How much does it cost to hire someone to finish my AI-built app?
Short answer: It depends on what is left, not on how much is already built. A small gap, such as one broken screen or a deploy, can be a few hundred to a couple of thousand dollars. A full list of gaps, such as login, payments, an admin screen and going live, usually lands in the thousands to low five figures. The ranges below are built from hours multiplied by hourly rates I state, so you can swap in the quote you actually get. Lovable, Bolt, Replit and Cursor are tools that write app code from plain-English instructions, and the code they leave behind is usually a solid start, not a finished product.
You have an app that is part built in an AI tool, you are stuck, and you want a real number before you talk to anyone.
I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.
The short answer: price the gaps, not the app
Developers do not price "an app that is 70% done." They price a list of jobs. Each job takes some hours, and the hours are multiplied by a rate. So the useful question is not "what does finishing cost?" It is "what is the list of jobs, and how many hours is each one?"
That is good news for you. A list can be written down, checked and priced as a fixed amount. The rest of this page gives you the jobs that usually remain in an AI-built app, an hour range for each, and a method to turn them into a fixed-price quote.
One note first. There is no single market average for this, and I will not make one up. The figures below are my planning assumptions, shown as hours and a stated hourly rate. Treat them as a way to sense-check a quote, not as a price list.
What usually needs finishing in an AI-built app
AI app builders are good at screens and basic flows. They are weaker at the parts that must be exactly right: who can see what, money, and running safely in the real world. If you want the longer reasons, read why AI-built apps break. In practice the leftover work falls into these groups.
- Login and accounts. Sign-up, password reset, email confirmation, and making sure one user cannot see another user's data.
- Payments. Stripe, a payment processor that handles cards for you, needs more than a pay button. Its documentation says it signs every webhook, which is the message Stripe sends your app when a payment succeeds, and your app should check that signature. It also says your endpoint should return a success response before you mark an invoice paid. These are the details that stay unfinished.
- Data and permissions. The database rules that decide who can read and change each row. The Supabase documentation, a database platform many AI builders use, says a table without Row Level Security (rules that limit access row by row) is readable and writable by any role with a grant on it.
- Admin screens and emails. A way for you to see orders or users, and emails the app should send.
- Going live. A real domain, a separate live version from the test version, backups, and error alerts.
- Polish. Phone layouts, empty states, slow pages, and the bugs you find when real people click around.
Cost by remaining work
The table shows hour ranges I would plan for each job, and the cost at an example rate of $100 an hour. Both the hours and the rate are assumptions for illustration. Developer rates vary a lot by country and experience, so replace $100 with the rate in your quote. Figures are as of November 2026 and in US dollars.
| Job | Planning hours | Cost at $100 an hour |
|---|---|---|
| Paid review: is it worth finishing? | 3 to 6 | $300 to $600 |
| Login, password reset, email confirmation | 8 to 16 | $800 to $1,600 |
| Payments with webhooks and refunds | 16 to 40 | $1,600 to $4,000 |
| Fix data rules and permissions | 12 to 40 | $1,200 to $4,000 |
| Emails and notifications | 6 to 12 | $600 to $1,200 |
| Admin screen | 10 to 30 | $1,000 to $3,000 |
| Go live: domain, live and test versions, backups | 8 to 16 | $800 to $1,600 |
| Polish, phone layout, bug fixing | 10 to 30 | $1,000 to $3,000 |
To use a different rate, multiply the hours by it. At $60 an hour, every dollar figure above drops by 40%. At $150 an hour, it rises by 50%.
A worked example with made-up numbers
Say a studio owner built a class booking app in Lovable, an AI tool that builds web apps from chat. It has a calendar and a booking form. It is missing a reliable login, card payments, an admin page and a live setup. The figures are invented to show the method.
| Job | Hours I would plan |
|---|---|
| Paid review | 4 |
| Login and password reset | 12 |
| Payments and webhooks | 24 |
| Admin screen | 16 |
| Go live | 10 |
| Total | 66 |
At $100 an hour that is $6,600. At an assumed $60 an hour it is $3,960, and at $150 it is $9,900. A developer who adds a buffer for the unknown would quote higher. A developer who has done the same jobs many times might quote lower. If the quote you receive is outside that spread, ask why.
Notice what the example does not include: rebuilding the calendar and the booking form. Those work. That is the main reason finishing is usually cheaper than starting over. The page prototype with AI, finish with a developer explains why that split works well.
How to turn the list into a fixed price
- Get the code and the logins first. You need the code, the database login and the hosting login in your own name. If the code is only inside the AI tool, export it. Lovable, for example, documents a two-way sync to GitHub, a service that stores code, and a download option. My guide on exporting code from an AI app builder covers the tools.
- Pay for a short review. Ask one developer to read the code and write the job list. This is the $300 to $600 line in the table. A review gives you a list you can hand to several people for quotes.
- Write each job as an outcome. Not "fix payments" but "a customer pays by card, the booking is marked paid within a minute, and a failed card shows a clear message."
- Ask for a fixed price per job. A fixed price means the developer carries the risk of overruns, so expect a modest buffer in it. Hourly with a cap is the other safe option.
- Pay in stages. Pay after each job works on a test version you have clicked through yourself.
A brief you can paste into a quote request
Copy this and fill the brackets.
- What the app is for, in two sentences.
- Built in: [Lovable, Bolt, Replit, Cursor or other]. Code is in: [GitHub link or exported zip].
- What works today: [list].
- What does not work or does not exist: [list, one outcome per line].
- Services used: [database, Stripe, email, hosting].
- Number of users expected in the first 3 months: [number].
- Deadline and why: [date].
- I need: a fixed price per job, the hours behind each, what you will not do, and who owns the code when you finish.
The hire guide hiring a Lovable, Bolt, Replit or Cursor developer shows what to look for in the person who answers.
What pushes the price up
- Money and personal data. Payments and anything private need more testing. That is where a quick fix becomes a careful one.
- A messy data model. If the database was built without a plan, a developer may need to reshape it before adding anything. Use the checks in rebuild or fix: a decision rule to see if you are in this case.
- No documentation or access. Every missing login costs time to chase.
- Changing your mind. New features during the finish are the largest source of bills above the quote. Write them on a "later" list.
- A native phone app. Apps for the App Store and Google Play add review steps and testing on top of the work above.
What brings the price down
- A clean export in GitHub with a readable history.
- Clear screens or a short written description of each flow.
- One or two small jobs rather than ten. Split the work: finish the core flow first, launch, and add the rest after.
- Doing the small safe jobs yourself with the AI tool, such as changing text and colours, and keeping the developer for login, money and live setup.
If you are weighing the AI tool against a person for the finishing work itself, can AI finish a half-built app or do you need a developer compares the two. For the cost of the AI tools themselves, see how much vibe coding costs, and for a build-from-scratch comparison see AI app development cost.
Red flags in a quote
- A single number for "finishing the app" with no job list.
- No mention of who owns the code and the accounts.
- A suggestion to rebuild everything before anyone has read the code.
- Payment taken in full before any work works on a test version.
- No plan for the live version, backups or what happens when something breaks in week one.
- A developer who says everything is easy. Payments and permissions are never easy to get right.
Who should not hire a developer yet
If the app is a test with no real users and no money flowing through it, the cheapest step may be to keep going in the AI tool until the idea is proven. A developer is worth the cost when real customers, real payments or private data are about to arrive. For this question, the review step is the right first spend.
How I'd approach this
I would not ask "what does it cost to finish?" until I had a list. So I would get the code and logins into my own name, pay for a short review, and turn the findings into jobs with an outcome each. Then I would collect two or three quotes on the same list and compare them line by line.
I would ship the core flow before anything else. A booking app that takes a booking and a payment, and sends one email, is a business. An app with ten half-finished features is not. Everything else goes on a "later" list.
I would also be strict about the human check. Anything that touches payments, logins or other people's data should be read by a person before real customers use it. That is the part of the finish I do not leave to the AI tool.
For reference, Blur Studio finishes and fixes start around $1,500, and the quote follows the same job-by-job method as above. If you want to see how that works, the page linked below explains it.
Questions people ask
How much does it cost to finish an app that was started in Lovable or Bolt?
It depends on the jobs left. In my planning example, a booking app with login, payments, an admin page and a live setup came to about 66 hours, or $3,960 to $9,900 at assumed rates of $60 to $150 an hour. A single missing screen would be far less. Get a paid review first so the list is real.
Is it cheaper to finish an AI-built app or rebuild it?
Finishing is usually cheaper when the screens work and the data is sensible, because you do not pay again for work that already exists. Rebuilding makes sense when the database is badly shaped or the code cannot be run by anyone else. Use the six checks in rebuild or fix: a decision rule.
Should I ask for a fixed price or an hourly rate?
A fixed price per job works best once the list of jobs is written down, because you know your total before work starts. Hourly with a cap on hours is a fair second choice. Avoid open-ended hourly with no cap on a project that has already stalled once.
What should I get from the original builder before I hire someone new?
The code, the database login, the hosting login, any API keys in your name, and a note on what each part does. If the code lives only inside the AI tool, export it first. The guide on exporting code from an AI app builder walks through it.
Can I finish the app myself with the AI tool?
For screens, text and simple flows, often yes. For payments, login rules and who can see which data, have a developer check the result. Lovable's own security documentation says its scans do not replace a thorough security review and recommends a professional review for sensitive data.
How long does finishing take?
From the example hours, a 66-hour job is roughly two to four weeks for one developer, depending on how many hours they can give you a week and how quickly you answer questions. These are planning estimates, not a promise. Ask each developer for dates per job.
More in this topic: Problem solving: AI-built and no-code apps (26 guides) →