Get the exclusive new guardrail tool that reduces Microsoft Excel calculation errors

This is obviously a fake headline, right? Can you imagine if that was a real headline?

Can you imagine if the headline was: "Microsoft announces new guardrail tool to reduce Excel calculation errors"

How did you react?

Reduce Excel calculation errors?

Did you say reduce?

As in… there are STILL going to be errors? In Excel? Your premium tool just helps to reduce errors?

Ok. So apologies Microsoft and please, Microsoft — and you, dear reader — indulge me for a moment. We know Excel doesn't make calculation errors. It doesn't hallucinate when you're doing SUM(A1:B2).

However: This headline is exactly what we are seeing from so many Conversational AI vendors today. Tools. Solutions. Offerings. To reduce errors with Generative AI.

Reduce?

Do you know how this reads to executives who are charged with deploying and taking personal responsibility for your services?

Not good. Less than ideal. Almost helpful.

So many of us are rushing to extol the virtues of GenAI today — and we are trying to hard to meet the expectations of industry. Dear reader, you are not.

When you are using language like “reduce”, that message is being heard loud and clear.

The message: This stuff ain’t ready for prime time.

Some customers will be fine. In any industry where there is limited criticality to the messaging content, it’s all good. And anything you can do to reduce errors is great. Anything you can do to make sure the thing doesn’t start giving out recipes when it’s meant to be telling you whether the handbag is in stock, is good news.

It’s all good if you can keep the chatbot on topic when I’m booking a service appointment for my car, or changing the appointment for the dog’s annual visit to the Vet.

You need a better, better option for any mission-critical industry though.

Be clear - these industries will absolutely pay (top dollar) – but not for “reduced errors”.

I was in a virtual room yesterday with serious executives from serious public companies, every single one of them a current Chief “something” (eg Digital or Data Officer) or a recent Chief, all from a wide array of industries.

This was essentially the dialogue: GenAI (today) in the context of Conversational AI is great if it doesn’t matter. It’s great for ideas, for summaries, for chatting. But at scale, and in mission-critical situations, give me something that works, today.

Don’t offer to simply reduce my errors.