AI consulting practice
We clean up the data mess with AI. Most businesses are sitting on more information than they can use: a dozen systems that were never meant to talk to each other, the same customer recorded three different ways, numbers nobody quite trusts. Think of an estate agency with tens of thousands of details spread across a property system, a mailbox, a spreadsheet and someone's notes. We sort that out, put AI to work on top of it, and keep you in control at every step. And we do not stop at advice. We build the systems and we run them.
What we do
Most businesses do not struggle with AI because the models are weak. They struggle because the information underneath is a mess: it sits in a dozen systems that were never meant to talk, the same customer is recorded three different ways, and half the numbers disagree with the other half. An estate agency with a few staff can easily be holding tens of thousands of details across systems that have never met. We clear that up, put AI to work where it plainly pays, and keep a person in charge of the outcome.
We start with the one job that gives you the quickest, clearest return, and prove it works before you spend on anything wider.
We give your information one trusted home, so AI works from facts your team agrees on rather than from scattered copies that disagree.
Everything we build runs under human review. You decide what happens; the system does the heavy lifting and shows its working.
What we do
A short list of jobs, each tied to something you will notice: hours back in the week, numbers that agree, work that stops falling through the cracks. We take on as much or as little as you need, from a first honest opinion to running the system day to day. All of it stays safe and under your control.
We look at how your business actually runs, say plainly where AI would help and where it would not, and give you a short plan you can act on.
Outcome: a costed shortlistWe hand the repetitive work to AI: the chasing, the re-keying, the checking one list against another. Your people get their week back for the parts that need a person.
Outcome: hours returnedWe pull your information into one trusted place, so the same customer is one customer and every system reads the same set of facts.
Outcome: one source of truthWe connect the software you already pay for so it works as one. No more copying things by hand between systems that were never meant to talk.
Outcome: fewer manual hand-offsWe show your people how it works and how to change it, so you are not dependent on us to keep it running. It stays safe and under your control.
Outcome: a team that can run itWe do not stop at advice. We build it, put it live, and where you want it, run it day to day so the benefit actually lands.
Outcome: systems in productionA simple first step
A fixed price for a fixed piece of work. We look at your business and tell you plainly where AI is worth the effort and where it is not. No open-ended discovery, no retainer, no pressure to buy anything else. It starts and it finishes.
We work out how the job actually gets done, find where the time and money leak away between your systems, and rank the options by what they are worth and what they would take. You end up with a written report and a costed shortlist you can act on, with us or without us.
Our approach
We look at how the job actually gets done before we talk about tools. Then we build only where it plainly pays, prove it on one job before you spend on anything wider, and keep a person in charge of anything that matters. Sometimes the honest answer is that you do not need AI here at all, and we will say so.
We learn how the job really gets done and where the time and money leak away. The technology comes after that, never before it.
We pick the one job with the most to gain for the least effort, and we tell you plainly where AI is not worth it. No building for the sake of it.
We give your information one trusted home. That is where AI actually pays off, and nothing built on top of a mess holds up.
Everything we put live has a name against it, a way to switch it off, and a person answerable for the parts that matter.
Going live is the start, not the finish. What it shows you tells you how to run the work better, and then we extend it to the next job.
People stay in control. The system handles the volume and the repetition; you set the rules, review the work, and hold the final say. That is the safe way to put AI into a business, and it is the only way we do it.
Selected work
These are four products of our own, live today and built principally on Anthropic's Claude models. Two are the AI environment and the training behind it. Two are working businesses in construction and estate agency, running on it every day. They are the evidence that we can make AI work on real, messy business data, not just talk about it.
The AI environment we install on a client's own server: Claude Code joined to a persistent knowledge graph that holds every person, document, conversation, and decision they have touched, with email, messaging, research, and document production wired in. It is the working version of the architecture we recommend, built by us and run by us.
Claude Code · knowledge graph · agent orchestration getmaxy.com → EducationA hands-on course, in London and online, that teaches knowledge workers to set Claude up around their own work without writing code. Twenty-eight modules across five stages, from the basics of the models through bringing your archives into a structured data layer to running the whole setup day to day. It is our applied Claude knowledge, written down and taught.
Claude fundamentals · data layer · applied practice maxy.institute → Deployment · ConstructionA back office for small builders and trades, built on our environment. Building work is a mess to keep track of: several jobs at once, quotes and approvals moving between people who each keep their own version of the truth. SiteDesk puts AI across all of it, under human control.
Small builders and trades · live deployment sitedesk.online → Deployment · Estate AgencyBuilt for UK estate agents, who sit on tens of thousands of details spread across systems that have never talked to each other. It handles the enquiries, the listings, the correspondence and the paperwork that moves a sale along. Same architecture as Maxy, shaped to how an agency actually works, and run day to day.
UK estate agents · live deployment realagent.network →Insights
Short, plain pieces on what people actually ask us: why connecting more systems together rarely fixes the mess, who answers when AI gets something wrong, and how to keep a grip on it so it protects your business rather than putting it at risk.
Each new connection between two systems is one more thing to maintain, and one more place for the numbers to disagree.
When every tool holds its own copy of the truth, your team spends its days checking one list against another. Wiring the tools together more tightly just moves the mess around.
The fix that lasts is one trusted place every system reads from. AI then works from facts your people already agree on, and the checking stops.
Read the view →Most teams are renting the layer they run on. The choice is whether to keep renting or to own the data and run the model directly on it.
AI agents arrive in waves, from a tool one person uses to AI woven through how the business runs. Knowing which wave you are in, and which one you are moving to, frames every other decision.
The durable position is to keep your data in a structure you own, a knowledge graph, and use the model directly on it. You gain the model's growing capacity with every release, and you never give away control.
Read the view →Using AI responsibly is less about principles on a wall and more about who answers when something goes wrong.
The questions that matter are simple. Whose system is this? What was it built on? Can we explain a decision to the person it affected? Can we show our working?
Done well, this is not a brake on progress. Knowing who is answerable, and keeping a record, is what lets you use AI with confidence instead of crossing your fingers.
Read the view →There are, and for most smaller businesses they are less frightening than the headlines suggest. The simplest approach is to hold yourself to the strictest rule that applies to you.
Europe's AI law sorts systems by how much harm they could do. The more a system could hurt someone, the more you have to be able to show about how it works.
Start with a simple list: which AI you use, what each one decides, and who is answerable for it. Most of the work is keeping notes and keeping an eye on things, not new technology.
Read the view →In nearly every one of these cases the AI did what it was asked. The harm came from an ordinary safeguard nobody had put in place.
Seven real cases, from a leak into ChatGPT to a coding agent that wiped a live database, come down to the same handful of missing safeguards. Not one is a new kind of risk invented by AI.
The fixes are unglamorous and they work: give the AI access to only what it needs, stop it short of anything it cannot undo, treat whatever it reads as untrusted, and keep a record of what it did.
Read the view →With the work, not the technology. The most impressive projects deliver the least, because they start with the clever thing and go looking for somewhere to put it.
Map a week of the job first, step by step, and the opportunities show up on their own: the stalls, the same details keyed in twice, skilled people on work that does not need them.
Then build only where an agent genuinely helps. Sometimes it does not, and going live is the start of the loop rather than the end of the job.
Read the view →About Rubytech
Rubytech is an AI consulting practice for smaller businesses. The problem we solve is nearly always the same one: information scattered across systems that were never meant to talk, the same customer recorded three different ways, and numbers nobody quite trusts. We clear that up and put AI to work on top of it.
We learned how to do that in some demanding places. Our experience runs across banking, insurance, healthcare, data and science, where a data mess gets very large very quickly. That is evidence, not a limit on who we work with. The skill that carries over is taking a tangle of systems and making AI work inside it safely, and it works just as well at ten people as at ten thousand.
What you get from that is judgement: which problem to solve first, which of your data can be trusted, and where AI earns its place rather than adding risk. We work as senior independent consultants, so you deal with the people doing the thinking, not a chain of hand-offs.
The proof is in what we have built. Our own products run on Anthropic's Claude models day to day, including two that run real businesses in construction and estate agency. The work above is the evidence under the capability, not a portfolio of slides.
We provide services in the United Kingdom and across the EEA.
Start a conversation
A first conversation costs you nothing but an hour. Tell us a little about your business and what is getting in the way, and we will tell you honestly whether AI is the right answer, and if it is, where to begin.