Software for startups and growing businesses
We build systems that are robust and easy to read, test and extend. The speed comes from AI. The quality comes from what it is given to work to, and from the people who check what comes back.
Fast and good
That trade-off goes away when the work before the build is done properly. Here is where each half comes from.
How a build runs
Most of the decisions that make a system cheap or expensive to own are made before the first line of code. So that is where the work starts.
A set period to pin down what the business does, what the product has to do for it, and what it doesn't need yet.
Domain-driven design splits the problem. The part that makes you different gets built. The rest is bought.
Usually small services, each owning its own data. The principles and practices the code must follow are written down.
Build once, deploy many. Each service deploys on its own, with its infrastructure, so one change never drags the rest along.
Before a ticket runs, a three amigos session agrees what it is for, how it will be tested and what will be measured in live.
One model manages the work and owns the tests. Another writes the code against them. Agents do the rest.
Integration tests at each service endpoint, automated by the tester rather than run by hand.
Done means live. Problems found by telemetry are fixed forward: a ticket is raised, approved by a person and built.
Services
The build
A system designed around how your business works, built with AI to a written standard and delivered through to production.
Quoted to deliver, with continuing support.
Afterwards
Once the system is live, it still needs someone holding it to that standard and planning what comes next.
Monthly retainer, a few days a month.
The team you need
It depends on the project, but a startup's first platform rarely needs more than this. The money a smaller team saves is better spent getting the product in front of customers.
Who also own how their code gets to production and runs there.
Who owns acceptance and automates the checks instead of running them by hand.
When the business needs someone dedicated to deciding what comes next.
Built this way
A property analytics platform that works out the most an investor should pay for a property, based on the cost of the deal rather than the asking price. A vision model reads the listing photos, and the analysis is available to AI assistants through MCP servers as well as the web app.
It runs as a set of services on Azure, each deploying on its own, with the build following the same process described above.
Earlier work, as lead or principal engineer
About
Owner and principal consultant
Ajentx is led by Andrew Dobson, a software engineer with twenty-five years of commercial delivery behind him, much of it as a lead or principal engineer brought in to set up teams and the way they work.
Everything he brings to a build with AI comes from doing the work by hand first: where a service boundary should go, which test will break when the code changes, what a requirement has left out. That judgement is what the models are held to.
.NET and Azure by background, with the language and framework chosen for the job. Based in Lancashire and working with clients across the UK.
Contact
A first conversation is about your business and what the product has to do for it. You'll get a straight answer on what it would take.