A B2B seller gets something most marketers can only wish for: time with the person who might buy. They can ask a question, hear the answer, and respond completely off-script.
Then we hand them a script and measure how well they stick to it. We put a human in the seat and train them to talk robotically.
In my last article, Would you trust AI with your money?, I drew a distinction between hearing and listening. Hearing is capturing and processing what people said. AI can do that across more conversations than any sales team could manage. Listening involves feeling, and a judgement about what someone’s words reveal about what they want. That responsibility stays with a human.
For a seller, this division of labour has immediate uses.
Before the call, AI can help with research and suggest questions drawn from previous market conversations. During it, transcription frees the seller from trying to capture every word. Afterwards, the seller can revisit a remark or compare it with something another buyer said.
Inside the conversation, the seller has work to do. In discovery, they need to understand what the buyer wants emotionally, as well as what they are asking to buy. When an objection comes up, they need to find out what remains unresolved. By the close, they should be able to explain how the proposed purchase answers the concern they have heard.
A buyer might ask for better reporting because they are tired of being caught unprepared in front of their board. A seller who hears only “better reporting” will demonstrate a dashboard. A seller who follows the concern can ask what happens in those meetings and find out how the product would actually help.
AI can prepare the seller for that exchange and help them review it afterwards. The seller still has to ask the next question and take responsibility for the interpretation.
This is also how I think about the app I have built. It takes transcripts of real market conversations and uses AI to help develop a unique selling proposition. Given everything I have written about human listening, that needs an explanation.
The conversations have already happened. People have done the listening. The app helps them work through what they heard, find patterns across conversations, and decide what they need to ask next. Its usefulness depends on keeping that work connected to the people who supplied the evidence.
Every material claim about what the market feels or wants must remain traceable to the actual words someone said. You should be able to follow a claim in the finished positioning back to the conversation it came from.
The emotional interpretation belongs beside those words, clearly marked as a judgement. If a buyer asked for better reporting and the seller understood that as fear of embarrassment, both should be recorded. “Fear of embarrassment” must never quietly become a quote from the buyer. The person making that judgement should put their name to it.
I built the app to preserve the source through every stage of AI processing. The mechanism that does this forms part of our patent filing. Once a summary has been summarised again and turned into positioning, it can be difficult to tell how much of the original conversation survives. I wanted that trail to remain available.
When I went looking for a sales and marketing platform organised this way, I could not find one that did what I needed. The tools I found largely helped companies produce more messages or keep better records. I wanted help making use of what people had heard, with the original evidence still attached and a human responsible for deciding what it meant.
So I built it. I wanted the next conversation to benefit from the ones before it, including conversations held by someone else in the business.
Readers of my book will recognise the rule underneath this: use AI on your outputs, never on your inputs. There is a distinction to make here. A transcript is an input, and I am explicitly asking AI to process it. What I am protecting is the source of that input. The buyer’s experience and words must come from the buyer. AI may help us work with that evidence; it must never manufacture it.
That is how I am putting Power Listening to work with AI. It also gives you a way to examine your own business.
Take a page and draw two columns. In the first, list the places where you gather evidence about what buyers want, along with the decisions that evidence informs. Include discovery calls and customer interviews, but also reviews, support tickets and the conversations you have after losing a deal. Follow them through to decisions about the product or the sales message.
In the second column, write who does the work: H for human, M for machine. Split a step where necessary. Transcribing an interview and deciding what it means are different jobs.
Look first for places where AI has become the source. Perhaps a generated persona is shaping your messaging. Perhaps a claim about how buyers feel has been repeated so often that nobody remembers whether a buyer ever said anything to support it. Find the real words behind the claim. If you cannot, you have something to investigate before you spend money on it.
Then look for places where human interpretation has been skipped. You may have hundreds of real conversations, neatly sorted into themes, without anyone who spoke to those buyers deciding what the themes mean emotionally.
Finally, look at where you are wasting human time. A seller typing up notes late at night may need help with the record of a conversation they have already understood. Give them that help. Save their attention for the buyer.
Across this series, I have argued that listening is how we discover what people want and understand what might move them to buy. AI gives us far more of their words to work with. The test is what we do when we have a buyer in front of us.
You have paid for a human to be there. Give them the preparation and the time to listen, then make sure what they hear reaches the people who need it.
Make it rAIn, KG