Don't overlook oversight

Whether you have started to dabble with ChatGPT at home or you're using Microsoft Copilot professionally, a key differentiator starts to emerge: AI fluency. The way we use AI and the quality of the results we get vary depending on how AI-fluent we are.

And that fluency matters more than ever, because we're entering a world where AI-generated text is becoming indistinguishable from something written from scratch.

There are tools that rewrite content specifically to evade plagiarism detectors. There are apps that sit silently beside candidates in remote interviews and feed them perfect answers in real time. Deception is becoming a service that you can purchase via subscription.

Would you use AI to apologise? Would you use it to console? Would you use it to deliver difficult news? And would the person receiving it feel differently if they knew?

In funeral service, these questions land harder than in most other sectors.

It goes without saying that AI can hallucinate and produce confident nonsense. That risk is shrinking, but not gone, and guidance must ensure staff remain responsible for whatever AI helps them create.

The pilot still needs stronger judgement than the copilot.

As you become more AI-fluent, you begin to see the pattern-based nature of the system. You show it the structure of a routine email, and it will recreate it flawlessly with different inputs. You hand it a form, and it will produce something tidy, formatted and almost uncomfortably "you".

These instructions can be passed to other team members in the form of a pre-configured chat. Depending on what platform you use, these have different names. Copilot = Agents. ChatGPT = GPTs. Gemini = Gems.

AI amplifies productivity: you can hand AI instructions to other members of staff to generate the same structure each time with different outputs. The temptation is to use AI for everything all the time, and that would be a mistake.

Human oversight is needed to ensure quality, consistency and authenticity. And human oversight is needed in terms of governance. That means setting and enforcing appropriate policies. Without direction, people fill silence with their own solutions. And in a world where AI tools are increasingly engineered for deception, the responsible route is unlikely to be simply banning things.

Ban AI outright and you create the perfect environment for shadow usage. Worse, you guarantee losing sight of the one thing you're ultimately responsible for. Your data.

A practical fix? Supply your team with secure, approved tools. ChatGPT for Business. Microsoft 365 Copilot. Google Workspace's built-in AI. Each of these gives you far more control over where information goes and how it's stored. More importantly, they give staff a clear, safe place to work, rather than mixing professional queries with personal AI apps on their phones.

And rather than unleashing AI usage without guardrails or guidance, have a conversation with your team. Decide when transparency is necessary through the lens of your customer or recipient. In some contexts, a simple disclosure is enough, such as "This was produced with the help of AI". It nudges the reader to apply some level of scrutiny to facts and accuracy, as we all know AI can get things wrong.