Decades of building software that has to keep working
We have spent decades building and looking after software for organisations ranging from start-ups to the NHS. AI has changed how quickly things get built, but not what it takes to keep them running, which is someone who understands the whole system and answers for it. Keep using AI. Just don't deploy it unchecked.
working != safe
Most AI-built software only ever gets the first of these checks, and a Readiness Review covers the rest.
“If you think it's expensive to hire a professional to do the job, wait until you hire an amateur.”
Who we help
You've built it yourself
That is a good thing, and we would encourage more of it. Getting it working is usually the easy part, though, so before it grows it is worth knowing who can read your customers' data, what it will cost to run, and what happens when it gets busy.
You're rolling out Copilot
That includes the kind of spreadsheet someone shared with "everyone" years ago and has since forgotten about. We check permissions, identity, labels and policy before Copilot goes looking, and leave you with a plan your own IT people can follow.
Your figures don't agree
That is usually because there is nothing solid underneath for it to work from. We build the data model, the agreed definitions and the components your AI tools use, so the figures stay put from one day to the next and you can trace each one back to where it came from.
If you run an agency that has lost senior people, you can borrow ours by the hour for architecture, infrastructure, security and code review, working under your brand.
What we found
Suppose a team builds an event booking plugin with an AI assistant. It takes bookings, sends the confirmations and looks right, and nobody has done anything wrong.
Given five days with it, we would quite likely find that any logged-in user can download every booking, that personal data is kept forever, and that the busiest page runs a separate database query for every attendee. None of this shows up when you click around, and all of it can be fixed in a few days once someone has spotted it.
How the review worksIllustrative example. A made-up report, not client data.
Production Readiness Review
| Area | Rating | What we found |
|---|---|---|
| Security | Red | The export endpoint checks that a request came from the site but not who sent it. Any logged-in subscriber can download every booking. |
| Personal data | Red | Names, emails and phone numbers are kept forever in a custom table, with no way to delete one person on request. |
| Performance and server load | Amber | The events page runs one database query per attendee. That is 40 queries today, and around 400 at the size you are planning. |
| Hosting cost | Amber | This is fine on current traffic, but the planned campaign would need a larger hosting plan unless the query problem is fixed first. |
The full report covers two more areas.
Why you still need people
We are not trying to compete with ChatGPT, Claude or Copilot; we use them ourselves. What we offer is the judgement around them, which a chat window cannot provide.
On our own projects, AI writes a good deal of the code, but it never has the last word: one of us reads every change, and a person puts it live. We would suggest working the same way, and we are happy to help you set that up.
Every review and every change we approve has a name on it, and if something is wrong you can pick up the phone and speak to that person.
We look at where your software runs, what it costs, what data it touches and who else relies on it, none of which can be seen from inside a chat window.
Software has to keep working, and keep being understood, long after whoever built it has moved on, so we write down why decisions were made and nobody is left guessing in two years' time.
We have spent decades building software for start-ups, small firms and large organisations including the NHS, Drake International and Wipro. The team also has formal AI training and years of experience running large language models (the technology behind tools like ChatGPT) in production.
What clients say
“Over the years Chris and the team at Elan have created and maintained a bespoke software system for PMA, shaping it to our changing needs and working closely and collaboratively with us as our business has evolved. We like working with nice people. They are nice people. So it's been a very happy and collaborative relationship.”
The people
Technical Director and founder
Chris leads our technical direction, from system design to the hosting, security and support that keep a system running for years.
Senior engineer and AI specialist
Calum leads our AI work with clients, from prompt design and evaluation to AI systems that learn from people's feedback.
When a job needs something we do not do every day, such as penetration testing, networks or incident response, we bring in specialists we have worked with and trust.
Articles
We no longer build one-off dashboards, but there is one kind we still design by hand: the dashboard inside software that collects its own data. The reasons it can be trusted are the same foundations we build for AI.
A chat tool can draw a dashboard in seconds. For most reporting, we now build the foundations that let AI produce correct, consistent dashboards every time, and that improve the more they are used.
You built a WordPress plugin with an AI assistant and it does what you wanted. These are the checks I would run before leaving it on a live site.
Production Readiness Review
The price is fixed before we start, from £1,500. You will receive a short report with a red, amber and green summary, and the ten fixes that matter most, in the order we would tackle them.