You built something useful with AI, and people now rely on it
Before you pay a developer to rebuild it, get an independent read on where it holds, where it doesn't and what comes next. Whoever built it, or a developer, makes the fixes, so there is no repair work to sell you afterwards.
Dr Colin Kelly
PhD in NLP (Cambridge)
Former Head of Applied AI Research
Fractional CTO who took an AI product from concept to public beta
Starts with an AI readiness review: $6,000 fixed, or $8,000 with AI inside
AI inside means the tool sends your data to an AI model, such as ChatGPT or Claude, to answer, decide or act.
When what you built outgrows the way it was built
You, or whoever on your team built it, made something useful with AI tools, quickly, and it does a job the business needed.
"It works" usually means it does the job on the cases you tried, used by people who know its quirks. That stops being enough once it is in production: more people use it, it holds customer data or acts for the business, and customers depend on it. It often shows when the platform changes or switches something off and it goes down, someone wants it rolled out more widely, a developer looks at it and hands back pages of changes, a customer sees something they shouldn't, or the AI or platform bill jumps.
The usual next move is to hand it over to a developer, who may rebuild it the way they would have done it. An independent read first tells you what comes next, including whether you need a developer at all.
I give you that read, written down, and the decisions stay with you.
Who this is for
Owners and MDs of small firms whose team built a business tool, agent or app with Lovable, Replit, Cursor, Claude or similar, and the business now runs on it
Founders who aren't technical, with something built with AI tools that customers are starting to rely on
Also a fit: a team lead in a larger firm who built something useful that other people now depend on, several tools that now pass work between them, or a developer who has started building with nothing written down on your side. The review starts from what exists.
Not a fit
- A personal tool only you use: AI help and the free go/no-go check
- Someone to make the technical decisions and lead your people: that's a Fractional AI CTO
How it works and what it costs
Every engagement starts with the review. What comes next depends on what it finds.
AI readiness review
$6,000 fixed, or $8,000 fixed when there is AI inside, one week, with no repair work to sell you afterwards
I check who can see your customers' data and how it is deleted, whether you own the accounts and could move off the platform, and who would notice a mistake. When there is AI inside, I also check what it does on your real cases and what it can do without a person. Where the fixes are a setting, a permission, a prompt or a log, the report is the whole job. I need read-only access to the code; without it, the report marks what depends on the code as Unverifiable.
Output: A verdict (Ready, Ready with conditions, or Not yet), the order of fixes and what kind of developer you need, if any, in a written report you can show to a client or an IT team
The design and business case
From $12,000, about two weeks, only if it needs redrawing
When the fixes go deeper, such as redrawing the workflow or who can see what, moving off the platform or changing AI supplier, I write the design for the next version and a short business case: rebuild it, buy a product, or set up tools you already pay for, each with a price range.
Output: A written design, and the brief a developer quotes against
The monthly check
A fixed monthly fee, with an end date agreed at the start
While the fixes or the rebuild go ahead, I review decisions as they come up, read developer quotes with you and check the work against the review or the design. I tell you in plain words when it strays, and change the design in writing if something better comes along.
Output: A plain read on each change, and the design kept current in writing
What I look at
Seven checks run on every review, and five more when there is AI inside it. Three others run where they apply, for example when it searches stored documents or lets the public sign up.
Every review
Customers and their data
- •Whether the keys that connect it to other services are hidden from visitors
- •Whether one customer can see, change or delete another customer's records
- •How customer data is deleted when someone asks
Every review
Running it and answering for it
- •Whether you own the accounts and the data, and could export the data and move off the platform
- •How you would switch it off, and what happens when the platform or a provider changes or goes down
- •Whether you could trace what happened if a customer complained
- •Who in the business answers for it
When there is AI inside
The AI inside it
- •Whether it gives the right answer on 20 of your real cases, plus 5 harder ones I write
- •Whether someone can type, paste or upload text that makes it ignore its instructions
- •What it can do on its own, and which actions need a person to confirm
- •Which customer data goes to which AI provider
- •Whether users know it is AI and can reach a person
Working with your existing developer
A developer or agency you already work with keeps responsibility for the build. The review, and the design if there is one, are the brief you both work from, so the business side and the technical side read the same document.
Related insights
Self-improving systems: the AI architecture pattern everyone talks about, nobody builds
Despite the hype, truly self-improving AI systems remain theoretical due to fundamental technical and organizational barriers, with today's "self-improving" implementations being merely constrained optimization within predetermined parameters.
Cost-effective LLM implementation: when to fine-tune and when to prompt
Most companies are burning money on LLM implementations by defaulting to expensive fine-tuning when sophisticated prompting could achieve comparable results at a fraction of the cost and complexity.
Frequently Asked Questions
How much does an independent AI architect cost?
Agathon's AI readiness review costs $6,000 fixed, or $8,000 fixed when the tool has AI inside, and takes one week. AI inside means the tool sends your data to an AI model, such as ChatGPT or Claude, to answer, decide or act. If the tool needs redrawing, the design and business case costs from $12,000 and takes about two weeks. The monthly check after either is a fixed monthly fee, agreed in writing at the start.
Do I need a developer to fix an app I built with AI tools?
Whether an app built with AI tools needs a developer depends on the fix: whoever built it can often change a setting, permission, prompt or log, but changing how the app is built usually needs a developer. Agathon's AI readiness review sorts the fixes and, where a developer is needed, says what kind and what to do.
Will a developer want to rebuild my app from scratch?
A developer taking over an app built with AI tools may want to rebuild it, and sometimes that is right, but a setting, permission or missing log needs no rebuild. Agathon's AI readiness review sorts the fixes by kind before the handover, and the design runs only if the app needs redrawing.
Why not ask an AI tool to review or fix my AI-built app?
An AI tool reviewing AI-built work answers only what it is shown, and grades its own kind of work. Agathon's AI readiness review tests the running tool on your real cases and ends in a next-step decision that I put my name to. For a personal tool only you use, AI help and the free go/no-go check are usually enough.
Can someone review my vibe-coded app before it goes into production?
A vibe-coded app, built by describing it to an AI tool, can be reviewed like any software, by checking who sees its data and what happens when it breaks. Agathon's AI readiness review covers any business tool, agent or app built with AI tools, adding the AI checks when there is AI inside.
Who fixes the problems an AI readiness review finds?
Agathon's AI readiness review leaves the fixes to whoever built the tool or a developer. The report puts the fixes in order and says which need a developer. There is no repair work to sell you afterwards.
How is an independent AI architect different from a fractional CTO?
An independent AI architect recommends and leaves the decisions with you; a fractional CTO makes the technical decisions and leads your people. With Agathon's AI architect service, I recommend and write down why, and you approve.
Can an independent AI architect work with my existing developer?
An independent AI architect can work beside your developer or agency, giving both the same written plan. With Agathon, your developer is invited to the review readout and stays responsible for the work.
Does Agathon's AI architect service use a team or one person?
Agathon's AI architect service is one person, me, with no juniors, doing the review, the design and the monthly check.
What do I need to provide for an AI readiness review?
Agathon's AI readiness review needs read-only access to the code (on GitHub or the platform it was built on) and to the settings, a 30 to 45 minute walkthrough, two test accounts and, when there is AI inside, 20 real cases with the answers you would accept. Without the code, the review runs from the outside and marks code-dependent findings, such as whether a developer could take it over, as Unverifiable. Where settings access isn't possible, you can show me them on a screen-share instead, and the report marks those findings as seen rather than checked directly.