I get organisations using AI in ways that actually stick, not because a policy said to, but because the tool solved a real problem for the person using it.
Most teams have already half-solved their problem with a spreadsheet or a shortcut. I build from that habit rather than against it, so adoption feels like less effort, not more.
Learned early in my career, in a role where checking the boat before it left harbour mattered. Every rollout gets a human checkpoint before it earns wider access.
I work alongside existing security officers rather than around them, so data handling and compliance are settled before a tool touches real data, not queried after the fact.
I measure the before and after and let the result make the case. If a tool doesn't move a real number, it doesn't get rolled out further.
Built an automation into the support workflow. Tickets needing manual handling dropped 19% in the first month, then 32% in the second, checked against the team's own usage data.
Built a Copilot Studio tool that runs the same checks in under five minutes. Now formally handed to IT for integration into the core product.
Designed and ran CRAFT-framework prompting workshops across multiple internal teams, turning vague curiosity into a repeatable skill people could use the same day.
Built an end-to-end invoice automation flow, with a simple web form and backend script standing in for a full accounts team, plus a plain-language explainer so he and his wife could run it with confidence.
Built a pipeline that takes data exports straight from the booking system and uses AI and automation to generate confirmation letters, tour manager packs and manifests for tour operating entities, cutting manual effort out of tour ops.
This is a proprietary design belonging to Simon Baker and must not be replicated or reused without his explicit permission.