folio

Does Copilot in Excel work for anybody? Anecdotally, nobody I know finds it useful. Queries very rarely work. On the odd occasion it does, the answers are not dependable and may incorporate hallucinated (aka incorrect) data.

Colleagues are more enthusiastic about its potential in other applications like Outlook and Teams. Yippee! I’m not one of those people, though, but that’s probably more to do with the way I work than necessarily a shortcoming in the platform.

Still, Microsoft Copilot has a long way to go to be genuinely useful for most people. It’s still very much a novelty, only occasionally useful. On those times I’ve asked it to help with a report, I’ve ended up just doing it myself because it took too long.

Which just goes to show that even if you’re the third richest company in the world, you can still fail to deliver products that offer significant value for your users. Just ask the richest company in the world — Apple — how their foray into AI is going with their botched launch, promising features they’ve had to quickly withdraw because they just don’t work.

None of this is to say AI is useless. I’ve found some genuine use cases, such as reviewing code or helping to make content more accessible. For undertaking initial research, it’s incredibly helpful, although you do still need to verify results given the aforementioned hallucination problems.

But those thinking AI will deliver massive cost-savings and profits are beginning to learn that’s not necessarily the case. Studios are already shunning generative video, realising its promise is much overhyped. And anyone working with critical data is realising that generative AI cannot be trusted for safety.

This could be bad news for governments planning to replace civil servants with technology. At best, in its current form, it can augment roles. Certainly, I see it augmenting my own role; well, it will have to, for the centre is set on shrinking corporate administrative services across the board.

I think it’s better to start thinking of AI as an accessibility aid than a panacea to all of our problems. AI is undoubtedly impressive, but the hype often oversells its capabilities.

The initial excitement around new technology often creates a rush to adopt everything available, but as the market matures, being selective becomes important. Not every new AI feature or platform will meaningfully improve our work or creative processes.

Focusing on the tools that consistently deliver practical benefits while being more cautious about investing in promising but unproven capabilities will help us find a better balance.

But, alas, balance is severely lacking in the decision-making at the top of nations, institutions, and corporations, where those in leadership positions lack genuine technical expertise. The risk is that their imbalance will ultimately sink the ship.