Improving Sales Performance with Artificial Intelligence

Startup came out of stealth mode recently, announcing a $6 million in Series A funding. Gong’s cloud-based artificial intelligence software helps improve the performance and productivity of B2B sales teams by analyzing their conversations with prospects and providing guidance about the best way to close a deal.


Amit Bendov, CEO, Gong

Says Amit Bendov, Gong co-founder and CEO: “There are probably a thousand books on successful selling on Amazon—and they are all based on anecdotes.” Instead, Gong brings science to the art of the deal, augmenting experience and intuition with facts and figures drawn from analysis of big unstructured data—all the conversations conducted by all members of the sales team.

Over the last 20-plus years (most recently as CEO of fast-growing business intelligence startup Sisense), Bendov has developed a number of successful sales teams, all selling remotely. “People do not expect to see you anymore,” says Bendov. Even if you meet with a customer onsite, more often than not they will start a WebEx or GoToMeetings session with their remote colleagues, as a B2B sale today involves 5.4 decision makers, on average.

In addition to the distributed nature of today’s decision making process, the economics of remote sales or ”inside sales” make it superior to traditional in-person “field sales.” The latter approach allows a sales person to see around 5 customers per week. With inside sales, the same sales person can interact with 25 to 30 customers and potential customers in the same time period, observes Bendov.

Because of these advantages, inside sales is becoming the preferred approach to selling in many enterprises today. While being more cost-effective than field sales, however, it hasn’t addressed the major cost contributor of any sales operations: The hiring of ineffective sales people. For both approaches, the mystery remains—what makes a successful sales person?

“We’ve always hired people in triplets because we knew one of them is not going to cut it,” recalls Bendov. “And we never knew exactly why they didn’t cut it.” Indeed, a 2013 survey by the American Association of Inside Sales Professionals found that “hiring and training people who possess [the right] skills is the top challenge facing inside sales leaders.”

Having identified the problem that needed to be solved, Bendov found the technologists that could solve it, specifically his co-founder (and Gong’s CTO), Eilon Reshef. In a few months, Gong’s engineers have developed an artificial intelligence software package that combines natural language processing (NLP) tools with machine learning to analyze, categorize, and quantify sales conversations and determine what works and what doesn’t at a scale that was not available before to sales management.

“The average B2B sales conversation runs 6000 words per hour, sales people hold around 20-30 conversations per week, and managers are often responsible for 6-12 reps—impossible math if you’re trying to gain a deep understanding of all those calls,” says Bendov. “Gong provides visibility so sales executives and reps can understand what’s happening on every single call.”

With the permission of both sales people and the customers, calls are recorded and stored in the cloud (as both audio and video files). Gong goes through a first-pass machine transcription of the calls and then applies NLP to work around the inaccuracies of the initial transcription and deal with the complexity of human to human interactions. In contrast to the relatively straightforward human to machine interaction in the case of Siri or Amazon Echo, the sales conversations Gong is tackling are free flowing, multi-party, and riddled with interruptions and speech fragments.

The text of the transcription is categorized by pre-set topics such as “budget,” “timeline,” “pricing,” “competitors,” and “urgency.” The analysis of the conversation shows which topic was discussed, when, and by whom, how much a sales person is talking vs. listening, which questions are asked, and the objections raised. “In 30 seconds, you can understand what happened in a 30-minute call,” says Bendov. In addition to the pre-set topics, after about 2000 hours of conversations, Gong runs a machine learning algorithm that identifies topics specific to each conversation.

Looking at the specifics of the call for each topic helps sales people and their managers grasp quickly the essence of the conversation or remind themselves of what was discussed or where the prospect is in the sales cycle. “Sales people like it because this is like meeting notes on steroids,” says Bendov. When preparing a proposal, the sales person can quickly check how many users are in the prospect’s company, what was promised on the call, and what action items were discussed by simply searching the call’s transcript. Another example of the value of capturing the text and context of the conversation, is that it allows for a quick understanding of how serious is the customer—with clues such as too many questions about pricing or too few, or whether a competitor was mentioned.

Finally, analyzing aggregate data across thousands of calls and correlating it with data about closed deals from a CRM system, allows Gong’s customers to improve their playbook and sales pitch: They find out the patterns of a successful sales conversation and the bad habits that should be eliminated (see “Pitch to listen ratio”); they discover the presence of competitors and how it changes over time, helping them adjust the product and how it is sold; they learn what’s best to mention at different stages of the conversation–for example, top sales people talk a bit about the budget in the beginning of the conversation but a lot more towards the end rather than make it a substantial topic of conversation early on (see “Frequency of Budget Discussions”).



Discovering what’s trending and what conversation patterns make a difference apply to any type of human to human interaction. This means that Gong’s software could be expanded to include text-based conversations and applied to other domains such as customer support and call centers, job interviews and recruiting, and marketing interactions.  “It’s a massive market opportunity,” says Bendov.

Originally published on