Customers Want AgentForce Agents Interacting with Google DialogFlow

This week during Salesforce's TDX conference for developers, I spent a few good minutes with Nick Johnston, the company's senior vice president of strategic partnerships and business development. Coming on the heels of Salesforce's big partnership announcement with Google, I wanted to learn how the AgentForce agent will integrate with Google technologies like Gemini and DialogFlow to help make things easier and better for both employees and customers. But beyond any one strategic partnership, I was really interested to see if Salesforce's approach to strategic partnerships has changes over the past few years and what role AI has played, if any, in that change.

In this clip, Johnston talks about the different layers of integration the partnership with Google addresses, including the following:

• Infrastructure Layer: Running Salesforce technology on Google Cloud Platform (GCP) addresses regulatory compliance and customer preferences.
• Data Layer: Integration with Google BigQuery and real-time data access via Google Search APIs enhance retrieval-augmented generation (RAG) capabilities.
• Model Layer: Support for Google's Gemini AI models in Salesforce's Model Garden leverages multimodal capabilities for complex prompts.
• Action Layer: Improved interoperability between Google Workspace (e.g., Gmail) and Salesforce enhances data sharing and agent productivity.
• Contact Center Enhancements: Collaboration with Google's Dialogflow enables seamless handoffs between bots and human agents, real-time transcription, and translation. These features improve contact center efficiency by blending digital labor with human expertise.

Edited transcript

Brent Leary: Take us through a little bit of the big announcement last week. It feels like it was a long time ago, but it was just last week with Google. I know you've had several partnerships before, but talk about the one that you just announced.

Nick Johnston: Google's been a long-standing partner for years. This one we were very intentional with. It's about customer choice, going back to what I was referencing before where customers are giving us feedback. There are certain aspects of the way we're delivering our technology that they have opinions on. And with Google we started getting a lot of very clear opinions and asks at multiple layers of our overall AgentForce strategy.

Starting with infrastructure, there are some customers who really want to run on other options than what we've historically had, and they want to run on GCP. So that's a starting point. Or they want their data to reside in places where Google has a big infrastructure footprint, and if we can bring our technology there it helps them from a regulatory standpoint. So that's like the trust layer, if you will. And then if you think back to what you saw in the keynote today you have the data layer. So we've been working with Google around all things data cloud integrations with BigQuery. 

But now we've been getting questions from customers in the world of RAG-type answers for AI and retrieval augmented generation. What might we be able to tap into given the vast amount of data that Google has access to in real time Wouldn't it be cool if AgentForce could tap into Google Search APIs and be able to pull that real-time information back in the way that AgentForce works?  So that's a component of the partnership.

Also at the model layer customers have been asking for Gemini. And so, as we think about the models that we can both host inside of AgentForce as well as those that we allow customers to bring to our platform, Gemini has some unique capabilities around multimodal: the breadth and complexity of a prompt that they can consume, and customers wanting us to be able to host that inside of our own model garden.

Then the last thing, at the action layer, some of the things you saw in Patrick's keynote, customers want to see us doing more between Google Workspace and Slack. Discoverability of documents and better interoperability between email and Slack is something that our shared customers are asking us for, as well as the ability for agentic experiences inside of Workspace to feed data into Salesforce and then vice versa; for Salesforce data to better inform emails and document generation that may be happening on the Workspace side.

So you're going to start seeing our agents interacting with Gemini agent capabilities at a deeper layer. It's kind of hitting all levels of our Action, LLM, data, and trust layers. That's how we framed up the whole thing and why we were so excited about it. 

Brent Leary: I know the contact center is where a lot of things are going on when it comes to agentic AI. How does that partnership deal with that?

Nick Johnston: I'm glad you mentioned that because I didn't hit on that directly. But it kind of fits in the data layer, the AI layer, and the action layer.

Google has a technology in their customer engagement suite that's really taken off. And we have a lot of customers who are using their Dialogflow experience for front-line agent bots, and we want to be able to have AgentForce agents interact with Dialogflow with effective handoffs. Then in the contact center scenario you may have an AgentForce agent supporting a customer service rep when a call gets sent to them. And you want to make sure that whatever context is in that DialogFlow natural language bot where the initial intake received is properly handed off to the person. But not in just dumping a bunch of text but in a natural language way where the person feels much more informed as they're getting on.

 So you're going to see us doing more between both of our respective channels and how we make the DialogFlow capabilities that they have work better with AgentForce. And then there are aspects of the things they're doing around real-time transcription, real-time translation of audio streams that can just make a contact center so much more effective and productive, especially as you're thinking about the blend of digital labor in the contact center and the human in the loop.  

How that all works together, and the stuff that we can bring with Google's advanced capabilities, and the things we're doing more on the agent experience side, we're pretty excited about it.  It's early for us, but there are a lot of big customers, especially in the retail space, that are pretty excited to see us working together. So you'll see more coming on that in the next couple of months.

Brent Leary: How has all this AI stuff, from generative to agentic, changed the definition it has of your strategic partnerships?

Nick Johnston: It hasn't changed the definition, it's changed the focus. The definition always comes back to who are the companies that matter most to our most important customers that we need to be working with, and everything we do starts with customer demand, so when customers tell us we need better interoperability with Snowflake or we need a better way to incorporate Anthropic models, any of these things are what usually starts the work that we do. It's just changed the conversation; now we need to understand how you guys are thinking about large language models. We have a point of view on large language models, so how do we incorporate our point of view with your point of view and how open are you? That's a lot of how the conversations have shifted. Data obviously has become more important, and openness of data, so a lot of the work we've been doing with our zero-copy partner network and Data Cloud has been a big focus for us. We've had a similar construct with allowing customers to bring their own large language models. If they've built their own strategy with a Google Vertex or an AWS Bedrock, and they want to bring those capabilities to our platform we'll let them swap the brain, if you will.

The other thing we see a lot of right now is our customers are asking us more than ever to give them direction. I'd say early on in some of these strategic partnerships we were working on they wanted to understand them and then they would make their own decisions on who they'd use for what. I think because there's just so much activity right now with all things agentic AI they want to know how to do it or what your recommendation is, which I think makes these partnership discussions more interesting than ever because you really have to get to the crux of agreeing with the partner on how you're going to talk about your shared value prop not just letting the customer decide. So it's been a fascinating part of the business to work in.