Before you sign with an AI platform, check eight things: how it keeps your data away from outsiders, how people join and leave, what record it keeps of every change, how it checks what the AI builds, how predictable the bill is, whether you can take your work with you, what onboarding and support you get, and how the vendor has handled past incidents. Get written answers for the exact plan you will buy, because on many platforms the governance features you need sit on the most expensive tier.
Choosing an AI platform for your team is not like choosing a note-taking app. Your people will put customer records, schedules, prices and employee details into the tools they build. The tools will become part of how work gets done. If the platform is wrong, the cost shows up later as a data exposure, a surprise bill, or a pile of tools you cannot move.
This guide is a complete due diligence checklist for owners, IT leads and finance leads. It covers the eight areas to check, the exact questions to ask, a table of where common governance features sit on well known platforms, a scoring sheet you can copy, and how to run a demo that tells you the truth. Every fact about other vendors is linked at the bottom, with prices as listed in October 2026.
The eight areas to check
| Area | The core question | Why it matters |
|---|---|---|
| 1. Data protection | Can anyone outside our company reach our apps or data? | Missing access rules are the most common cause of exposed records in AI-built apps |
| 2. Access and identity | How do people join, what can each role do, and how do they leave? | Access that does not end when employment ends is a standing risk |
| 3. Record of changes | Can we see who did what, and when? | Without a record, problems cannot be traced or explained |
| 4. Build quality | How is AI output checked before it becomes part of a tool? | AI output is often almost right, and almost right breaks things |
| 5. Pricing | Can our finance lead forecast next month in one sentence? | Usage-based bills and per-user charges surprise teams |
| 6. Ownership and exit | Do the tools belong to the company, and can we take our data with us? | Tools in personal accounts leave when people leave |
| 7. Onboarding and support | Who helps us get our first real tool working? | A tool nobody adopts is money spent for nothing |
| 8. Vendor track record | How has the vendor handled past incidents? | Responses to past problems predict responses to yours |
1. Data protection
The documented failures in AI-built apps come mostly from access rules that were missing or wrong. In March 2025, researcher Matt Palmer found 170 of 1,645 scanned Lovable apps exposed through missing database access rules, recorded as CVE-2025-48757. Lovable disputes the CVE, saying each customer is responsible for protecting their app's data, according to the US National Vulnerability Database. That tells you to ask every vendor directly who is responsible for protecting the data inside the apps your team builds.
Questions to ask:
- Does every app live inside our company, with every request checking that the person belongs to our company?
- Can an app be published to the open internet? If so, who can do it and how do we prevent it?
- Who is responsible for protecting the data inside apps we build: you or us?
- Is any of our data used to train AI models, on the plan we will buy? Our guide to what AI app builders do with your data has the full list of data questions.
2. Access and identity
Three plain definitions help here. Single sign-on (SSO) lets staff log in with one company account. SCIM lets your staff directory add and remove people in other tools automatically. Role-based access means each person's role decides what they can see and do.
Questions to ask:
- How do people join? Is it invite-only, and do invites expire?
- What roles exist, and can people who only use apps be prevented from changing them?
- How do we remove someone, and does their access end immediately?
- What happens to the tools a person built when they leave?
- Which of these features are included in the plan we are pricing?
That last question matters because on many platforms the features in this section sit on the top tier. Here is what vendors list, as of October 2026:
| Platform | Where common governance features sit (as listed in October 2026) |
|---|---|
| Lovable | SSO starts at Business; SCIM and audit logs are Enterprise only |
| Bolt | SSO, audit logs and data retention policies are Enterprise features |
| v0 | SAML SSO and role-based access control are Enterprise only |
| Replit | SSO, SCIM and audit logs are Enterprise only |
| GlideOS | SSO and audit logs are Enterprise only |
| Softr | SSO, SCIM and audit logs are Enterprise only |
| Retool | SAML/OIDC SSO and SCIM are Enterprise only; audit logs start at Business |
| Google AppSheet | Okta, Active Directory and OIDC sign-in, plus governance policies, are Enterprise Plus only |
| Airtable | SAML SSO is available on Business and Enterprise Scale only |
None of this makes these products bad. It means the price on the front of the pricing page may not be the price of the plan your IT lead needs. Our guide to role-based access for employee-built apps explains what roles you actually need.
3. Record of changes
Questions to ask:
- Is every build and every change recorded with who asked for it and when?
- Can admins see who invited whom and who changed roles?
- How long is the record kept, and is it included on our plan?
4. Build quality
AI output is often almost right. In the 2025 Stack Overflow Developer Survey, 66 percent of developers named "AI solutions that are almost right, but not quite" as their top frustration. Ask how the platform catches that before it reaches your tools.
- What is checked before a change becomes part of a tool?
- Is security checked on every piece as it is built, or once before publishing?
- Can the AI act directly on the data our team relies on while it builds? See what happened when an AI agent deleted a production database.
- Do we pay when the AI fails?
5. Pricing
Pricing models vary widely. Some vendors charge by credits or tokens, some per seat, and some charge for every person who uses an app. A few examples, as listed in October 2026:
- Power Apps: Premium is $20.00 per user per month paid yearly, and the people who run the apps need a license too.
- AppSheet: $5, $10 or $20 per user per month, and every app user, including guests, needs a license.
- Retool: Business is $50 per builder and $15 per internal user per month billed annually.
- Softr: each plan caps the people who can log in to your apps, with extra app users billed monthly on Pro and above.
- Lovable: no per-seat charge; plans are priced by credits, and credits cover building, hosting and AI features inside apps from one balance.
- Bolt: Teams is $30 per member per month, and tokens are not shared between members.
Questions to ask:
- What do people who only use the apps cost?
- What happens when we run out of credits, tokens or builds? Does anything bill automatically?
- Can we set a cap, and do we get alerts before we reach it?
- Do failed attempts by the AI count against our allowance?
Our guides to AI app builder credits and preventing surprise AI software bills go deeper.
6. Ownership and exit
- Do tools belong to the company, or to the person who built them?
- Can we export our data in a common format at any time?
- What happens to our data if we stop paying?
- How are backups and recovery handled? See our guide to backups, rollback and recovery.
7. Onboarding and support
- Who helps us build our first real tool, and is it on our work or a demo?
- Is there a walkthrough for our IT lead?
- How do we reach support, and how fast do they respond?
8. Vendor track record
Every vendor eventually has a problem. Look at how it responded. A few public examples:
- Lovable, April 2026: The Register reported that Lovable first said "We did not suffer a data breach" before apologizing for an exposure of public projects' chat history and source code.
- Base44, July 2025: Wiz found a flaw letting anyone with a public app ID register on private apps, bypassing SSO. It was fixed in under 24 hours, with no evidence of exploitation.
- Microsoft Power Apps, August 2021: UpGuard found 38 million records exposed across 47 portals; Microsoft first called it "by design."
- Retool, September 2023: an SMS phishing attack compromised 27 cloud customers, all in crypto, and Retool published a detailed account.
Ask each vendor: what was your most serious security incident, how did you find out, how fast did you fix it, and what did you change?
A scoring sheet you can copy
Score each area from 0 to 2 for every vendor you consider. 0 means no clear answer, 1 means partly, 2 means a clear written answer that meets your need on the plan you will buy.
| Area | Vendor A | Vendor B | Vendor C |
|---|---|---|---|
| Data protection | |||
| Access and identity | |||
| Record of changes | |||
| Build quality | |||
| Pricing predictability | |||
| Ownership and exit | |||
| Onboarding and support | |||
| Track record | |||
| Total out of 16 |
Run a demo that tells the truth
Do not let the vendor choose the demo. Bring a real workflow.
- Pick one real, painful workflow, such as job dispatch or purchase approvals.
- Ask the vendor to build it live, described in plain words by someone on your team.
- Sign in as a limited role and check what it can see.
- Open the record of what was just built.
- Ask what that session would cost on the plan you would buy.
How FORGE handles it
Here is how FORGE answers the eight areas, in plain terms.
- Data protection: every app lives inside your company, and every request checks it. Anyone outside gets nothing back.
- Access and identity: invite-only by email, every invite works once and expires, and there is no public sign-up. Owners, admins, builders and app users each see and do exactly what their role allows. Removing someone ends their access immediately.
- Record of changes: every step of every build is recorded with who or what did it.
- Build quality: a panel of top AI models plans each tool, a dedicated AI crew builds it, every piece is checked for quality and security before it is saved, and the tool is checked at every step. A build FORGE fails is free.
- Pricing: $49 a seat a month for each person who builds, with 3 builds included. Changes are a small fraction of a build and only count when something changes. Packs of 10 builds are $150. App users are free. Caps and alerts at 80 and 100 percent, and nothing bills by itself.
- Ownership: every tool your team builds stays with your company when people leave.
- Onboarding: our team onboards you, starting with a workflow workshop on your real work, including a walkthrough for your IT lead. Most teams have their first real tool the same day.
- Apps your way: web, mobile and desktop apps.
Bring this checklist to the workshop. Talk to us.
Questions people ask
What should we check before buying an AI app platform?
Data protection, access control, the record of changes, how AI output is checked, pricing predictability, data export and exit, onboarding and support, and the vendor's incident history. Ask for written answers for the plan you will actually buy.
What is SSO and do we need it?
Single sign-on, or SSO, lets staff log in with one company account instead of a separate password for each tool. It is useful for companies that already manage staff accounts centrally. On many AI platforms it is only included on business or enterprise plans, so check the price of the plan that includes it.
What is SCIM?
SCIM is a standard that lets a company's staff directory add and remove people in other tools automatically. Larger companies use it so a departing employee loses access everywhere at once. Smaller companies can get the same result if the platform makes removal a single step.
What are audit logs?
Audit logs are a record of who did what and when inside a platform. They let you answer questions like who changed a tool, who invited a person or who deleted a record. Several AI platforms list them only on enterprise plans.
How do we compare AI platform prices fairly?
Model a realistic month for your team: how many people build, how many only use the apps, and how much building you expect. Include charges for app users, usage-based overages, and the plan you need for the security features you require.
Should we ask about a vendor's past security incidents?
Yes. Every vendor will eventually face a problem. How it responded, how quickly it fixed it and whether it accepted responsibility tells you more than its marketing does.
How does FORGE answer these questions?
Every FORGE app lives inside your company, people join only by invite, four roles are enforced, every piece is checked for security before it is saved, and every step is recorded. It is $49 a seat a month with 3 builds, app users are free, and nothing bills by itself. Your IT lead gets a full walkthrough in the workflow workshop.
See what FORGE builds for your team
We onboard every team ourselves, starting with the work that costs you the most time.
Talk to usSources
- Lovable plans
- Bolt pricing
- v0 pricing
- Replit pricing
- Glide pricing
- Softr pricing
- Retool pricing
- Google AppSheet pricing
- Airtable: configuring SSO
- Microsoft Power Apps pricing
- NVD: CVE-2025-48757
- The Register: Lovable denies data leak (April 2026)
- Wiz: critical vulnerability in Base44
- UpGuard: Power Apps portals data exposure
- Retool: MFA isn't MFA
