AI Engineering
The engine, not the act.
AI as a marketing layer is everywhere. AI as engineering — the discipline of using it well — is rare. We sit on the engineering side. Context engineering, second brains, and AI as an enabler of human knowledge and experience: that's what we build, what we use internally every day, and what we offer to clients.
Our position: AI is an enabler, not a replacement.
AI doesn't replace the engineer, the consultant, or the domain expert. It amplifies them. The work that gets transformed is the cost and speed of getting from problem to first answer; what doesn't change is the judgement about whether the answer is right. The teams getting the most out of AI are the ones who treat it as a tool that augments human knowledge — not a system that runs without one.
Our practice is built around that principle. Every workflow we engineer keeps a human in the loop where judgement matters and frees them up where it doesn't. The output is better than either alone could produce.
Foundation discipline
Context engineering.
Without context, AI is a coin-flip — sometimes brilliant, often wrong, never reliable. Context engineering is the discipline of structuring AI inputs — prompts, memory, examples, tools, evaluation — so outputs are repeatable, traceable, and useful in production.
It's not a single technique. It's a stack of practices: durable memory systems, structured outputs that can be validated programmatically, tool integrations that let AI act on real systems, evaluation harnesses that catch regressions before they reach customers, and prompts that survive a model upgrade. Done well, it's the difference between an AI demo and an AI tool.
- → Memory systems that retain the right context across conversations
- → Structured outputs — JSON schemas, validation, retries — so AI can be relied on inside software
- → Tool integrations via Model Context Protocol (MCP) — connecting AI to live business systems
- → Evaluation harnesses — catching quality regressions before customers do
- → Prompt and capability versioning — surviving model upgrades without breakage
Knowledge as infrastructure
Second brains.
Most businesses have knowledge trapped in people's heads, scattered across email threads, and undocumented in workflows that "just work" because someone remembers them. When that person leaves, goes on holiday, or simply forgets — the knowledge goes with them.
A second brain is structured knowledge infrastructure your team and your AI tools can both query. It captures the decisions, processes, customer history, technical detail, and reasoning that make your business work. The result: faster onboarding, less re-discovery, decisions backed by what the organisation already knows, and AI tools that operate with real context — not generic answers.
We build these for our own operations and for our clients. The mechanism is straightforward; the discipline of doing it consistently is where the value lives.
Dogfooded
What we build for ourselves, we build for you.
The AI claims we make on this site are backed by tools we actually use, not slides we present. A handful of the artefacts running our business right now:
- → An internal knowledge assistant with persistent memory that triages inbound work, drafts proposals, and routes client requests to the right project folder.
- → The chatbot widget being built for our customer site-build pipeline — turning a conversation into a structured brief that drives static site generation.
- → Document automation pipelines — Statement-of-Work generation, project hand-off documents, status updates — that turn meeting transcripts and decisions into deliverables.
- → MCP integrations into Xero, calendars, and our project filing system — so AI tools have current, accurate context every time we use them.
None of these are demos. They're production tools, used daily, evolving as our customers and our practice change.
What this looks like in your business.
AI engineering isn't a category to "deploy AI" in. It's a way of identifying where AI changes the economics of an existing process — and engineering around that change so the gain is captured reliably.
Customer enquiry handling
Turning inbound emails, forms, and calls into structured tickets with the right context attached — before a human gets involved.
Document and proposal generation
From scoping notes to a draft SoW, retainer outline, or quote — minutes instead of hours, ready for a human to refine.
Internal knowledge query
A second brain over your business — process docs, customer history, technical decisions — that your team can search in plain English.
Decision support
Surfacing the relevant information at decision points — pricing, risk, vendor selection — so the human call is faster and better informed.
The serious part
Data governance. The minefield we take seriously.
This is the question most AI providers don't answer well: where does your data go, who can see it, is it used to train models, how long is it retained, and where does it physically live? For any business handling customer data, regulated data, or trade secrets, these aren't side-considerations — they're the difference between AI adoption and AI exposure.
We treat data governance as a core part of every assessment and every engagement. Specifically:
- → Provider terms — what gets logged, what gets retained, training-data opt-outs. Anthropic's API doesn't train on your inputs by default; not every provider can say the same.
- → Data residency — where the model physically runs, where audit logs live, what crosses borders.
- → Context minimisation — sending only what the task needs, never the whole database. Smaller context, smaller exposure.
- → PII handling — redaction, tokenisation, structured-output validation. Personal data leaves only where it has to, and never by accident.
- → Audit trails — logging what was asked, what was returned, who saw it. The basis of any retrospective investigation.
- → Sector-specific considerations — UK GDPR, financial services, healthcare data, public-sector procurement. The compliance shape changes by industry; we tailor accordingly.
We're upfront about scope. For lightly-regulated SME work, modern AI providers are well-suited and we'll engineer accordingly. For highly-regulated data — financial services, healthcare, legal — we'll tell you where the fit is strong and where you'd be better off elsewhere, and we'll point you there if needed.
Get started
AI Day Assessment.
£500
Fixed fee. One day.
A focused, one-day deep-dive into where AI fits in your business — and where it doesn't. We sit with you (on-site or remote), examine specific processes or pain points, and produce a written report.
What you get
- → Written report with prioritised opportunities
- → Honest "don't bother" call-outs where AI won't help
- → First-step recommendations with rough cost ranges
- → Context engineering specifics for any opportunity worth pursuing
- → Data governance posture review — provider terms, residency, PII handling, sector specifics
What it isn't
- → A sales pitch for a follow-on engagement
- → A generic "AI strategy" deck
- → A demo of what's possible — it's specific to your business
- → A commitment to more work — you walk away with the report
- → A substitute for legal advice on regulated data
Looking for the broader picture?
AI engineering sits alongside our IT services, Forge bespoke builds, and product portfolio. Tell us what you're trying to solve — we'll point you at the right starting place.
Get in touch