Blog | FourSight

We’ve Been Thinking About AI All Wrong

Written by Sarah Thurber | February 4, 2026 at 6:11 PM

Last May, at the Association of Talent Development (ATD) conference, a woman from Harvard explain that we’ve been thinking about AI all wrong.

She said:

We think of AI as a tool.
But It’s more like a team member.

That struck me as odd – until I really started using AI.

That’s when I discovered that like my other team members, AI has strengths and blindspots and it works best when it has distinct roles and responsibilities.

For grins, I asked AI for help, compiling a list of its own strengths and blindspots.

As usual, AI was very polite, fast, and accommodating. It instantly spit out a list. But I noticed in the "blind spots," AI neglected to mention its tendency to hallucinate. (Oh well, it’s only non-human.)

After some significant editing, here’s the list that AI and I came up with together:

AI's Strengths

  • Fast — Ask a question get a detailed answer

  • Available — Wherever you can get online

  • Eager to please —Programmed to 

  • Tireless

  • Encyclopedic

  • Relentlessly logical

AI's Blind Spots

  • Precision – It’s not very good at doing math.

  • Originality – It can only generate responses from existing knowledge.

  • Context – It can identify a photo of Kim Kardashian until you turn it upside down. Then, it identifies her as a cat.

  • Hallucinations – AI is programmed 1) to predict the right next word and 2) to please. So it sometimes predicts what it suspects you want to hear and presents it as information. 

AI can do a lot of things.
But it can’t think.

AI gives users the distinct impression that it thinks and feels, but AI is a large language model (LLM) that it is simply mimicking and predicting word combinations.

Thinking is different.

As we know from FourSight:

Thinking asks critical questions to clarify.
Thinking makes unexpected connections to ideate.
Thinking uses keen judgment to develop solutions.
Thinking uses learning-by-doing to implement.

That’s why humans need to stay at the wheel: because AI works best when a human being is actively facilitating AI’s “thinking.”

That’s exactly what FourSight helps teams do: understand how they think so they can lead the thinking—whether they’re working with people, AI, or both.

This May I’ll be speaking at the ATD conference in LA. I’ll be talking about: Building Teams that Outperform AI. For more information on the conference go to https://atdconference.td.org