Sales enablement is that activity you know is essential but never really get to. It always comes last. Part of the reason is because it is so time intensive: One-on-ones don’t scale, and even if a manager carves out time to listen to calls at 1.5x speed, they can only cover one or two reps. Enablement is also rarely an emergency because it’s not clear it has failed until reps underperform.
AI is starting to change all that. Natural language, call recordings, and automated analysis can, in some scenarios, allow sales leaders to spread their attention across more reps and identify “unknown unknowns,” or emergencies that demand their attention but which they had no inkling of. Because how else could you instantly spot all the sales reps who have a great value-selling talk track, but never actually use it? And give no outward signs you should be concerned?
What follows is a brief explanation of a Claude CRM integration use case that Neil Harrington, director of sales at LeanData, built along with six ways you might use it with other tools. This all might sound technical but we promise, it is not. It only requires copying and pasting. Neil says he put it together while looking after his spirited toddler.
You can view the full recorded session here, and we have summarized the instructions below.
This is an installment in a many-part series based on our GTM AI Labs virtual meet-ups where our portfolio companies get their hands dirty applying AI to GTM problems. If you’d like to join our community, please register here.
But wait, why is AI training helpful?
Because you can’t be everywhere. Neil has always felt there was a gap between what he could figure out by listening in on sales calls and what he would request from his data science team. Listening to calls didn’t scale, and the information data teams surfaced never seemed to be relevant enough, partly because it focused on quantitative metrics.
Neil needed an actual readout based on everything his sellers were doing. “I wanted to know, what’s happening with every rep? Are they evolving? Getting better? Getting worse?,” he says. He realized that with Salesforce and Claude’s Model Context Protocol (MCP), he could give a large language model access to a good swath of his sales reps’ activities.
Claude’s MCP is, at the time of writing, the easiest way to make this CRM connection. It is “an API on steroids,” says Neil, as it can push code into your systems to run scripts and do its own research. A cautionary note: This also means you must be careful about what you give it access to. (We recommend giving it only standard user permissions so it can’t rogue edit or accidentally delete anything important.)
How this setup works:
- Download Claude desktop (necessary for MCP)
- Set up your own Claude MCP server by going to Node.js
- Give Claude Salesforce access (ideally, your sales call transcripts are logged there)
- Ask Claude questions
A note of caution: Claude will sometimes err. It told Neil one of his sales reps seemed checked out and needed immediate attention. But that rep was on a well-deserved vacation. Don’t just trust. Verify.
6 use cases for sales enablement
1. Calculate the dollar value of your at-risk pipeline
Have Claude analyze your sales pipeline via the opportunity object for pipeline risk. Claude can summarize that data, those activities, and all the call transcripts to make assessments and generate a quick-read report.
In the example pictured, Claude actually came up with the idea of using a donut chart to convey the exact value of each at-risk segment. “Seeing it by dollar value is interesting because it’s not a view I’d thought about before,” says Neil.
Use it to:
- Analyze on a regional level
- Analyze on a team level
- Analyze on a manager level
- Analyze on a rep level
- Compare to the prior quarter or same time last year
2. Rate sellers on how well they apply your sales methodology
Neil’s team was trained on Sandler sales methodology. If your preferred methodology is well-known or you have documentation that you can share with Claude, the model can analyze whether sellers are, say, setting upfront contracts, closing off stages, or falling into “mutual mystification” traps and ending calls without next steps.
You can go further and assess whether that sales methodology is even helping your team. Different methodologies apply better to different industries and buyer temperaments. Does yours actually help you close deals? Is a rep who scores high on the sales methodology more likely to hit quota? If they are not, that’s valuable intel.
Use it to:
- Produce a rep-by-rep assessment
- Find phrases / attitudes correlated with closed losses
- Suggest who should mentor whom
- Produce specific rep guidance for your sales trainer
- Create a GPT to act as a sales trainer
3. Analyze low performers
Have Claude give you a second opinion on how low performers can adjust their technique before they’re put on a plan. Ask it, “Where should they focus? What should they do? How can you hold them accountable?”
You can ask Claude to support all its findings with verbatim quotes from that rep or the buyer. “It is very articulate about what some reps’ challenges are,” says Neil. “I think a lot of sales leaders don’t do this type of analysis until too late so they can’t take corrective action in time.”
Rather than resist AI-generated guidance, Neil found one of his reps was relieved by the information. It was difficult to argue with two years of call recordings summarized with verbatim quotes, and it opened a conversation that finally helped them improve.
Use it to:
- Create a performance improvement plan
- Create a personal model coach
4. Generate account summaries
Generate account summaries with account statuses and next steps. This analysis is only limited by the accuracy and consistency of your Salesforce data. You can also ask for sentiment analysis of calls, the trend over time, and if it has access to product data and usage trends.
Those summaries are useful gut checks for reps, especially if compared to their notes. Is there anything they’ve missed? Do they tend to be too optimistic and let competitive deals slide when they should actually be more concerned?
The summaries can also show leadership what’s happening without the constant cycle of check-ins. Plus, it makes its assessments with a greater amount of data than reps alone have available to them.
5. Onboard new reps
Instead of rote, multiple-choice quizzes, create a living “quiz” reps can chat with just like a prospect, avatar and all. Sellers may find this more encouraging and forgiving than typical quizzes, and it supports more learning styles—reps aren’t dinged for minor terminological or grammatical infractions if they understand the general concept.
Another benefit: Reps can onboard at their own speed. If that model is operating with the full understanding of your team’s current pipeline, including the full history of their territory, they can start planning and tailor their learning to certain industries or objections from day one.
All the while, this mode of onboarding can save you, the manager, from spending time on the basics and more time coaching. It can also save you and your sales enablement team from updating the training materials. Simply upload the new pricing or messaging deck and the quiz updates itself.
“You of course still need a human touch,” says Neil. “But it can create much more variety in the training.”
Use it to:
- Create a persona quiz: “Which persona does this sound like?”
- Create a use case, pricing, and packaging quiz
- Allow the rep to ask their own questions and self-guide
6. Self-coaching
This Claude model can act as a coach and guide reps through realistic role-play. This can be as simple as a text-based chat or you can use an AI-generated avatar tool such as Tavus.
The model can also give feedback on how reps are prioritizing their days. Give it access to their calendars and note-taking apps and they can ask it how they should prioritize their time. Given its access to the CRM and call recordings, it can tell them which deals have deadlines rapidly approaching and which accounts are at risk.
As manager, you can have it coach you through difficult conversations, and try coaching avenues you hadn’t thought of yet. Bringing this full circle, you can have it create a coaching guide for each rep based on where it believes your team needs help.
Use it to hunt for unknown unknowns
Using Claude to follow your curiosity can help you uncover “unknown unknowns.” For every sales training issue you know about, there are five you don’t, and Neil finds that this tool helps surface them. It requires experimentation and practice, though, and it is easy to fall into information overload. There is only so much you can be aware of and act on—hence the summary reports. Use it to tell you which issues matter most and dig into Salesforce and your call recording tool to verify before drawing any conclusions.
Where it really grows useful is in providing horizontal insights. Yes, you can learn about individual reps, but are there certain phrases or prospect attributes correlated with losing deals? Winning them? And beyond the presales process, what customers end up being profitable, worthwhile, and easy to support? Which types are most likely to refer others and generate bluebird deals?
We are in the early stages of experimenting with these tools, but based on Neil’s initial findings, they seem promising.