Andrew CheesemanSpeciality
Everyone wants AI.Your estate is not ready.
The model is the easy part. The hard part is the twelve-year-old .NET service nobody wants to touch, the cloud account that was clicked together and never written down, and customer data spread across four systems with no clear owner. I move the estate somewhere it can be changed safely, get the data into shape, and then build the first AI feature that earns its keep.
Which of these is you
The platform nobody wants to touch
.NET Framework on Windows servers, a hosting contract that renews every year, or an AWS or Azure account built by hand. Every change carries risk, so nothing changes, and the AI plan waits behind it.
Moved in waves, each one held in Terraform, deployed through a pipeline and with a rollback plan. Lift what should be lifted, modernise what pays back, retire the rest.
AI is on the roadmap. The data is not.
Customer and operational data lives in a CRM, a database, three spreadsheets and somebody’s inbox. Nobody can say which copy is right, or what an AI tool would be allowed to see.
The data mapped, a system of record chosen, access drawn on purpose, and the useful parts made retrievable, with data protection handled from the start rather than bolted on.
Your developers already use AI, without rules.
Copilot, ChatGPT and Claude run daily, the licences are scattered, code review has not caught up, and nobody can say whether it is making the team faster or just busier.
The tools chosen, the guardrails written down, AI review wired into CI, and a measured pilot with one team, so you know what changed and what it cost.
What that looks like
- An estate and dependency inventory that tells you what you actually run, what it costs and what depends on what
- Migration in waves you can stop between, each one paying for itself, rather than an eighteen-month rewrite
- Old .NET brought forward to a current version, on a platform held entirely in Terraform
- Data readiness for AI: a system of record, an access model, and retrieval whose quality is measured rather than assumed
- AI-assisted engineering with the rules written into each repository, and review in the pipeline rather than on trust
Being straight about the AI half
The migration half is thirty years of shipped work. The AI half is newer, and I would rather say so than have you find out. What I bring is the platform, the data and the engineering practice that AI depends on, plus daily hands-on use of AI tools on a live production codebase. I badge machine learning as the newest line on my services page. If you need someone to design models, hire that person; I will make sure your estate is ready for them. And if you want a strategy deck about transformation, I am the wrong call.
Where this comes from
Three platform moves, each with a business that could not stop while it happened, and a small studio taken into AI-assisted engineering on a live Fortune 500 platform.
- Benchmark Capital: designed and ran the move off RackSpace onto Azure, provisioned the CI/CD pipeline, and took the systems from a monolith to event-based services, carrying the Lloyds Bank and Schroders joint venture and £10bn+ in managed assets
- ETZ Technologies: moved the platform onto Azure and adopted SOA, NServiceBus and a microservices design as the load and the team grew
- Graffic Jam: a multi-account AWS estate held entirely in Terraform, then the engineering taken from code completion to agentic tools working across 28 repositories, with the conventions written down in each one
Who this is for
Companies without a CTO
A managing director or operations lead who knows the platform is a risk and wants a straight answer on what to do about it, in what order, and for how much.
Scale-ups outgrowing version one
A head of engineering whose team ships features but has no time to fix the foundations underneath them. I take the foundations, and leave the team able to keep them.
Teams told to do AI
The board has asked for an AI plan, and the tech lead needs the estate and the data ready before anyone promises a feature. I do that part.
The way in
One day. £1,500. A written answer.
I spend a day inside your systems and write back what I found, what I would change, and what it would cost. You keep the document whether or not we ever work together again.
After the review
Three fixed prices, then a retainer if you want one.
Migration & AI Readiness Assessment
£3,500
Fixed, ex VAT · two weeks
Estate and dependency inventory, migration waves in order, data readiness for AI, guardrails for AI tools in your team, and a costed 90-day roadmap you can run with or without me. Half the fee is credited against follow-on work booked within 90 days.
AI Adoption Pilot
from £7,500
Fixed, ex VAT · three to four weeks
One real use case built into one of your applications, with retrieval quality measured, costs tracked and guardrails written down. Or AI-assisted engineering rolled out to one team, with delivery measured before and after.
Migration Wave
from £9,500
Fixed, ex VAT · quoted from the assessment
One defined set of services moved to its new home, held in Terraform, deployed through a pipeline, with a tested rollback plan and documentation your team can run.
Fixed-price work is paid in three stages. For a longer programme, the Fractional Chief Architect retainer starts at £2,850 a month for three days.
Tell me where you are stuck.
A short conversation costs nothing and usually gets further than a written brief. Tell me what is in the way and I will tell you what I would do about it.