NEWFollow the cyber threat vectors into the dark. Register for Jedi Declassified Threat Briefing >Register Now
Back to the Jedi Archives
AI 5 min read

Are You Putting Too Much Trust in AI?

AI tools save time, but a confident answer isn't always a correct one. Learn where AI needs a human check and what your team should never paste into it.

Are You Putting Too Much Trust in AI?

Think back to when tools like ChatGPT first appeared. Most of us treated every answer with healthy suspicion, reading it carefully and checking whether it had got things right.

A couple of years on, how many people in your business are still checking quite so carefully?

It's an understandable shift. When AI has helped you hundreds of times, you start to trust it a little more. Then you hand it more important jobs because it did so well with the easy ones. That's exactly the point where a bit more caution pays off.

Confident isn't the same as correct

Imagine someone in your accounts team gives an AI tool a spreadsheet and asks it to analyse the figures. It comes back with a tidy summary and highlights an interesting trend. It looks sensible, so it goes straight into the monthly report.

But how did the AI reach that conclusion? Did it read every row? Did it misunderstand a column? Did it simply fill a gap with a guess?

AI is already being used for important work in small businesses: financial information, expenses, payments and reporting. It's useful because it can get through a lot of information very quickly. But it can get things wrong, and when it does, it's often very convincing.

A bad answer doesn't always look like a bad answer. It can arrive beautifully written, neatly formatted and completely confident. If nobody checks it, that answer can end up in a board report, a client proposal or a pricing decision.

What is your team putting into AI tools?

There's a second question worth asking: what information is being fed into these tools?

AI use usually starts with harmless jobs, like tidying up an email or summarising a document. Once people get comfortable, they start uploading more: a spreadsheet of customer details, a contract, last quarter's figures.

At that point you need to know which AI tools your staff are using, whether they're on free personal accounts or proper business ones, and what happens to the information they upload. If personal data is involved, that also has implications under UK GDPR.

There's a big difference between a business AI tool that works inside your existing Microsoft 365 security and data protection, and someone pasting client information into a free consumer chatbot.

Decide where AI helps and where a human checks

You don't need to ban AI. You need a sensible view of where it can help freely and where someone with the right knowledge has to check the work.

  • Low risk: drafting an internal email or meeting notes that someone reads and tweaks before sending.
  • Medium risk: summarising documents or research where accuracy matters, so check the key points against the source.
  • High risk: anything that affects your finances, customers, contracts or a major decision. A knowledgeable person should review it before anyone acts on it.

A short AI usage policy that lists your approved tools, what information must never be uploaded, and when AI output needs a second pair of eyes goes a long way. It doesn't need to be complicated. One page is often enough.

Let AI save you time, not make your decisions

Let AI take the repetitive work away. Let your team experiment with better ways of working. Just don't get so comfortable that nobody stops to ask, "Is this right?"

Not sure how much AI is being trusted in your business?

At Jedi Services, we help UK businesses use AI safely, from choosing the right tools to putting a simple AI policy in place and making sure your data stays protected. If you'd like a quick review of how AI is being used across your team, get in touch.