The AI Agent That Assumes: Why Automation Needs to Ask, Not Guess
AI agents are good at filling in gaps quietly, which is exactly the problem. Here's why business automation needs to be built to ask rather than assume, with real examples.
AI prototype to production
Turn your AI-built prototype into production-ready software
You built something real with AI. Now it needs to be safe, reliable and maintainable enough to put customers, payments and a business behind it. That's the gap I close, as an experienced developer who's shipped production software.
01Overview
AI development tools have changed what a non-developer, or a small team, can build in a weekend. That's a genuinely good thing. But the same speed that gets you a working prototype tends to skip the parts that only matter once real people, real data and real money are involved. I take what you've built and make it production-grade, without throwing away the work that already works.
02The gap
A demo that works is not the same as a product you can trust
Tools like Lovable, Bolt, Replit and Cursor, and AI-assisted development generally, are excellent at turning an idea into something you can click through. They get you to a tangible prototype faster than has ever been possible, and that's worth a lot when you're testing an idea.
The catch is that a prototype is optimised to look right in a demo, not to survive contact with real users. The things that keep a live product safe and stable, proper access control, sensible data handling, error recovery, payments that don't misfire, tend to be exactly the things that get skipped to move fast. That's not a criticism of the tools. It's just a different job, and it's the job I do.
03What I review
Where the gap usually shows up
When I look at an AI-built prototype, these are the areas that most often need attention before it's ready for customers.
Architecture & data
How the app is structured and how your data is modelled. Prototypes often work fine until the data gets real, then hit walls. I make sure the foundations hold as usage grows.
Security & access
Authentication, permissions and, for multi-tenant products, keeping each customer's data properly separated. This is where prototype shortcuts cause the most harm.
Payments & correctness
If money changes hands, the payment flow and the awkward edge cases around it need to be right, with tests on the parts where a bug would actually cost you.
Reliability
Error handling, background jobs, retries and logging, so when something goes wrong you find out and recover, instead of a customer telling you first.
Secrets & configuration
API keys, credentials and environment setup handled safely and kept out of places they can leak from.
Deployment & monitoring
A proper deployment with real environments, plus monitoring so you can see how the live product behaves and how it holds up under load.
04Where you're at
Which of these sounds like you?
Wherever you are with your prototype, there's a sensible next step.
You have a prototype but aren't sure what needs fixing
I review it and give you a clear, honest picture of what's solid, what's risky and what to do first, in plain language.
Ask for a reviewYou know what needs fixing
You've spotted the gaps and want an experienced pair of hands to close them properly rather than patch over them.
Help me productionise itIt has real users and problems are emerging
The prototype got traction, and now the cracks are showing: bugs, slowdowns or things breaking under real use. I stabilise it and make it dependable.
Help me stabilise itYou want someone to take over development
You'd rather hand the code to an experienced developer and focus on the business. I take ownership of the codebase and carry it forward.
Take over my build
05FAQs
AI prototype questions
What's wrong with building in AI tools like Lovable or Bolt?
Nothing. Tools like Lovable, Bolt, Replit and Cursor are genuinely good at turning an idea into a working prototype quickly, and that is valuable. The gap is between a prototype that works in a demo and software that is safe and maintainable enough to put customers, payments and a business behind. That gap is what I bridge.
Will you rewrite everything from scratch?
Usually not. I start by understanding what you have. Often a lot of it is fine and the sensible move is to fix the specific weak points, harden the risky parts and build on the rest. I will tell you honestly when a rebuild of a particular area is genuinely the better call.
Can you just review my prototype first?
Yes. A review is a good, low-commitment starting point. I go through the code and give you a clear picture of what's solid, what's risky and what to prioritise, so you can decide what to do next.
Do you provide security guarantees or penetration testing?
I apply sensible, professional security practices and fix the issues I find, but I do not sell formal security certification or penetration testing, and I would not claim a system is guaranteed secure. Where a formal audit is needed, I will say so and can point you in the right direction.
I'm not technical. Can you explain what needs doing?
Yes. You will not get a wall of jargon. I explain what I have found and why it matters in terms of your business and your customers, so you can make informed decisions.
Who owns the code once you've worked on it?
You do. You own the code and the accounts, with no lock-in, exactly as you should.
Can you keep developing it afterwards?
Yes. Many people come for a review or a fix and stay for ongoing development. Once I know your codebase, I am well placed to keep building on it.
Insights
Related reading
You Inherited a System Nobody Can Explain. Now What?
When you inherit software with no documentation and no developer to ask, the temptation is to rewrite it fast. Here's a safer way to work out what it actually does first.
Your AI Coding Assistant Doesn't Remember Last Week's Decisions
AI coding agents don't carry memory of past decisions between sessions the way a human developer does. That gap shows up as inconsistency in your codebase, and it matters more than most people realise.
Show me what you've built
Send me a link or a login and a couple of lines on what you're trying to do. I'll give you an honest view on what it takes to make it production-ready.
Or email me at josh@thackr.co.uk