We help your teams use AI to automate dull, repetitive work
We work alongside your people and teach them to do it themselves.
Our services, west to east
Select a summit to open its field notes.
-
Corstorphine Hill 162 m
We work out where AI is worth using in your organisation and what to do first.
- A readiness assessment covering data, systems, skills and current AI use
- A phased rollout plan with owners and costs for each stage
- Measures of progress agreed before the work starts
-
Craiglockhart Hills 176 m
We run hands-on workshops that show your engineers how to use AI tools at each stage of the software development lifecycle.
- Setting up agentic coding tools and using them in daily work
- Prompt patterns and workflows the team can reuse
- Reviewing and testing code that AI has written
-
Castle Rock 131 m
We set out how staff may use AI with your code and data, then configure the tools to match.
- Secure settings for each AI tool your teams use
- Rules on which data can go into which tool
- Review gates before AI-written changes reach production
-
Braid Hills 213 m
Attackers using models a step behind the frontier labs can find vulnerabilities faster than you can patch them. Clear-box testing finds them first.
- Frontier models and security tools testing with full access to your code
- Confirmed vulnerabilities fixed by severity, as changes your team reviews
- The same checks automated in your CI/CD pipeline
-
Blackford Hill 164 m
We index your organisation’s documents into a knowledge base, so staff can ask questions and check the source of each answer.
- Data curation that can be audited and verified
- LLM classification of content, measured against agreed success criteria
- Retrieval-augmented generation over your own documents
-
Calton Hill 103 m
We build custom software for the repetitive tasks your staff do by hand, and show your teams how to build their own.
- Tools built around the processes you already run
- Agents and MCP servers that connect to your existing systems
- Your own team trained to build the next tool themselves
-
Arthur’s Seat 251 m
We match the approach to the problem, from classical machine learning to large language models, and build it to run reliably.
- Classical machine learning for tabular, image and time-series problems
- Fine-tuned models when a general LLM only gets most of the way
- Tested on your own data, then deployed and handed over
Our view of AI
You pay people for their judgement, but much of their week goes on the admin it takes to get there: copying figures between systems, chasing documents, rebuilding the same report each month.
Software could always do that work. It cost more to write than it saved, so it stayed manual. With AI, custom software is quick and cheap to build, and those tasks are now worth automating. That gives people back time for the work that needs their judgement, what you pay them for.
The AI itself is a small part of this change. In the short to medium term, most of the gains will come from custom software written by AI, with AI built in at a few key points. Most automation follows fixed rules, so it belongs in ordinary code that does the same thing every time, rather than in an AI model or agent. That code runs on computers you already pay for, while a frontier model charges for every token. Tools beat tokens.
Project managers call the old trade-off the iron triangle: good, fast or cheap, pick two. Now we get all three.
The iron triangle
- Good and fast, so not cheap. Experienced engineers build it properly and quickly. The bill is more than the admin was ever worth, so the work is never approved.
- Fast and cheap, so not good. A macro or a quick script does the job until something changes. Then it breaks, and nobody knows how it works.
- Good and cheap, so not fast. Someone builds it carefully between other priorities. By the time it is finished, the process has moved on.
- Built with AI: good, fast and cheap. AI writes much of the code, and engineers review and test it. A well-built tool now takes far less time and money than it did, so small jobs are worth automating.
Engineers from your own teams
When the next feature or system is just a prompt away, the best people to build it already work for you: they know how the business runs and how its parts fit together. We help you find them, then train them to build software their colleagues can rely on.
About Seven Hills
Seven Hills is an AI transformation consultancy in Edinburgh, Scotland, named after the seven hills the city is built on. We work remotely and in person with clients across the UK and internationally.
We have more than 15 years of experience building machine learning and AI systems at large technology companies, and more than 10 patent applications in ML, information security and AI.
Talk to us
Tell us which tasks are taking up your team’s week. The first call is free and lasts 30 minutes.