Full Transcript: How to Use AI to Prioritize Expansion Opportunities
Why Sales Teams Struggle to Prioritize
AI can combine CRM records, call transcripts, and external company signals to rank expansion opportunities by likelihood of success. The output is a prioritized list with supporting evidence and suggested next actions, giving sales leaders a clear starting point for where to focus.
In today's walkthrough, I'm going to show sales leaders how to get really meaningful insights using AI in a surprisingly easy way. If you're a sales leader, you probably know that using AI for sales isn't all that intuitive. Using it for researching an account or crafting emails is about as far as a lot of salespeople or leaders get, and I understand that. Beyond the chat experience, using AI isn't all that obvious in this context.
But at the same time, commercial teams often struggle to prioritize the right opportunities for a number of reasons. CRM fields are often stale or incomplete. Rep notes are inconsistent and all over the place. Transcripts contain tons of useful information but are hard to review at scale. Important outside developments at accounts often go unnoticed. And then teams end up chasing the loudest opportunities or the ones closest to renewal, but not always the best ones.
What's interesting is that everything I just talked about is really valuable data. In the before times, you would've maybe needed a data expert to harness all of it, but now you just need to know how to use AI to do it for you. In this walkthrough, I'm going to show you how any sales leader can get started using AI to get real insights about a book of business, using a very simple prompt package I put together and uploading all this data. We're going to get AI to tell us which accounts are prioritized, why, and how much opportunity it thinks there is. And we're going to do it all with a regular AI chat tool.
Data Preparation
The data prep for this workflow is straightforward: a CRM export, CRM notes, external research, and optionally call transcripts. Each file covers a different angle on the same set of accounts, and AI combines them into a single ranked view that no individual file could produce on its own.
The first thing we're going to do is go over our data prep checklist. Part of that is picking the right AI model. I'm going to keep it really simple: you want to use the latest model from the AI provider that you're using, and you want to use the most advanced one. Usually that's something like the thinking or pro model.
Now let's get into the data. The first thing is an export of your CRM. We want to limit this analysis to about 25 accounts at a time. Any more than 25 and the models tend to not finish the job or hallucinate or have problems. If you wanted to do this at scale and have it run through all of your accounts on a daily or weekly basis, that's where you'd hire a company like mine to come and build that for you. When you go to export it, you want to export it as a CSV file and not a regular Excel file.
Next, you're going to need CRM notes. Just grab the notes and export them for the 25 accounts that you're doing this analysis for.
The final thing you'll want, but is optional if you don't have it, are call transcripts for those 25 accounts. If your company is using Gong, Chorus, Fireflies, Granola, or even just Google Meet or Teams, there's a very good chance that those call transcripts exist somewhere. Don't export more than three per account because we don't want to overwhelm the AI. If you don't have call transcripts, don't worry. You can still do the analysis. But conversations with sales reps are rich with context and can really help AI do a better job.
Running the External Research
External research gives AI context about what is happening at each account right now: recent news, leadership changes, acquisitions, and expansion signals. Without it, the analysis relies only on internal data. With it, AI can connect outside developments to the opportunity picture your CRM alone would miss.
We're going to start having some fun because we actually get to use AI. I'm going to copy the external research prompt from the starter kit, go over to our first tab, and paste it. At the bottom it says "companies to research are attached and/or below," so all we have to do is paste our list of companies there.
I'm going to switch over to my Excel sheet. You can see my list of 25 accounts. Now I'm going to copy these from my synthetic CRM data and paste them into the prompt. All we have to do is send it, and now ChatGPT is going to work, basically going out and finding information about these companies. This could take a few minutes, maybe anywhere between three to ten minutes.
The results are in. It provides a summary for each account with a URL for every source. That's a rule we gave it, to make sure it stays grounded and doesn't just start inventing stuff. And we tell it that if it doesn't find anything for a particular company to specifically say it could not find anything. That way it doesn't try to just satisfy us by making something up, which these AI tools often tend to do. That's why we put some guardrails into the prompt.
Running the Master Analysis
The master analysis prompt contains a scoring methodology, a rubric, and instructions that tell the AI how to evaluate the data across multiple dimensions. You copy it from the starter kit, paste it into a fresh chat, and upload your files one at a time as the AI requests them.
This prompt is a little more complicated than the research one. It has a scoring methodology, a rubric, and all these things telling the AI how to analyze the data we're giving it. You're welcome to go in and read it, evaluate it, and take a look. I actually had AI write it for me. AI is really good at that kind of stuff.
We copy the prompt, paste it, and send. GPT tells us it understands the role and summarizes what it's supposed to do. Then it starts giving us instructions. The first thing it wants is a CRM export. It's reminding us that it has to be a CSV file and not to include more than 25 accounts. You don't have to remember any of that stuff. Even if it's in the guide and you forget, the tool is going to remind you. That's because we wrote the prompt to do that. We gave it instructions to give us instructions. Kind of weird, but pretty effective.
It asks for the CRM export, then the CRM notes, then the external research. For the external research, all we have to do is come back to the first tab, copy the response, and paste it. Then it asks if we have call transcripts, reminds us it's okay if we don't, and that we should upload only 10 at a time. After uploading all the files, it triggers the analysis. This took about five minutes total.
Walking Through the Output
The output is a ranked table of all 25 accounts, scored by expansion opportunity. Each row includes a priority tier, estimated dollar value, confidence level, supporting evidence from internal and external sources, and a suggested next action. The table can be copied directly into a spreadsheet for sharing with the team.
It scored all 25 accounts for expansion opportunity. Competitor pressure that only threatens the current contract was treated as a retention risk, not expansion upside. So that's good to know.
On the first column we have the ranking. Basically it ranked the 25 accounts based on the priority of the opportunity using the opportunity score. The tag in the priority tier column is based on the numbers. Anything from 75 to 100 is listed as "Act Now." Below 75 down to 50, we go to "Develop," still a great opportunity just not immediate. Below 50, we see "Monitor." And below 25, you'll see "Maintain."
The estimated opportunity value is interesting. Our CRM data includes how much the company is currently spending with us. AI takes that and looks at other things going on with the transcripts and the news and does its best to figure out what could be the potential upside in terms of revenue. But the dollar amount is not the only or most important thing when it comes to prioritizing. A company might rank higher even with less upside because other signals point to urgency.
The confidence level is driven by how much data AI has for a particular company. The more data, the higher the confidence. And if the data is mostly pointing in the same direction, it enhances the opportunity score. If different transcripts have two people saying different things, or some piece of news contradicts something from a call, that creates doubt and lowers confidence.
Evidence and Next Actions
Let's look at CP Group. High confidence, about $90,000 upside. CP Group is taking over a new Orlando high-rise and explicitly wants a fire and safety life audit after the transition. That's as clear as it gets. One of the transcripts has a contact saying "we do want a full fire and life safety audit once we have the keys."
The key internal evidence includes another quote: "If this new tower has three different contractors for alarms, sprinklers, extinguishers, I already know how that movie ends." She's saying she wants to consolidate. That's a huge opportunity. If somebody says they want to consolidate, they're essentially saying whoever gets this contract is going to have it for all their locations.
The external signals found that CP Group acquired an eight-building Central Florida office portfolio and reported leasing momentum at Miami Tower. Fantastic insight. We know this account just bought a building and they're going to need services.
The suggested next action says our rep should contact the key decision maker within two weeks to schedule a Q2 operational audit for the Orlando high-rise. Not only do we have a ranking, we actually have a suggested next action, which is super useful.
Going Deeper with AI
The initial output is a starting point, not an endpoint. You can ask AI for a full action plan on any account, drill into specific dimension scores to understand the reasoning, or ask follow-up questions about patterns across the portfolio. The value compounds the more you engage with it.
This is a fantastic start to figuring out who we should focus on. But it shouldn't stop there. What I'm going to show you now is going deeper with AI. All we have to do is ask: "Can you give me an action plan for CP Group?"
It gives us a goal, a first email angle, makes the audit scope narrow enough to approve quickly. It even produces a recommended audit scope that the rep can bring to the account. It gives a suggested talk track and a 30-day action timeline in a table we can copy into a workbook. We could send this directly to the rep, and they have a concrete plan. You can do this for every account if you wanted, or maybe just the Act Now accounts since they're the highest priority.
I also want to show one more thing. You might be wondering where the dimension scores came from and what they mean. Instead of just wondering, we can ask: "Can you give me details on the dimension scores for CP Group?"
It spells out each dimension: evidence convergence and signal strength, coverage gap and white space, relationship momentum, timing and urgency, and external signals and market context. Each has a score, an explanation, and the sources it drew from. You can see the weights: dimension one has 30%, dimension two has 20%, and so on. These are set in the prompt. It does the math to get the final score.
Building the Habit of Going Deeper
The reason I wanted to show that is because I'm hoping that the people who watch this feel empowered to go deeper. It's a habit that I think is really important as we transition into an AI-native economy. The habit of asking follow-up questions, going deeper, using your curiosity to your advantage is really important. And if you can't think of what questions to ask, you can even ask AI what else to ask it. That's a little trick I've learned to do myself when I'm kind of stuck. Just talk with it and see where it takes you.
I hope that this walkthrough was helpful and created value for you. Please don't hesitate to ask us questions in the comments, or you can email us. You can find our contact information on our website. The starter kit is in the description below. And don't be shy about reaching out and letting us know if you want to see us do any other particular analysis. We're happy to do so.