags when hiring an AI agency.
When you are ready to compare teams, how to choose an AI development company covers the questions to ask.
What if you can't answer these yet?
Then you have found the real first job. That is useful to know before you pay for a build.
There are two routes. You can fill the gaps yourself and come back. Or you can work through them with a developer in a structured session. What happens in an AI software discovery phase explains how that works.
Devwiz has built 200+ apps since 2015, and the Software Plan Workshop is built for this gap. It is $5,000, credited to a build. You get a written software plan, a clickable prototype and the first-version scope, ready to quote.
Take your answers to the first call. Even half-written notes make that call useful.
What is your next step?
Write the eight answers today, one or two lines each. Circle where you get stuck. If three or more leave you stuck, book the Software Plan Workshop with Devwiz and we'll work through them together.
Frequently asked questions
What should I prepare before talking to AI app developers?
Prepare eight short answers: the one problem and the one user, the core workflow as steps, where your content and data live, apps you like, your must-haves and later list, who decides, your deadline and why, and how you will know it worked. One or two lines each is enough.
Do I need a full spec or a tech stack before I hire a developer?
No. A full spec is the developer's job to help you shape. A tech stack is their call, and they should explain it in plain words. What they need from you is the problem, the user, the workflow and your goals.
What if I only have an idea and no detail?
That is fine, but say so. Start with the one problem and the one user, then sketch the workflow as steps. If you get stuck, a discovery session or the Software Plan Workshop is the right first step before a build.
How do I explain what I want if I am not technical?
Use examples. Show two or three apps you like and say what you like about each. Write the workflow as numbered steps in plain words. Developers can work from that without any technical terms from you.
Do I need a pitch deck to get an accurate quote?
No. A pitch deck is for investors. For a quote, a developer needs the problem, the user, the core workflow, the must-have features and your deadline. Those five things matter more than any slide.
Why do developers ask who makes the decisions?
Because slow sign-off can hold a project up. If one person approves scope and changes, work moves fast. If several people each want changes, the developer ends up waiting.
What is the Software Plan Workshop?
It is a Devwiz package for people who cannot yet answer the questions above. It costs $5,000 and is credited to a build. You get a written software plan, a clickable prototype and the first-version scope, ready to quote.
About James Killick
10+ years building digital products · 200+ apps shipped since 2015
James is a co-founder of Devwiz and an AI product specialist. Since 2015 he has helped ship 200+ apps for founders, businesses and government, including work for NSW Government, Briometrix and Huskee. He builds AI-first platforms and writes about turning a proven program into software. He also hosts the Up in the AI podcast.
More articles by James · James's personal site · LinkedIn · AI Orchestrators
Tags: Hiring


