Mining Valuable Marketing from Successful Sales Calls with AI
Man Bites Dog I guardianmarketing.substack.com
Successful sales calls, where prospect becomes customer, are a goldmine of ideas and copy which can be turned around and used to bring in new prospects and turn current ones into customers.
The challenge is often taking the time to rake through all of your calls for this information.
AI is an exceptional tool for processing large amounts of information quickly, which is essentially the position you are in when:
you are trying to figure out copy and marketing materials
you have a number of recorded successful sales calls which you can dig into for key information about what leads people to buy and why
Without going into any reasonable depth on copywriting, one of the best places to look for language to figure out what to say to lead new people to buy ...
Is to start with the language used which already has led people to buy.
Groundbreaking, I know.
Personally ... there's little that is as mind numbing as listening through hours of sales calls and interviews to note out key statements, ideas, sentiments, and beliefs of your prospects turned customers. However, it's kind of foolish to not do this.
Enter,
Ai.
While I don't like AI to do my thinking for me, one task it is exceptionally good at is processing text information and telling me what's in it.
Which is exactly this situation, when we have sales calls and a desire to pull solid copy from them.
So,
I'm gonna show you how to do that, and include the sorts of questions you can ask the AI to get the kind of information you need.
Let's go ...
Getting valuable ideas from sales calls quickly
(In my specific example, I use a paid ChatGPT account, but you can do this with any chat based AI so long as it allows you to upload text files - beware that some other chats may limit the amount of text they are able to process at any given time, and you may need to break up your files or just copy smaller chunks of text straight into the chat)
For the examples below, we were starting with Loom recordings of sales calls. Loom is a fantastic service for this sort of thing because it automatically provides a transcript, and a transcript is what we are after. You can upload video recordings to Loom to take advantage of their transcripts (most likely you'd want a paid tier, but it appears you can get transcripts on their free tier, you're just limited to 5 minute videos).
If you're looking for free options to try this out, another decent route is to upload a video to youtube and then use one of a few free youtube transcription services (which you can google). I just find these are less accurate.
If you really get into this, you can use some sophisticated techniques with OpenAI's Whisper API to mass produce transcripts - something I demonstrate in the AI Email Masters program1.
Here are the steps in short:
1) Start with a transcript of the interview. As mentioned, Loom is excellent for this. In Loom you can find (on desktop) the transcript on the right side. By default it's "Activity" but there's a tab that says "Transcript." Enter the transcript and then what I do is "copy" that transcript into a regular text file (I just use text edit on mac).
2) I save that as a txt or rtf.
3) Then in GPT (latest model) I said this:
I need your help summarizing and organizing information from some sales call transcripts. I have a few different transcripts from different sales calls and I want you to summarize the calls, give me bulletpoints of important details and then for me to be able to walk through the call by asking you specific followup questions. Can we start with the first transcript?
4) GPT says sure go ahead, and I just send the txt file with the transcript in it.
The base summary from that instruction has been so far very useful.
Right off the bat GPT identified the purpose of the call, the background information of the person on the call (organizing available demographics and key interests of the prospect). GPT identified the goals of the call and outlined the key concerns and objections which the prospect had that the sales person overcame.
So an excellent base summary to start digging and asking further questions.
5) What I do from that point is read through the summary GPT provides, and identify points I want clarified. In this case I've been looking for specific talking points, how the prospect talks about them and how the sales person responds to them (which ultimately results in the sale).
This has been highlighting some key concerns, pains, and interests which I can more clearly talk about in the copy (in this case it's an email project).
I probed specific questions about the content which I was interested in knowing. How this sales person talked about the content and the program and how the prospect responded to those ideas.
I also had GPT give me a detailed analysis of who this prospect turned customer is, along with why they decided to purchase. (And repeated this process through several calls with different people)
If necessary I can go back into the transcript (searching for the talking point) and then the timestamp in the original interview to listen.
Overall, the process is straightforward to start, and then turns into a back and forth conversation as you find new ideas and details to dig for.
Some more examples of followup questions I've used:
"in the conversation when {PROSPECT} seeks reassurance about program legitimacy and safety, what does {SALES PERSON} tell him?"
"Based on this interview what can you tell me about {PROSPECT}"
"why did {PROSPECT} decide to have a conversation with {SALES PERSON} to begin with?"
"How, specifically, does {SALES PERSON} respond to {PROSPECT}'s {fear}?"
"Based on this interview can you tell if there was one dynamic which convinced {PROSPECT} to buy?"
Other questions and ideas to dig for
The purpose of using AI to process through sales calls and interviews like this is to try to find not just more understanding about the person who has become a customer and why they did, but specific phrases we can use, fears we can use, roadblocks we can hear they overcame or discern they overcame, etc.
All of that can become the basis of the words in the copy itself, as well as ideas for campaigns, emails, ads, etc.
Things we want to dig for ...
What do they fear and why?
What do they desire and why?
What has kept them from getting the result/outcome they are now considering buying?
Why are they here talking to us vs anyone else?
What are they looking to *justify*? (Many times people's reticence to buy is just a teetering balance of justification between the cost and the promised outcome - can we figure out what they are teetering over and what one thing tips them into buying?)
What beliefs do they appear to have?
Belief is an interesting one and something I've played around with in respect to R3 for Email - where I focus heavily on belief based ideas simply because i know that if my prospects believe what I believe, they'll follow and listen to me.
ALSO
Beliefs are a great place to work on stealth influence2. If we know what they believe, or what they at least assume about their reality, and we can identify evidence to the contrary, then we can position some of our content to put simple cracks in their view of reality, ultimately possibly leading to breaking down the belief and replacing it.
How to do this with AI ...
We can't just exactly ask AI for this information. "What do they believe?" is a question worth asking because sometimes it leads to interesting threads of discussion with the AI as we uncover new information.
But AI doesn't itself have a concept of belief.
Because the idea of "belief" means something specific to me, it's on me to be specific in my conversation with the AI in order to pull out useful information. I categorize 'belief' as philosophy and principle. It's easier to get AI to identify principles because they are known/definable/measurable.
I can also try to identify what the person is assuming about reality. What foundational assumptions do they appear to be making, because we can focus on breaking those as well.
So, "What do they believe?" is a fine place to start, but we must be flexible and willing to dig back and forth with questions to find the definable answers (if we’re asking AI for opinion then what we get is not very useful).
We can ask
"what principles appear to be driving their decisions?"
"what does the customer seem to know to be true?"
"what facts appear to influence the customer's point of view, actions, and decisions?"
Fundamentals with AI
I find it useful to approach working with AI as if I'm having a conversation with another person. I recognize that the AI is not a person, but these chat based AIs are actually built this way. So if you approach this work as if you're having a conversation, I find you will get better faster more useful results.
Remember what you are seeking in the outcome.
Clarity of understanding about your prospect's fears/desires/interests. What drives them. What turned out to be the tipping point in their buying decision. What they needed confirmed/denied in order to make the sale happen.
And finally,
Confirm the information.
Ai is a powerful tool, but it also has a certain level of freedom built into it and sometimes it makes stuff up, or just gets facts wrong and does not process information the same way you or I would.
So far, when I've done transcript analysis this way, I've found that the AI references the sections of the transcript where the facts come from. I can also ask it to tell me where in the transcript it got an idea or understanding about the customer. Then I can use time stamp to check the transcript and even the audio recording.
If it's critical information, I'm going to check it against reality.
It may sound like a lot of extra work, but the reality is that processing sales calls this way takes 70-90% less time than listening to the whole thing yourself.
Want to go deeper?
You don't need to become any kind of expert or seriously integrate Ai into your workflows to use the above technique.
Got sales calls and want to pull copy/marketing/ideas from them? Run the transcripts through and ask the AI questions about the conversation.
Relatively simple.
If you have a deeper interest to integrate Ai into your workflows, especially if you are a copywriter and/or focusing on email marketing, you may be interested in AI Email Masters - my principle based approach to integrating AI as a copywriter and marketing strategist.
AI Email Masters is currently available to MBD+ subscribers:
Be Useful. Be Present. Love the Journey.
Joseph Robertson, CMO Man Bites Dog
Ready to Step Into The Arena?
Ready to engage the field? Man Bites Dog paid subscribers have comment access unlocked below. (They’re also sent a killer welcome package in the mail with all kinds of opportunities that are not available in digital format)
Here are some other options:
Get on the waitlist to join the Arena: engagethefield.com
Check out the Engaging The Field Handbook
Grab your your own copy of the R3 system (it’s a book and it’s not cheap)
Ninja stuff. Always looking for new and different angles for leveraging ChatGPT to amplify and accelerate data collection, assessment, and strategizing. Thanks for this, Joseph!