Prompt Engineering For Sellers

Before I get into this article, credit for the idea comes from a conversation with my friend, Jack Malcolm. Thanks for provoking the idea.

The term, “prompt engineering,” is relatively new to most of us. While software developers and AI specialists have understood the concept for year, it’s very new for the rest of the world.

But as the LLMs have emerged and grown, the concepts of prompt engineering are starting to fill our feeds. We now have lists of AI generated prompts for just about any scenario, “Generate a prospecting email for this…., How do I handle this objection…, What are the strategies I should develop to do this…., How do I build a sales organization to achieve these goals…, What marketing outreach programs should we be thinking about…, What are the key issues CROs are worried about in this economy?”

Leveraging these prompts results in the LLMs giving us answers. In some cases, they do the work for us, generating 1000’s of emails. In some cases they help us identify key issues in certain situations.

When I look at the most sophisticated users of these tools (not just in selling), I see them not just using a single prompt, but engaging the LLMs in a conversation that drills deeply into the issues. The deeper one drills, the better the quality the answers these tools provide.

Prompt engineering is one of the hottest jobs in the market. These roles have specialists developing very rich prompt sequences, providing huge insight.

The reason prompt engineering is so important, is these tools are designed to give the user answers. They want to please the user, so they tend to give the answers they think the user prefers. But that doesn’t mean they are giving the best answers or even the right answers (this is not just about hallucinations, but about any answer). Recognizing this weakness in the tools, prompt engineers get into very rich conversations. They might start with, “Give me all the benefits I should be talking about this product to this cusotmer…” Once they have gotten the response, they might say, “What are the top 3 objections customers will have to each,” or “Imagine you are a competitor, how would you respond,” or “What should I be asking to make sure the customer deeply understands this issue,” or “What should I be asking to understand the impact of each issue?”

Prompt engineers don’t accept the first response these tools provide. They drill down seeking to better understand each one. They ask for differing points of views. They create debate scenarios where the debaters would argue the pros and cons. They ask if these are the most critical issues or if there might be things we are missing. They drill deeply into each one, looking for differing points of view until they have the richest understanding possible. And prompt engineering is the means by which they conduct these conversations.

And the process is a shared learning process. The responses from the LLMs provoke different thinking, questions and prompts. In turn, as the LLMs respond to these refined prompts, they learn more, consequently provide us higher quality responses.

Now what does this mean for sellers?

How do we leverage these principles into helping sellers become better prompt engineers? I don’t mean in their use of prompt engineering with the LLMs, though that is important. But think about how we might apply the principles of prompt engineering in our conversations with our customers.

In most of our conversations, we leverage relatively simple questions (prompts) with our customers. And they provide us, like the LLMs with very simple responses. For example, “What are your needs and requirements?” The customer responds with their needs and requirements. Then we go to the next question, “What solutions are you considering to help meet those needs,” to which they give a simple response.

We march through our prompt scripts getting the surface level answers. But we and the customers aren’t getting much out of these conversations. When we drill down, asking things like, “Why is this an issue…., What’s the impact…, What if you chose not to address it…., What if you changed the way you think about it,” we get much richer answers and understanding of the customer.

As we drill down with our prompts, we get smarter based on the responses, and the customer gets smarter in this process. We have far richer conversations.

Conversations, whether with LLMs or with people become much richer when we don’t accept the initial responses, but use those for follow on prompts, drilling into the issues much more deeply. What we get from LLMs is much deeper levels of response, from which we learn more. Likewise, with much deeper level of prompts in our conversations with customers, we and the customer get smarter in the process.

There’s one level deeper to think about the process of prompt engineering our conversations with customers. The limitation of most LLMs is they are trained to give answers. They don’t tend to ask questions (they struggle with that when prompted). The conversations tend to be one way, oriented to satisfying us with the answers we seek.

But imagine what might happen in our conversations with customers. Our prompt engineering forces us to drill down, getting richer answers, perhaps getting the customer to think differently. But our customers have much richer intelligence than the LLMs. These conversations can engage them in their own prompt engineering of the conversation.

Now we are engaged in very rich two-way collaborative conversations.

How are you developing your prompt engineering skills? Managers, are you coaching your people in their prompt engineering? Sales Enablement, are you providing your sellers with rich prompt engineering skills, both for their use of LLMs, but in enriching their conversations with their customers.