Do you need a custom AI build? Usually you already have the tool

One of the top hotels in the UK came to me expecting to need an outside partner, new tools and a sizeable automation project. It turned out they already had Microsoft Copilot across the whole business, and several staff were using it. Most of what they wanted was within reach of what they were paying for. I showed them how to get it out of Copilot, and the few things Copilot couldn’t do became a short, clearly scoped list instead of a big project.

That pattern comes up all the time. A painful problem gets assumed to need a complicated fix, but the size of the problem doesn’t set the size of the solution. When I look at an operational bottleneck, it usually falls into one of three tiers.

Training. You already pay for the tool that fixes it and nobody has shown the team how to use it. Writing emails, drafting reports, pulling data out of standard documents: if you have Copilot, ChatGPT, Claude or Gemini, you have the engine already. What’s missing is a template, a clear prompt and a short training session. No developer needed.

Integration. Two systems won’t talk to each other. Your CRM holds the client data, your back office needs it for billing, and someone spends Friday afternoon copying between them. Automation platforms like Zapier, Make or n8n connect them. That’s configuration rather than code, and usually a few days to build and test.

Custom build. A process that reads unstructured documents, applies compliance logic and makes judgement calls before a person reviews the result. No off-the-shelf tool does that safely, so it needs proper engineering, API connections and strict guardrails, and it takes weeks.

The consulting industry’s incentives point the other way. A big custom project pays for an agency’s overhead and a prompt doesn’t, so the build is what gets sold.

I start at the bottom. If a prompt works, we use the prompt. If the workflow spans several systems, we build an integration. Only when that fails do we look at a custom build.

That saves money, but the bigger benefit is what it does for the team. When a problem gets solved with tools they already have, they stop seeing AI as a black box built by expensive outsiders. A paraplanner who learns to format a report with AI will work out the meeting agendas next, without being asked.

So check what you already pay for first. It nearly always covers more than anyone expected.