Kelford Labs Daily: Prompting the model

In generative AI, there’s a concept of “prompt design.”

Basically, while the large language model only has its existing training data and parameters, there are techniques you can use in your “prompt,” or input, that will induce it to generate different outputs of varying style, quality, or accuracy.

“Prompt design” is the simply the skill or speciality of being able to get superior outputs by optimizing your inputs.

I bring this up because people are the same.

The information in your brain is the information in your brain, but the way you’re prompted—or asked—a question influences your ability to answer it. It changes the ideas, concepts, experiences, or events that you recall in that moment.

So, when we’re talking to our clients, it’s important we ask them questions that get the most useful and accurate responses.

For instance, what I don’t ask clients is: “What’s the single most valuable thing you deliver to your customers?”

Because I know I’m not going to get the best answer. What immediately pops into mind—what almost certainly just popped into your mind—is the thing you want to be true. It may also be true, but I know for sure that you want it to be.

What the question, or prompt, ends up outputting is where you’ve invested most of your energy. Where you try the hardest, or what you like most. Or what you’ve be least willing to part with.

And that’s all great to know, but not the question. Because what if it’s not actually what the customer values most? That’s the answer I want.

So, instead, what I ask is: “Of all the different elements of your business—product lines, service offerings, technical components, departments, etc.—which single one, if removed, would eliminate the most value? Which part of your offering, if it went away, would take the rest of the business with it?”

Now, suddenly, the mind leaps to where the real value is. What the customers are actualy buying most. Where you’re seeing the biggest bang for your buck.

The questions—the prompts—are essentially the same, but the answers—the outputs—are vastly different.

But why am I telling you this today? Because, when you ask your clients for case studies, testimonials, or referrals, it’s important to ask the right questions to get the right answers.

By default, people who give testimonials say what they think you want to hear, instead of what your other prospects would be most moved by.

So here’s a question you can try asking instead of “What did you like about working with us?”

“If you referred us to a friend or peer and they said we were too [expensive / small / local / far away / insert your own common objection], how would you respond to convince them to give us a shot?”

Try using prompts like that, and I think you’ll be surprised by the quality of the outputs.