-->

KMWorld 2024 Is Nov. 18-21 in Washington, DC. Register now for $100 off!

When is good enough enough?

Article Featured Image

We live in a world rife with disinformation, which has always been the case. It’ s a bit different now in that disinformation is propagated and disseminated faster and more widely than ever before, but, ultimately, there always has been inaccurate information, and there always will be. In the world of information and KM, we are (quite rightly) obsessed with accuracy. As the mantra goes: The right information to the right person at the right time. But the question is, must the right information be 100% accurate?

It sounds anachronistic even to ask the question but consider, for example, these two contrasting, yet common, KM use cases—legal and customer service.

Legal firms require information sources to be as nearly 100% accurate as possible: There is no room for error and certainly no room for “hallucinations” to find their way into legal matters. As for customer service, 95% accuracy in meeting a customer’ s inquiry is likely a massive improvement over the FAQ index-reliant chatbots relied on today. If a customer service response is not 100% correct, the customer will escalate to an agent. If legal advice is incorrect (and, worse still, if AI generated the error), the client may escalate the issue to a state disciplinary board. These are two common areas of focus for KM—two radically different scenarios when the “right” information isn’t always 100% accurate.

I can add a third scenario: marketing. One hundred percent accuracy is not required here—compelling, practical and not illegal are the key parameters. Excuse the cynicism, but hopefully, you get where I am going with this.

Generative AI

You may have noted that I didn’t reference AI, particularly generative AI, until paragraph three, even though it seems to be the law that all columns written in 2023 should lead with the topic. But I couldn’t avoid it forever, as generative AI is at the peak of its hype cycle and dominates the tech landscape. Moreover, it’s a tech advancement that directly impacts the world of KM. Technology vendors that sell into the KM market are almost rabid in their enthusiasm to embrace generative AI, and, overall, it’s fair to say, with good reason. But to invoke a phrase from my homeland, it’s “horses for courses.” In other words, what runs well in one place may not run so well in another.

A 100% accuracy rate is a worthy goal for KM, but it’s not realistic. Experts that manually curate a specialized knowledge-base will make errors, not intentionally, but it’s not and never will be feasible to validate every datapoint in information sources. Generative AI cannot, and will never, deliver 100% accurate results. Likewise, it’s simply not feasible to validate every datapoint before processing it or every output before giving it to the user. Mistakes will happen; that is just how it goes. But the criticality and burden of those mistakes vary enormously.

KMWorld Covers
Free
for qualified subscribers
Subscribe Now Current Issue Past Issues