Where To Draw The Line With AI

Anu Choudhury is the Chief Technology Officer of Reputation, a global leader in reputation intelligence.

Companies of all shapes and sizes are embracing AI, and many are seeing real benefits. But there are certain tasks AI should not do on its own, and that’s where I, as a CTO, draw the line.

For example, I don’t want AI posting publicly on behalf of my company, or any other company for that matter, on its own. This means no autonomous AI comments on the internet, nothing pushed live on LinkedIn, without a human in the loop to check it first.

Unfettered AI like this is a liability waiting to happen. We all know AI has a tendency to hallucinate, and it’s your good name that suffers if your AI loses the thread and says something that isn’t true. That’s a serious risk because once your reputation gets damaged publicly, it’s hard to win back trust. You sure won’t do it by deleting that one rogue post.

Here’s what unfettered AI can look like in the wild. Let’s say a customer gives you a one-star review because they were unhappy about a particular service. Your AI agent might see that review and respond right away, professionally. But in its quest to always be helpful, it might also invent a discount code out of thin air. Because nobody at your company is monitoring the AI’s outputs, that unauthorized discount code is now sitting out there in the wild for everyone to see. And the original customer? They get even angrier when they try to use the code and it doesn’t work.

That’s not a situation any company wants. And it’s why my number one rule for AI is to have a human in the loop for any public-facing outputs, such as review responses and published content. I’ll hold onto this rule at least until AI can earn my trust and prove to me that it can generate appropriate answers again and again, at scale.

Not Everything Needs Oversight

Don’t get me wrong. I’m not an AI luddite. I actually think most workflows are better off fully automated, with no human intervention required.