Find the value
Before any build, we find where AI, data, and automation would actually move your numbers, and the processes that should be fixed or killed rather than automated. We do not automate your mess faster.
AI automation & operations
We use AI, data, and automation to create measurable business value. Most AI spend goes nowhere because it starts with the tool, not the problem. We start by finding what is actually worth doing, then build and ship the systems that do it.
We find what is worth doing, then build it.
Before any build, we find where AI, data, and automation would actually move your numbers, and the processes that should be fixed or killed rather than automated. We do not automate your mess faster.
We design, build, and ship the systems: AI workflows, agents, data pipelines, integrations, full applications. Working operations that create value, not a slide deck.
Ongoing, embedded help running and expanding your AI and data operations over time. Judgment and hands, not just advice.
How we work
Most AI and automation fails for one reason: it is pointed at the wrong question. So before we write code, we run a short, structured pass to find the load-bearing assumption a project is resting on, and the thing actually worth building. This is the assumptology method, and it is why what we ship creates value instead of just running.
What are we actually trying to do, and what are we not?
What does the plan presuppose? Which assumption is load-bearing?
What breaks if that assumption fails? What would count against it?
What is actually worth building, once the premises are explicit?
A CTO who ships in serious places.
Behind Erotetic Arts is a career of building production systems where getting it wrong is expensive: data pipelines and warehouses on Redshift, DBT, and Snowflake with real testing, BI at scale, payments infrastructure from Open Banking to card acquiring, and systems run inside regulated finance. We build for production, not for a demo.
Recent independent builds
A custom AI knowledge base built for a six-figure legal case. Every email and OCR'd attachment is cross-referenced in one place instead of held in memory, and each new email is analysed as it lands, so the team argues from the full record rather than their recollection of it.
A SaaS product I designed and built end to end. It watches the things that quietly break, email configuration, certificates, and protocols, and is built for action: fixing what is failing, not just telling you it went down.
Repeatable migrations from ageing WordPress to fast, modern Astro apps with a git-based CMS. Content is versioned and separated from code, so every change is auditable by regulators or stakeholders, with better security, scalability, and speed.
A Dagster data pipeline and API around UK Companies House charges data, the security interests (mortgages and debentures) lenders register against a company's assets. It turns a slow public register into a clean, queryable feed for credit and due-diligence decisions.
Whether you want to automate a process, build an AI or data system, or work out what is actually worth doing, we would love to hear what you are working on.
Start a project →