Gen AI Innovation
Last week we hosted another AI Productivity Workshop for the ActionCoach team. The room was packed, but it was the "aha" moments at the end that stood out most:
"I never knew there were so many AI tools out there."
"I'm amazed how much better some of these are than the paid version I've used for years."
"Now I have a team of AI assistants for my daily tasks."
The biggest takeaway of the day, in one line: the "best" AI doesn't exist. Only the best AI for the task in front of you.
Why Tool Selection Matters More Than People Think
Selecting the right large language model is the difference between a tool that "sort of works" and one that actually transforms your workflow. Most people pick one tool early on, stick with it out of habit, and never revisit that choice — which means they're often using a perfectly good hammer on a job that needed a screwdriver.
Here's how I'd map the current landscape.
The Strengths — Why Use Them
Different models are genuinely built for different strengths. Some are specialised for nuanced creative writing, others remain a powerhouse for logic-heavy tasks, and some are the clear choice when you need real-time information with citations attached.
Some models are also built around "chain of thought" processing — reasoning step by step through a problem — which suits engineering or complex maths far better than a model optimised for quick, fluent answers.
And it's not always about picking the most powerful option available. Smaller, leaner, often open-source models are frequently faster and cheaper for high-volume, simple tasks where you don't need the heaviest tool in the box.
The Reality Check — What to Watch For
Every tool comes with limits worth knowing before you rely on it.
Knowledge gaps. Always check the cutoff date. An AI is only as current as its last training session — ask it about something recent enough, and it simply won't know.
Hallucinations. You are the expert. Check the output. Even the best models can sound completely confident while being completely wrong — confidence in tone is never proof of accuracy.
Context window. Some models start "forgetting" the beginning of a long document faster than others. If you're working with a lengthy brief or a big dataset, this matters more than most people realise.
The Bottom Line
Don't marry a single tool. Build a toolbox.
The ActionCoach group's reaction says it all — most people have no idea how many options exist, or how differently those options perform on the exact same task. Once you start matching the tool to the job instead of defaulting to whatever you opened first, the quality gap becomes obvious fast.