Salesforce launches AI-powered relationship research as part of its B2B industry pitch

Profile picture for user pwainewright By Phil Wainewright June 9, 2021 Audio mode
Summary:
A new capability offers B2B sales people more information about prospects and their connections - now rolling out as part of the Salesforce investment banking cloud

Einstein Relationship Insights goliath example
(Product screengrab via Salesforce)

Amidst a splurge of new product announcements to coincide with its Industries Summit today, Salesforce has unveiled a new AI-powered assistant that helps B2B salespeople discover connections that will help them close deals faster. We spoke to Jason Briggs, Senior Director of Product Management, about Einstein Relationship Insights (ERI) and how it will augment the reach of salespeople in investment banking, healthcare and other industries.

Unlike other AI-powered systems that attempt to find meaning in massive unstructured databases, Einstein Relationship Management takes its cue from the individual user's browsing activity. Imagine a sales person is researching a specific company and has various browser tabs open as they do so. Based on what's open on the screen, the Einstein tool augments that search and looks for additional relationships. Briggs explains:

The system is going out and running searches on its own, on Google, in news sources, to try and find key information that I would have probably searched for if I had infinite time to Google, mess around, make a bunch of queries and do all that kind of stuff. But I don't, and so it's going to try and do as much of that as it can and bring me back the prioritized best relationships.

So that's the key right here. The key feature that we enable is discovering unanticipated relationships — things you probably wish you knew, but didn't have the time to search for.

Surfacing weak ties in information

The technology comes from Salesforce's acquisition of Diffeo, where Briggs was CEO, in November 2019. It differs from other AI solutions in three key ways, he explains. The first is that it focuses on augmenting the user's existing knowledge, rather than replacing it with a different dataset. Secondly, it's dynamic, responding to user activity on the fly, so that for example, if you return to a specific subject at a later date, it will do further searches in response. Most important of all, it's working with unstructured content. He explains:

It's the loose, connect-the-dots that you don't get from a lot of the other structured databases, but can often be some of the most helpful relationships, simply because they're not out there. They're not structured already, so people aren't using them. So it gives you an extra edge.

This reminds me of the notion of 'weak ties' that were frequently spoken about during the early buzz around enterprise social networking. Often it's those weak links to new spheres of influence or expertise that are the most valuable elements of your network, and which humans — rather than computers — excel at identifying and making use of. It's as though ERI dynamically builds a serendipitous knowledge graph from which the user can then identify the valuable connections. Interestingly, connections found in Twitter are among some of the most useful. Briggs comments:

It's using the unstructured content that it's found, the same way a human would think about it. We don't have a knowledge graph in our heads ... We see two mentions in a webpage [such as Twitter], and that's what digs into our brain, 'Oh, okay. They were on a panel together.' We're exposing the mentions in the unstructured text, instead of trying to build out the full rich knowledge graph.

Therefore this becomes a collaboration between the human user and the AI machine, acting as what Briggs calls "the teammate who's trying to show you interesting relationships that even it doesn't know." The machine isn't very good at classifying the edge cases it finds, but by surfacing them it allows the user to exercise their judgment and determine what's important. Users can drill down to understand why the algorithm chose to surface a particular connection. Briggs says:

It will also show you all the the evidence that substantiated that relationship, so you can see why it's being recommended. That explainability is something that we found is really important to be able to dig into.

Financial services use cases

While there is a generic baseline model that Salesforce runs in the background to identify what people, companies or events are important, the additional learning that's layered on top of that by individual users within an organization remains confidential to that account. It then gets written to the Salesforce CRM system and becomes part of the organization's shared knowledge. Briggs explains:

It's not just a separate process running off somewhere else, it's actually engaging with and also creating new CRM records for you to be able to use in the future, and to support the growth of institutional knowledge ...

Let's say someone else, either instantly, or maybe a year later, two months later, is looking at a different deal where that person is a key relationship. You'll be able to see and open that record and see all the information they gathered with hindsight.

While AI has often been associated with mass market applications — and in financial services has acquired something of a shoddy reputation for creating social media spam — the use case for Einstein Relationships Insights is focused around higher value, B2B sales. Briggs says:

This is not meant for high-volume, low-value type of engagements, where maybe you call 10 people every 10 minutes, and you don't really do much digging. This is for the case where it's a higher value engagement. So you're going to do a little bit of [digging] to go and find out more about them.

ERI is a key feature within Salesforce's Corporate and Investment Banking for Financial Services Cloud, also announced today. Other features in this new industry offering include integration to Tableau CRM to monitor deal analytics, access management to ensure compliance with confidentiality rules, and integrated views of external data sources alongside internal metrics.

My take

This is a great example of artificial intelligence augmenting and working with, rather than replacing, human intuition and knowledge. I've asked Briggs to put me on the waiting list for the journalism edition of this AI-augmented research capability — I can see a lot of value in its ability to quickly surface hidden connections.