Architecture and due diligence
Monolith or microservices, containers and deployment, and whether the codebase can carry what you want to do next. Written from building these systems, not from reviewing them from outside.
Independent technology judgement, from someone who has built and run it.
For founders, boards and executives making technology decisions they will have to live with — architecture, build or buy, AI, scale, and the teams that deliver it. We are not selling you a platform or a development team, which is why we can tell you when the answer is no.
What this practice covers
Six subjects, not a lifecycle. Most engagements are two or three of them at once, because that is how the decisions arrive.
AI & automation
Build or buy
The comparison done properly, including what the build costs to maintain after year one and who is left supporting it.
Part-time CTO support
Architecture, infrastructure, security, hiring and delivery decisions, at the point where they are made. Carried rather than advised on, in companies he founded and companies he was brought into.
Remote and distributed teams
Hybrid, fully remote, or half an office and half not. Twelve years running an engineering team across eight countries and five time zones.
Scale and infrastructure
What breaks first as traffic grows, and what it costs to fix then rather than now. Run by small teams at a few hundred thousand users a month.
Or keep one of us on hand
Most of this work is not a project. Simon works as a fractional CTO with a small number of clients at a time, over months or years — in the architecture decisions, the vendor conversations and the hiring, at the point where they are made rather than in a report afterwards. How much time, and how you want it used, is agreed with you.
Inside AI and automation
Where AI earns its place, and where it does not. For leaders who are being told to do something with AI and would rather do the right thing with it.
AI Opportunity Assessment
Which of your processes would genuinely benefit, which would not, and what it would take. Including the ones to leave alone.
Business Process Automation
Removing manual work by connecting systems that already exist, usually cheaper and more durable than adding intelligence.
AI Advisory
Build, buy or wait. Model choice, data boundaries, cost behaviour, and what happens when the vendor changes the terms.
Proof of Concept to Production
The gap most AI work dies in. Evaluation, failure modes, monitoring, and who owns it on a Tuesday.
System Integration
APIs, data flows, and the unglamorous plumbing that decides whether any of it reaches a real process.
AI Governance
What the model is allowed to do, what it must never do, and how anyone would know. Especially where finance data is involved.
Where this meets treasury
- AI for cash forecasting
- Reconciliation and exception handling
- Treasury reporting automation
- Where not to use AI in treasury
Understanding one side is common. Understanding both is not.
Who does this work
Simon Potter — Technology & AI
Simon has been a CTO for twenty years — co-founder and then CEO at Kwiziq, and now fractional CTO at SalesPond and academy+. He moved Kwiziq’s architecture from a monolith to API-based services while it served 400,000 monthly users on a small team, and has put generative models into production rather than into pilots.
Belinda Davis leads the work where automation meets treasury process and accounting. Belinda’s experience.
Related thinking
- Do we actually need AI for this?
- The AI pilot works and we cannot get it into production
- AI agents or ordinary automation, which do you need?
- We have a technology strategy and nobody uses it