https://markgerrard.ai/ Essays on running AI in production. en-gb https://markgerrard.ai/writing/reliability-is-a-budget/ https://markgerrard.ai/writing/reliability-is-a-budget/ Reliability Sat, 14 Feb 2026 00:00:00 GMT The interesting failures are never the model. They are the seams between systems — the retry that wasn’t idempotent, the timeout nobody set, the row that two processes wrote at once. https://markgerrard.ai/writing/verification-before-autonomy/ https://markgerrard.ai/writing/verification-before-autonomy/ Agents Fri, 09 Jan 2026 00:00:00 GMT Autonomy is not a slider you turn up. It is something you earn, one verified decision type at a time. https://markgerrard.ai/writing/the-crm-is-the-hard-part/ https://markgerrard.ai/writing/the-crm-is-the-hard-part/ Integration Sat, 22 Nov 2025 00:00:00 GMT Calling a model is a weekend project. Letting it write to the system that runs your business is a year of unglamorous work, and that is where the value actually lives. https://markgerrard.ai/writing/orchestration-without-the-orchestra/ https://markgerrard.ai/writing/orchestration-without-the-orchestra/ Orchestration Fri, 03 Oct 2025 00:00:00 GMT Most “orchestration” I see is elaborate routing that exists to justify itself. The useful version is smaller and more boring. https://markgerrard.ai/writing/planning-domain-retrieval/ https://markgerrard.ai/writing/planning-domain-retrieval/ Retrieval Sun, 17 Aug 2025 00:00:00 GMT We spent months tuning embeddings before realising the problem was that our documents were the wrong shape. https://markgerrard.ai/writing/logging-the-prompt-that-shipped/ https://markgerrard.ai/writing/logging-the-prompt-that-shipped/ Observability Mon, 30 Jun 2025 00:00:00 GMT The single highest-leverage thing we did for reliability was unglamorous: record the exact, fully-resolved prompt that was sent, every time.