Shipulski On Design

AI may be wonderful in some contexts, but AI makes it more difficult to be an effective engineering leader.  Using AI in engineering elevates the training, development, and mentorship requirements that companies must address to make their engineering leaders successful.

When asked questions, AI gives answers.  The answers may be correct or not.  In day-to-day life, that’s not such a big deal.  But in engineering, that’s not good enough.  In engineering, the answers must be right.  Not almost right.  Not maybe right. Not kinda right. The answers must be right. When engineering leaders design products, those products must work, function as advertised, satisfy customers’ needs, meet the cost target, and not hurt people.  If an engineering leader follows AI blindly, there’s no guarantee the product will be successful, profitable, or safe.

When asked the wrong question, AI gives an incorrect answer to the question that should have been asked.  If an engineering leader misunderstands the context, there’s a good chance they’ll ask the wrong question.  And they won’t know it. At best, this will create rework and project delays. And in the worst case, the company could launch a product that hurts people.

And there’s an even more troubling situation: when the engineering team presents solutions to the engineering leader without informing the leader that AI was used to generate the solutions.  To protect against this failure mode, the engineering leader must recognize when AI has been used and dig into the questions/prompts the AI was given.  From there, the engineering leader must use their knowledge of the design context to decide whether the prompts were valid and whether the answer is useful or viable.  This is a heavy lift and requires highly capable engineering leaders who have been trained to recognize AI use and verify the validity of its solutions.

Can your engineering leaders use their knowledge of the design context to provide the right prompts to an AI?

Can your engineering leaders challenge the applicability of an AI’s answer even when they think they’ve provided good prompts to the AI?

Can your engineering leaders detect when their teams have used AI to generate solutions?

Can your engineering leaders challenge the engineering teams when they suspect AI has misled the engineering team?

I hope the answer to those four questions is yes.

Image credit — Richard