If you've attended a conference recently, read a business publication, or had a conversation with someone who's just discovered ChatGPT, you'll have heard some version of "AI is going to change everything."
That's technically true, in the same way that "the internet is going to change everything" was technically true in 1996. Yes, and also: most of the predictions were wrong, most of the hype didn't materialise as described, and the businesses that actually benefited were the ones that approached it pragmatically rather than evangelically.
Here's the honest version.
What AI is genuinely good at right now
Pattern recognition in large datasets. If you have a lot of data and want to find patterns, anomalies, or predictions — AI does this exceptionally well. Better than any human, faster, and cheaper.
Text generation and summarisation. AI writes competent first drafts, summarises long documents, and generates structured content from unstructured inputs. Not perfect. Good enough to save significant time.
Classification and routing. Sorting emails, categorising support tickets, routing enquiries to the right department — AI does this reliably and at scale.
Transcription and extraction. Meeting notes, document analysis, extracting structured data from unstructured text — accurate and fast.
Repetitive task automation. Anything that follows a rule, happens on a schedule, or involves processing the same type of input repeatedly is a candidate for AI automation.
What AI is not good at
Original creative judgment. AI can generate ideas. It cannot evaluate whether they're genuinely good ideas for your specific context. That's still a human job.
Navigating ambiguity in high-stakes situations. AI is confident even when it's wrong. In situations where being wrong has serious consequences, you need human judgment in the loop.
Deep domain expertise that isn't well-represented in training data. If your business operates in a niche that isn't well-documented online, AI models will have a blind spot.
Understanding your specific business context. AI doesn't know your customers, your team dynamics, your history, or your strategy. Every AI implementation requires investment in giving the system the context it needs.
The common mistakes
Buying tools before defining problems. Half the businesses I talk to have subscriptions to three AI tools they're not really using. The tools aren't the problem — the lack of a specific use case is.
Expecting AI to work without configuration. Out-of-the-box AI tools are general-purpose. Making them useful for your specific business requires setup, testing, and refinement. This work isn't free.
Underestimating the change management. The technology is usually the easy part. Getting your team to actually change how they work is harder and more important.
Trying to do everything at once. The businesses seeing the most value from AI have picked one or two specific use cases and implemented them properly, rather than trying to "AI-transform" the whole business simultaneously.
A practical starting point
If I were advising you right now, I'd say this: pick your biggest, most painful, most repetitive process — the thing your team hates doing because it's just mechanical work — and ask whether AI could handle it.
Usually it can. Start there. Implement it properly. Measure the time saved. Then look for the next opportunity.
That approach — specific, measured, iterative — is what delivers real value. Not a strategy that sits in a slide deck.
If you want to work out what that first use case is for your business, the AI diagnostic on my homepage is a good starting point. Or book a call and we'll figure it out together.