Using AI to Build out Personas

ChatGPT can analyze your call transcripts to figure out what your prospects want

“I know my prospects and I know what they want.” 

Stacy was getting frustrated. I had listened to a couple dozen of their first calls with prospects and I thought I saw a pattern there, or at least the lack of a pattern.  Stacy is the CEO and cofounder of a data engineering company and I had been working with her on improving the conversion of first calls to opportunities.

I noticed on all these calls that whenever the prospect talked about the challenges that they had, they never mentioned the need to document how data progressed through their pipelines. However, this was a feature that Stacy‘s rep always brought up on the first call. I was merely pointing out that the reps were spending valuable time talking about this feature, but it didn’t seem to be one of the primary challenges that they were trying to solve.

“Let me go through a bunch of the first calls and pull out what problems your prospects are trying to solve.” 

I realized that the pitch and demo that the AES were doing was just not hitting the mark. Prospects liked it but it’s not like they jumped out of their seats, saying yes, this will solve the problems that I have.  Stacy agreed with this approach and said I could go through and build up what I think were the different personas and what problems they had.

Categorizing the personas 

I knew I couldn’t create just one list of challenges that their prospects were facing.  I knew that they had different types of roles, so one list would be two general and not hit the mark for some of those people.  Knew that the people they first interacted with were at different levels. Sometimes it was individual contributors, sometimes it was managers and directors, and sometimes they were VPs or even c level folks.

So the first thing I had to do was put these people in different buckets. I got a list of the title of the primary person that they met with in all of their first meetings and there were several hundred of these. I thought, how do I even start creating different buckets for these?  

I realized that ChatGPT could be a useful tool so I loaded those titles into a prompt, and I asked ChatGPT to categorize them along two different dimensions – level and type of roll.  The first output was pretty granular, but I realized that level could probably be broken down into IC, manager, and executive. Roll was a little more challenging, but I was able to combine the roles into about five different types. – data engineering, analyst, data scientist, engineering, product management, and other.  Once I had these, I asked ChatGPT to create a table with these titles, their level, and their role.

Who are the best personas?

With these categories, we were now able to get some information about them. How many first calls do we have and what was the win rate with each of them? The most interesting thing we found is that when they talk to individual contributors first they were most likely to turn into a customer and the team thought they should only reach out to management and ideally the executives. There were very few first calls and they turned into opportunities the least.  The company was only targeting management and higher in the outbound.  this wasn’t something we were looking for when we started, but it changed who the team went after for prospecting

What challenges do they have?

Now it was time to answer the initial question – what challenges did these prospects have?  In order to figure this out I got the audio of about 40 first calls.  

Side note: The tool that they use for not has an integration with ChatGPT so we could just leverage that and it’s already set up with the configuration not to use the data for training. Just note that your company might have different policies when it comes to using ChatGPT or similar tools.

I use this following prompt for each of the transcripts.

I'm uploading the transcript from an initial discovery call. $person is from $company and their role is $role. $vendorpeople are also on the call and they are from $vendor. You can find information about their product at $vendorwebsite. Based on the transcript from the call, can you provide details on the problems $person has, which they are looking for $vendor to potentially solve?

$person: 
$company:
$role:
$vendorpeople: 
$vendorwebsite:
$vendor:

Summarizing the data

I compiled the information into a document so that we had the information for each persona and whether it was a winner or loss and then used this prompt to analyze them.

I'm uploading a document that details the challenges faced by a group of people at different companies. Based on what's in the document, what are the common set of challenges seen by these individuals? List them out from most common to least.

I played around with this data a bit and found some other interesting insights:

• Since the calls were broken down by wins and losses, I was able to ask what was the difference between the winds and the losses.  Not a ton but there were a few nuggets on what they should look out for.

• Beyond 10 meetings there wasn’t much incremental value with each call. In fact, it probably could be limited to five transcripts each.

• There was very little difference between individual contributors and management. That was nice because they didn’t have to change their messaging depending on who they were talking to.

• Some of the personas were pretty much the same so they could be consolidated as well

So What Happened?

“I’m shocked”

Stacy couldn’t believe the level of insights.  Their product roadmap and their messaging was focusing on a whole set of issues that weren’t on the list of top challenges.  Of course a lot of the thing on the list were issues that they addressed in the product, but this provided a much tighter list of challenges that their prospects and customers were having.

This exercise, which didn’t take long to do, changed several thing:

• Sales team now added ICs to their prospecting and removed the C-Level executives.  It doesn’t mean they don’t want that connection as part of the sales cycle but it changed the perspective on who started projects

• AEs changed their discovery and demo scripts.  They spent more time focusing on solving the challenges that prospects typically had and asked questions around these.  It also gave them some good talking points around “these are the typical challenges we see with other companies similar to yours.”

• Project Management changed the roadmap.  There was nothing brand new but it allowed them to change the prioritization on things that were more relevant based on what the prospects said vs. what Product Management THOUGHT was important

---