In-house AI App Review
Your team built it. We make sure it is safe to run.
More New Zealand businesses are building their own apps with AI tools like Claude Code, Lovable, Cursor and Copilot. That is a good thing: the people closest to the problem are solving it. But an app that works on screen is not the same as an app that is ready for customers, staff data or your brand.
We review what your team has built, tell you plainly where the risks are, and help you close them. We do not take it off you or rebuild it.
Why this matters now.
AI coding tools are fast at making things work. They are less reliable at the parts nobody sees until something goes wrong: who can log in, what data is exposed, where keys are stored, and what happens when the person who built it moves on.
The research is consistent:
-
Veracode testing of over 100 AI models found 45% of AI-generated code samples introduced a common, well known class of vulnerability, and that rate has not improved from 2025 to early 2026.
Veracode, via the Cloud Security Alliance -
Georgia Tech researchers traced 35 published vulnerabilities to AI-generated code in March 2026 alone, more than in all of 2025.
Georgia Tech, Vibe Security Radar -
Wiz found one in five vibe-coded apps it studied had serious vulnerabilities or configuration errors.
Wiz, via Kaspersky
None of that is a reason to stop building. It is a reason to have someone independent check the work before employees, customers or partners start using it.
What we check.
Does it do the job?
We start with the outcome you need, not the code. We check the app against what the business asked for, test the paths people actually use, and flag where it quietly does something different.
Privacy and personal information
What personal information it collects, where it is stored, who can see it and how long it is kept. We check it against the Privacy Act 2020 principles, including IPP 3A, which since 1 May 2026 covers personal information you collect indirectly.
User access and controls
Who can sign in, what each person can see and change, and whether one customer can reach another customer's data. Admin access, password and sign-in handling, and what happens when someone leaves.
Data security basics
Exposed keys and passwords in code, database access rules switched off, unprotected endpoints, out of date packages, and what the app sends to outside AI services.
Ownership and handover
Is it in your accounts, your code repository and your hosting? Can someone other than the builder deploy it, test it and roll it back? Is it documented well enough to survive a staff change?
How it works.
-
Discovery call (free, 45 minutes). What the app does, who uses it, what data it touches and what worries you.
-
Statement of work. We'll provide a SOW to outline the scope, requirements and outcomes we'll deliver to ensure your custom solution is ready to go live.
-
Review. We look at the app, the code and the setup in your environment, with your team. Access is read only unless you agree otherwise, and it stays in your accounts.
-
Findings, in plain English. Each issue rated by risk, with what it means for the business and what to do about it. No jargon dump.
-
Fix it your way. Your team or your developer makes the changes, with us on hand. Or we make them, under a fixed scope you approve first.
-
Re-check and sign off. We confirm the high risk items are closed, so you can go live with confidence.
What you get.
- A findings report your leadership team can read, rated high, medium and low
- A prioritised fix list your builders can work through
- A short go-live checklist for this app and the next one
- A handover pack: how to deploy, test and roll back, and who owns what
What this is not.
- Not a rebuild. Your team keeps the app and the momentum.
- Not a penetration test or certification. If your risk calls for formal security testing, we will say so and bring in a specialist partner.
- Not legal advice. We check against the Privacy Act principles and good practice. Your lawyer signs off on legal questions.
Who it is for.
Businesses where a manager, analyst or small internal team has built a tool with AI and it is about to face customers, partners or staff data
Leadership teams who said yes to in-house building and now want to know what they have actually approved
Organisations publishing an internal app to a second brand, site or customer group
An example.
A New Zealand building products manufacturer had built two customer-facing configurators. We moved them into the client's own infrastructure, reviewed the code, and set up a test, deploy and roll-back process their team now runs. When they needed one configurator to serve two brands from a single deployment, it was a small, fixed scope change rather than a rebuild, because the foundations had been checked first.
Investment.
Fixed fee, scoped to the size of the app and what it touches. We quote after the Discovery call, and you approve the SOW before any work starts.
Questions.
We built it with AI. Is that a problem?
No. How it was built matters less than whether it has been checked. We review apps built with any tool.
Will you need access to our systems?
Yes, to your code and setup. We work inside your accounts, read only by default, and never ask for passwords in email or chat.
How long does a review take?
Most single apps take days, not weeks. We confirm timing in the Discovery call.
Can you just fix everything for us?
We can, under a fixed scope you agree first. Most clients prefer their own team to make the fixes so they learn the patterns for next time.
What about the next app?
The go-live checklist is yours to reuse. Teams on our AI Operating System get this kind of check built into how new AI workflows are designed and approved.