Services
AI & Automation
AI wired into the work you already do, with the security questions answered first.
What you get
- Workflow audit identifying where AI actually pays back, and where it does not
- Custom AI assistants trained on your own documents, pricing and procedures
- Automated intake, quoting, scheduling, dispatch and follow-up
- Document and email processing pipelines with human review built in
- Internal tools and dashboards powered by AI over your own data
- AI security review — data handling, prompt injection, access control, vendor risk
Outcomes
- Hours per week returned to the people doing billable work
- Faster quote turnaround and fewer leads lost to slow response
- Consistent output quality that does not depend on who is on shift
- A clear, written answer to "where does our data go?" before anything ships
Capabilities
AI for trades and field services
Plumbers, electricians, HVAC and contractors: automated call and message intake, quote drafting from photos and notes, scheduling, and follow-up that stops leads going cold.
AI for retail and consumer businesses
Product content generation, inventory and demand analysis, customer support assistants, and review handling that sounds like your business rather than a chatbot.
AI for finance and investing
Research summarisation, report generation, data extraction from filings and statements, and analysis pipelines. Built by someone who has run production financial data systems.
AI for professional services
Document drafting against your own templates and precedent, intake triage, meeting-to-action-item pipelines, and knowledge retrieval across your archive.
Custom AI applications
When an off-the-shelf tool does not fit, we build the application: retrieval over your documents, agents that call your systems, and interfaces your team will actually use.
AI security and governance
Prompt injection defence, data residency and retention, access control, audit logging, and honest vendor risk assessment before you hand anyone your data.
Most AI projects fail for unglamorous reasons
Not because the model was not capable. Because nobody mapped the workflow first, nobody decided what happens when the output is wrong, and nobody asked where the data goes.
We start at the other end: what does your business actually do all day, which parts of it are repetitive and expensive, and what is the cost of a mistake in each one.
Where the returns actually are
Across the businesses we have worked with, the pattern is consistent. The wins are rarely the impressive demo. They are:
Intake. Every missed call, unanswered message and unreturned form is a lead someone else will get. Automated capture, triage and response is unglamorous and pays for itself fastest.
Quoting and estimating. Turning photos, notes and a rate card into a draft quote in minutes instead of evenings. The human still approves it. The evening comes back.
Document work. Extracting structured data from invoices, statements, filings and reports. Drafting against your own templates rather than generic output.
Research and reporting. Summarising, comparing and assembling the recurring report that currently eats a day a week.
The part most vendors skip
An AI system that can read your email and take actions is a new, privileged component inside your business. It deserves the same scrutiny as any other one.
We cover, in writing, before deployment:
- Data flow. What leaves your network, to whom, under what contract, retained for how long.
- Prompt injection. Content your AI reads is untrusted input. If it can trigger actions, that is an attack surface, and we design privilege separation accordingly.
- Access control. The assistant should see what the user is allowed to see, not everything in the index.
- Audit. A log of what was asked, what was returned, and what actions were taken.
- Failure behaviour. What the system does when it is unsure — which should be “ask a human”, not “guess confidently”.
This is where our cybersecurity practice and AI work meet, and it is the main reason clients come to us rather than to a pure automation shop.
Who is doing the work
Twenty-five-plus years of production engineering, including financial data platforms where numbers being quietly wrong is a serious event, and aerospace certification work where “it usually works” is not an acceptable standard.
AI is a powerful tool with a well-documented tendency to be confidently incorrect. It should be deployed by people whose professional instinct is to verify.
Frequently asked questions
We are a small trades business. Is AI realistic for us?
Often more so than for a large company, because the wins are concrete and the decision path is short. The highest-return starting points are usually intake and quoting: capturing every call and message, drafting the quote from photos and notes, and following up automatically. Those directly affect revenue rather than being an experiment.
Will our data be used to train someone else's model?
Not if it is set up correctly, and this is the first thing we settle. Business and API tiers of the major providers contractually exclude your data from training, and for genuinely sensitive material we can architect around self-hosted or on-premise models. You get the data flow written down before anything is connected.
What is prompt injection and should we care?
It is the attack where instructions hidden inside content your AI reads — an email, a PDF, a web page — get treated as commands. It matters enormously the moment your AI can take actions or read confidential data. We design for it explicitly: privilege separation, treating retrieved content as untrusted, and human confirmation on consequential actions.
How do you decide what to automate first?
We look for tasks that are high frequency, low judgement, and currently done by someone expensive. Then we check the failure cost. A task where a wrong answer is cheap and obvious is a good first candidate. A task where a wrong answer is expensive and invisible is not, regardless of how tempting it looks.
Do you build on OpenAI, Anthropic, or something else?
Whatever fits the constraint. Model choice is driven by the task, your data sensitivity, your latency budget and cost. We build so the model is a replaceable component rather than a dependency you cannot escape.
Related services
Software Engineering
Architecture, subscription platforms and data pipelines — built to be maintained.
Learn moreCybersecurity
Incident response, ransomware recovery, and hardening that holds up under audit.
Learn moreInfrastructure & Backup
Servers, clusters and backups designed so that failure is survivable and boring.
Learn moreTell us what is breaking — or what you are trying to build.
You get a senior engineer on the first call, not a salesperson. If we are not the right fit, we will say so and point you somewhere better.
Active incident? Write “URGENT” in your message and we prioritise it.