AI is supposed to be a competitive differentiator and your sales organization doesn't want to get left in the dust. So like many others, you purchased an AI sales solution that promised to transform your sales motion. You spend countless hours implementing the tool and training your team and wait for it to live up to all the promises you heard throughout the purchasing process. Then you get a panicked message from one of your reps. Uh oh, they accidentally sent an email generated by the new AI tool to a VP referencing the wrong meeting and details about a completely different account.
So, what went wrong? The new AI tool has been matching your sales reps activities like phone calls, emails, calendar entries, and meetings to the wrong accounts and opportunities in your CRM. And AI is only as good as the data it has available. Inaccurate and incomplete data means your AI outputs will be generic or, in the case of the example above, flat-out wrong.
Fueling AI with “good data” is not just about collecting activity information. Every sales AI vendor will tell you they can do this and many of them have impressive demos and a great UI to “prove” it. But, the most important part of sales AI is actually happening behind the scenes with activity matching, where data is organized and categorized. It’s what turns raw data into clean data.
Great sales AI requires correctly associating all sales touch points (aka rep activities) with the correct account and opportunity in your CRM so that it’s A. accurate and B. easy for AI to use. This process is called activity matching and effective sales AI tools do it extremely well. But it’s one of the most challenging parts of the process for AI vendors to get right.
When reviewing the health of a deal or identifying potential risks, sales managers will seek answers to questions like:
The answers to all of these, “What happened?” questions are contained in those reps’ activities. The old way of getting these answers was done by reps manually entering data into the CRM or sharing anecdotes during deal reviews.
Now, automated activity capture ensures that all of your reps’ activities and the people they’ve interacted with are captured automatically. That’s the easy part.
The hard part is taking all of that activity data and properly matching it to the correct accounts and opportunities in your CRM.
The activity matching process has two key steps:
Once both steps are completed, your sales AI tool can start feeding all of this data into its algorithms to get accurate and real-time insights. However, if step two is not executed correctly, you could end up with activities attached to the wrong accounts or opportunities. For instance, an important meeting with a new executive may not be accounted for or a sales manager might walk into an important meeting armed with the wrong history of the account or opportunity.
Without accurate activity matching, you risk working with incomplete or incorrect data when analyzing deals or the performance of your team. This compounds into incorrect insights or even hallucinated outputs from your AI tool, leading to less effective and ill-advised decision-making.
“The devil’s in the details. It sounds great to capture activities in context but if you don’t do it properly, you end up with a bunch of confusing data that doesn’t make sense. The accuracy of being able to do that with automation was really important to us, and we felt that People.ai had the best solution in terms of how it was being approached.” - Esther Friend, VP of Sales Efficiency and Transformation, Five9
Nearly every sales organization is searching for high quality AI solutions but it’s hard to know how to differentiate an average AI product from an exceptional one. Many of you have tried AI point solutions for forecasting or conversation intelligence or a public tool like Chat GPT and the generic and unhelpful outputs generated by those solutions are just not cutting it. That’s because those AI tools are fueled with inaccurate or incomplete data.
Accurate and trustworthy engagement data depends on a sales AI tool's ability to match accurately, even when CRM setups are complex and CRM hygiene is poor.
Many AI tools rely on good CRM hygiene and complete activity data entered manually into the CRM by all reps, which is extremely uncommon. The problem becomes even more pronounced when a company’s CRM is heavily customized. Standard activity matching techniques lean heavily on standardized go-to-market (GTM) practices. When an organization customizes their CRM, many revenue intelligence providers cannot utilize this data because it doesn’t fit into their rigid parameters.
When you’re evaluating sales AI vendors, ask them the following questions about their activity matching capabilities. The goal is to understand their technology to ensure they can perform high-quality matching, despite poor CRM hygiene and complex CRM implementation.
Q1: Can your tool automatically create net-new contacts and domains when they are not already in the CRM?
Why this is important: If your sales team is doing a good job, they will be spending time with new contacts all the time. If the AI tool cannot automatically create new contacts or domains in your CRM, data from those activities will be lost or associated with the wrong contacts/accounts/opportunities.
BONUS: When a tool can create net-new contacts and domains in your CRM, it is actually improving your CRM hygiene. By filling in missing data, it facilitates future activity matches, leading to better revenue intelligence in the long run.
Q2: What information about activities and your CRM do you use to match?
Why this is important: You can gauge how sophisticated AI technology is based on how comprehensive this list is. One of the signals to look for in a great AI tool is Natural Language Processing (NLP). NLP is used to extract content from activity text that can be used to find the relevant account or opportunity.
Q3: How does your system differentiate between similar CRM records when finding a match?
Why this is important: When there's an activity that might match to multiple CRM records (e.g. multiple opportunities in an account), can the system use all of the signals to decide between them to find the best one? A rules-based matching system helps the tool select the most specific account or opportunity match based on multiple categories of signals (ideally 10+). This advanced method allows the tool to make accurate matches even when the available data is less than ideal.
Activity matching is the top factor determining whether AI is getting good data vs bad data. By accurately capturing and associating data, you can gain crucial insights into the health of your deals and make more informed decisions. A tool with great activity matching capabilities can be a testament to the power of AI in revolutionizing sales strategies and driving success.