Lock-in and exit plans: what you can take with you from AI app builders

The short answer

Lock-in with AI app builders is more than whether you can export code. It includes your data, the logic that only runs on the platform, a pricing meter the vendor can redefine, product lines that split, and knowledge that lives in one person's head. Vendors including Bubble, Vercel, Microsoft and Glide have changed pricing or products after customers built on them. Before you commit, write an exit plan: what you would take, in what format, who would run it, and what it would cost.

Nobody buys software planning to leave it. But the companies that get hurt worst by a platform are rarely the ones that left. They are the ones that wanted to leave and found they could not: the price changed, the product split, the person who built everything quit, or the data sat in a format nobody could use.

AI app builders make this question more urgent, because they make it so easy to build a lot quickly. In a few months, a team can go from zero tools to a dozen that the business runs on. This guide explains the five kinds of lock-in to watch for, shows real changes vendors have made after customers built on them, and gives you an exit plan you can write in an afternoon.

Why "can I export the code?" is the wrong first question

Vendors and buyers often treat lock-in as a single yes or no: can you export the code? It matters, but for a business without developers it is rarely the deciding factor.

Exported code is a set of files. To be useful, someone has to run those files on a server, connect them to a database, set up sign-in, keep the security current and fix what breaks. A team that chose an AI builder because it has no developers usually has nobody who can do those things. The export exists, but leaving still means rebuilding.

So the better question is: if we had to move, what would we keep, who would run it, and what would it cost? That breaks down into five kinds of lock-in.

The five kinds of lock-in

1. Data lock-in

Your records (jobs, customers, inspections, approvals) are the most valuable thing in any internal tool. Ask:

Where does the data live: on the builder's own hosting, a third-party database, or your systems?
Can you export all of it, including files and photos, in a standard format such as spreadsheets?
Does the export keep relationships, for example which inspection belongs to which property?
Can you export it on demand, or only on a higher plan or through support?

2. Logic lock-in

The rules in your app (who approves what, what happens when a job is marked done, which fields are required) are your business process written down. On a visual builder, those rules live in the platform's own format. On an AI builder, they live in generated code that only runs where the builder set it up. Either way, moving means re-creating the rules somewhere else. Keep a plain-language description of each tool's rules outside the platform, so a rebuild starts from a spec, not from memory.

3. Pricing lock-in

This is the one buyers underestimate. Once your team depends on a tool, the vendor's pricing becomes your cost structure, and vendors change pricing.

Bubble introduced workload unit pricing in April 2023, a meter based on how much server work an app does. The change drew heavy pushback on Bubble's forum, where one user wrote that the estimates "do not reflect what a lot of users are seeing." Customers who had built on the old model had to re-evaluate apps that were already running.
Vercel moved v0 to token-metered credits on May 13, 2025. In its own words: "Usage is now metered on input and output tokens which convert to credits, instead of fixed message counts."
Microsoft says the Power Apps per app plan has not been available to new customers through some purchasing channels since January 2, 2026, according to Microsoft Learn.

None of these vendors acted in bad faith. Pricing evolves. The point is that a platform's meter can be redefined after you have built on it, and a meter you cannot predict today is harder to predict after a change.

A related form is the shared meter. Lovable's pricing page says its credits pay for building, hosting with Lovable Cloud and AI features inside your apps from one balance. That is convenient, and it also means your running apps and your building draw on the same pool, so hosting and building cannot be separated easily if you move one and not the other.

4. Product lock-in

Sometimes the product itself changes shape. Glide's pricing page states that GlideOS is a separate product from Glide Classic with its own plans. Teams on one product line need to read which pricing, features and roadmap apply to them. When a vendor launches a new generation, ask what happens to apps on the old one, and whether moving between them is supported.

5. Knowledge lock-in

The most common lock-in has nothing to do with the vendor. It is the employee who built every tool. When that person leaves, nobody knows why the approval rule works the way it does, which fields feed the monthly report, or what was changed last spring. AI builders can make this worse, because tools get built fast through a long chat history that only the builder ever read.

Knowledge lock-in is fixed with records and people: a record of who changed what and when, more than one person trained to build, and a short written description of each tool.

Side by side: what to ask about each kind

Kind of lock-in	What it looks like	The question to ask	Warning sign
Data	Records you cannot get out cleanly	"Show me a full export of one app's data, with files."	Export only through support or on a top tier
Logic	Rules that only run on the platform	"If we rebuilt elsewhere, where are our rules written down?"	The only description is the app itself
Pricing	A meter you depend on and cannot predict	"How has your pricing changed in the last three years?"	Units nobody on your team can forecast
Product	A product line that splits or is retired	"What happens to apps on your older product?"	No answer, or no migration path
Knowledge	One person understands everything	"Who else could change this tool tomorrow?"	A single name, and no record of changes

What leaving actually costs

It helps to put a number on the worst case, because it tells you how much protection is worth paying for.

If you had to rebuild your tools somewhere else, the work would fall into three buckets:

Moving the data. Usually the easiest part if you have a clean export, and the hardest if you do not. Budget time for checking that every record, file and photo arrived.
Re-creating the rules. Every approval step, status and permission has to be rebuilt. With a plain-language description of each tool, this is a few days per tool. Without one, it starts with interviewing the people who use it.
Retraining the team. Every person who uses the tool learns a new screen. This cost grows with headcount, which is one more reason app-user pricing matters.

If the rebuild requires a developer, the cost climbs fast. The U.S. Bureau of Labor Statistics put the median annual wage for software developers at $135,980 in May 2025, before benefits, recruiting and the months it takes to get up to speed.

What a good export looks like

When you run a test export, check for these:

Every table, not just the main one, in a format such as spreadsheet files you can open.
Attached files and photos, with a way to tell which record each belongs to.
History: who created and changed each record, and when.
A list of users and their roles.
A description of the app's rules, even a basic one.

If an export gives you only some of this, write down what is missing. That gap is the real size of your lock-in.

When something goes wrong

Lock-in matters most in a crisis, when you want options and do not have them.

In July 2025, Fortune reported that a Replit agent wiped SaaStr's live database, holding data on more than 1,200 executives and over 1,190 companies, during a code freeze. Replit later promised separation of development and production data and a planning-only mode. A business in that position needs to know, in advance, where its backups are and who can restore them.

Lovable says a February 2026 regression exposed public projects' chat history and source code from February 3 to April 20, 2026, and that it fixed the problem within two hours of the public report. The lesson for an exit plan: know where your tools' source and history live, and who can see them.

Write your exit plan in one afternoon

You do not need a lawyer or a developer to do this. You need an hour per important tool.

List every app. Name, owner, what it does, how many people use it, and what breaks if it stops.
Rank them. Mark the three the business could not run without for a week.
Do a test export. For each of those three, export the data and open it. Check that files, photos and history came with it.
Write the rules down. In plain words, describe who does what, who approves, and what each status means.
Name a second person. For each important tool, name someone besides the builder who could change it.
Record the cost. Write what the platform costs today and what it would cost if usage doubled.
Check the contract. Find the notice period for price changes and what happens to your data if you cancel.
Review it twice a year. Add new apps, retire dead ones, and redo the test export.

How to choose a platform with less lock-in

Prefer pricing you can count. A meter you can forecast in one sentence is less likely to surprise you after a change.
Look for pricing that does not charge app users. Then growing the rollout never deepens the dependence on a per-user contract.
Ask for a test export before you sign. Do it with one real tool during the trial.
Require a record of changes. It is the cure for knowledge lock-in.
Insist on onboarding for more than one person. Two trained builders are worth more than any export button.
Read the vendor's pricing history. Search their forum and blog for past changes and how they were handled.

How FORGE handles it

FORGE is built so that the company, not a single employee and not a meter, stays in control of its tools.

Every app lives inside your company. People join by invite only, and each person's role (owner, admin, builder or app user) decides what they can see and change. There is a complete record of who did what, so the history of every tool is written down, and the next person can pick up where the last one left off. Our team onboards you ourselves, starting with a workflow workshop, so more than one person on your team knows how to build.

Pricing is built to be counted, now and as you grow. It is $49 a seat a month for each person who builds, with 3 builds included. A change is a small fraction of a build and only counts when something actually changes. Packs of 10 builds are $150. People who only use the apps are free, so rolling a tool out to everyone never raises your bill. A build FORGE fails is free, owners get alerts at 80 and 100 percent, and nothing ever bills by itself.

The tools themselves are built to last. A panel of top AI models plans each one, a dedicated AI crew builds it, and every piece is checked for quality and security before it is saved, with the tool checked at every step. FORGE builds web, mobile and desktop apps, and most teams have their first real tool the same day.

To walk through your own exit plan and what your first tools would look like, talk to us, or read about security and pricing.

Questions people ask

What does lock-in mean for an AI app builder?

Lock-in is anything that makes leaving a platform costly. It can be data you cannot easily take out, app logic that only runs on that platform, pricing that rises after you depend on it, or knowledge of how the tools work that only one person or the vendor holds.

If a builder lets me export code, am I safe?

Not entirely. Exported code is only useful if someone can run it, host it, connect it to its data and keep it secure. For a team without developers, a code export is often a file nobody can use. Ask what the export includes and who could run it.

Have AI and no-code vendors changed pricing after customers built on them?

Yes. Bubble introduced workload unit pricing in April 2023 and drew heavy pushback on its forum. Vercel moved v0 to token-metered credits on May 13, 2025. Microsoft's Power Apps per app plan has not been available to new customers through some channels since January 2, 2026.

What is the most overlooked kind of lock-in?

Knowledge lock-in. When one employee built every tool and nobody else understands how they work, the company is locked in to that person, regardless of the platform. A record of who changed what, and more than one trained builder, are the fix.

What should an exit plan include?

A list of your apps and what each one does, where each one's data lives and how to export it, who could run each app elsewhere, a test export you have actually done, and the cost of rebuilding the most important tool somewhere else.

How does FORGE reduce lock-in risk?

Every app lives inside your company, there is a complete record of who did what, pricing is countable with no charge for app users, nothing bills by itself, and our team onboards more than one person so knowledge does not sit with a single builder.

See what FORGE builds for your team

We onboard every team ourselves, starting with the work that costs you the most time.

Talk to us