Let me take you back to a pivotal moment in AI history. In 2011, IBM’s Watson competed against Ken Jennings in Jeopardy and won – a breakthrough that helped launch the modern AI revolution.

But here’s something few people realize: Watson achieved this landmark victory with only 71% accuracy.

Think about that for a moment. For those of you with children or nieces and nephews, when have they ever come home with a C- on a test and you’ve said, “Job well done”?

You don’t.

Yet this “C-minus” performance was enough to beat one of the greatest human Jeopardy champions. This moment fundamentally changed how we think about machine intelligence and kicked off the artificial intelligence wave we’re experiencing today.

“Perfect Is the Enemy of Progress”

In my years advising enterprises on AI strategy, I’ve noticed a common misconception: the belief that AI must be flawless to be valuable.

This expectation creates paralysis and prevents organizations from deploying solutions that could deliver immediate benefits despite occasional imperfections.

The question isn’t “Is this AI perfect?” but rather:

“Is it better than the human error rate occurring within your current business process?”

This shift in perspective can transform how you approach AI implementation.

Here’s a quick example: Imagine you run a customer support center where human agents resolve tickets with 85% accuracy. You implement an AI assistant that handles routine queries at 90% accuracy.

It’s not perfect — but it’s an upgrade, and it frees up your team for higher-level issues. That improvement, however marginal, compounds in efficiency, speed, and customer satisfaction.

Your AI doesn’t need to be perfect—it just needs to improve upon what you’re doing now.

Three Principles for Realistic AI Expectations

As you develop your AI strategy, consider these principles for setting realistic expectations:

1. Benchmark against your current reality, not perfection:

Compare AI performance to your existing processes, not to an idealized standard no human could meet.

2. Start with well-defined problems:

Choose applications where context is clear and data boundaries are established—like customer operations or anomaly detection.

3. Maintain the human in the loop:

Use AI to augment human capabilities, not replace the contextual understanding and judgment only humans can provide.

While setting realistic expectations about AI’s capabilities is important, there’s one expectation you should never compromise on: human oversight.

No matter how capable AI becomes, the human-in-the-loop remains essential. The organizations succeeding with AI today aren’t waiting for perfect technology.

They’re thoughtfully applying valuable-but-imperfect AI to appropriate problems while maintaining human oversight.

After all, a C+ solution that’s actually implemented often delivers far more value than an A+ solution that never leaves the lab.

Until next time,

Sol

P.S. What area of your business could benefit from a “better than now” AI solution — even if it’s not perfect yet?

Every Friday morning, I deliver 1 actionable insight to help you navigate the post-AI landscape, simplifying complex transitions into a clear path for business impact. You can click here to subscribe.