One of the most common complaints I hear about Salesforce is that sales reps don’t keep their records updated. They’re on calls all day, meeting with prospects, talking to customers, discussing budgets, competitors, timelines, and next steps. But when someone else opens the Account in Salesforce, much of that information is missing.

It’s not necessarily because the sales reps don’t care about the CRM. They’re focused on selling, and after a day full of calls, updating Salesforce often falls to the bottom of the priority list.

The result is that Salesforce doesn’t always reflect what’s actually happening with the customer. A manager might look at an Account and see a few old notes, outdated fields, or an Opportunity that hasn’t been updated in weeks. Meanwhile, the sales rep has had several meaningful conversations that could completely change how the company approaches that relationship.

This is an area where I think AI and automation can make a real difference. Rather than asking sales reps to spend more time entering information, we can start capturing information from the conversations they’re already having.

The Information Is Already There. It’s Just Not in Salesforce.

Meeting recording and note-taking tools have become incredibly useful. Products like Granola can capture meeting conversations and generate transcripts and summaries, giving sales reps a record of what was discussed without requiring them to take detailed notes during every call.

That’s a big improvement, but it creates another challenge.

If the transcript and summary live only in Granola, the information is still disconnected from the CRM. The sales rep might know where to find it, but what about their manager? What about the Customer Success team? What about someone who takes over the Account six months later?

We recently worked on a project to address this exact problem by connecting Granola to Salesforce using Make.com.

The initial goal was straightforward: take a meeting transcript from Granola and associate it with the appropriate Account in Salesforce. Once we proved that worked, we could start doing more interesting things with the information.

Using Make.com and AI to Update Salesforce Automatically

The next step was to make the transcript useful without requiring someone to read the entire conversation.

We used Make.com to retrieve meeting information from Granola and process it through an AI summarization step using Claude. The goal was to produce a concise summary of the conversation and write that information back to the corresponding Account in Salesforce.

On the Salesforce side, we added a long-text field for the meeting summary and a separate date field to indicate when the summary was updated.

That way, someone opening an Account could immediately see a summary of a recent conversation without having to search through another application or ask the sales rep what happened.

We also wanted the process to run automatically rather than relying on someone to push a button after each call.

During the rollout, we tested the integration, enabled recurring synchronization, and used Salesforce reporting to validate that summaries were being populated on the expected Accounts.

What I like about this approach is that it doesn’t require sales reps to change how they work. They continue having conversations with customers, and the integration helps make that information available to the rest of the organization.

Getting the Information Onto the Right Record Matters

Of course, capturing a transcript is only part of the challenge. We also need to make sure it gets associated with the correct Salesforce record.

A sales rep might meet with several people from the same company. A meeting could involve contacts from multiple organizations, or the attendees’ email addresses might not match existing Salesforce Contacts.

Ideally, the integration uses meeting attendees and other available information to identify the relevant Contact and Account records. When the match is clear, the information can be associated automatically. When it isn’t, I’d rather flag the meeting for review than risk writing customer information to the wrong Account.

This becomes even more important if we eventually use AI to extract information and update structured Salesforce fields.

For example, suppose a customer mentions a budget of $250,000 during a call. Is that the budget for a particular Opportunity? Is it the company’s annual budget? Is it a preliminary estimate that hasn’t been approved?

An AI tool might successfully identify the number, but that doesn’t mean we should automatically write it into a Salesforce field.

The same applies to competitors, implementation timelines, purchasing decisions, and other information discussed during calls.

My preference is to start by capturing reliable summaries and then gradually introduce structured field updates once we’ve agreed on exactly what information should be extracted and how it should be validated.

Don’t Overwrite Valuable Conversation History

There’s another design decision worth considering: how should Salesforce store summaries from multiple calls?

A single summary field on the Account is useful because it gives someone a quick view of the latest conversation. But if that field gets overwritten every time a new meeting takes place, we risk losing the history of previous discussions.

For a simple initial implementation, a current summary and a date field can be a good starting point. Over time, though, I’d prefer to preserve individual meeting summaries as Salesforce activities or related records, while using the Account field for a concise view of the most recent or most important information.

That gives users both the convenience of a quick summary and the ability to look back at earlier conversations.

It also makes the integration more useful for handoffs. If an Account changes ownership or moves from Sales to Customer Success, the new owner can review the history instead of relying entirely on what the previous rep remembers.

Start with Summaries Before You Try to Automate Everything

Once you have transcripts flowing into Salesforce, it’s tempting to start extracting every possible piece of information.

Budgets, competitors, decision-makers, project requirements, next steps, and follow-up tasks are all potential candidates. AI makes it possible to identify much of this information automatically.

But I wouldn’t start by updating dozens of fields.

First, make sure the transcripts are consistently associated with the correct records. Then make sure the summaries are accurate, readable, and genuinely useful to the people working in Salesforce.

Once that’s reliable, identify a small number of structured fields that would benefit from automation. Define the rules for updating them, decide when a human should review the proposed changes, and test the results before expanding the process.

There are also some practical considerations around recording consent, access permissions, and what customer information should be sent to an AI service. Those decisions should be part of the design rather than something addressed after the integration is running.

The objective isn’t to fill Salesforce with as much AI-generated information as possible. It’s to capture the information that helps people make better decisions.

Make Salesforce More Useful Without Creating More Work for Sales

For years, companies have tried to improve CRM data quality by asking sales reps to be more disciplined about entering notes and updating fields. Sometimes that works, but it often creates tension because the sales team sees it as administrative work that takes time away from selling.

Tools like Granola, Make.com, and Claude give us another approach.

Instead of asking sales reps to document every conversation manually, we can capture the conversations they’re already having, summarize the important information, and make it available in Salesforce.

That creates value beyond the individual sales rep. Managers get better visibility into customer relationships, Customer Success teams have more context when they take over Accounts, and the company becomes less dependent on information living in someone’s memory.

I think this is one of the more practical uses of AI in CRM today. It isn’t about building a complicated AI assistant or trying to automate the entire sales process. It’s about solving a basic problem that’s existed for years: important customer information isn’t making it into the CRM.

Before you use AI to do more with your Salesforce data, use it to make the data you already have more complete and useful.

Sometimes the best automation is the one that makes Salesforce better without asking your sales reps to do anything differently.