GOVERNANCE

Trust, scale are front and center for agent adoption

Nat Rubio-Licht: What are enterprises getting right, or wrong, in their deployment of agents into the workforce?

Madhav Thattai: Everyone has correctly assessed that data is really critical for agents. I think that understanding and that realism about, if you want the agent to be effective, it has to have access to the right structured information … Everybody understands that the data question is really important.

The second thing that people are understanding more is consumer experiences and expectations are evolving significantly. We have entirely new channels, modalities, ways in which we can interact with technology right now. Accelerating these user experiences, whether it is internally or externally, is really critical. 

The third is you really don't get anywhere without trust … We work with customers in regulated industries, customers with really, really deep and complex systems. And so the trust question is really important.

Rubio-Licht: What are the most prevalent concerns for CIOs about agent adoption?

Thattai: Number one, they care about trust. Where's the data? How are we controlling it? How are we governing it? Number two, they care about how we are designing and building these agents. How are you going to connect up the right data to make that agent effective? What's the user experience feel like, if I want to do it in text, if I want to do it in voice, how is it going to feel? And so that design, I think, is really front and center. 

The third thing that CIOs really care about is scale. Take a company like Falabella, a huge retailer with 40 million customers, and they launched our experience on WhatsApp. They went from 40,000 interactions a month to five times that. When you think about that kind of scale, a CIO is going to ask: How are those agents performing? What is the analysis I need in order to know where I need to improve the agent? So we call this observability. If I'm going to put this on a massive channel where it's going to reach a lot of customers, I want to understand what that scale and that performance is.

TOGETHER WITH CODER

For Platform Engineering, AI Is Everywhere

Think we’re exaggerating? Think again. Engineering orgs around the world are embracing AI, with over 89% of surveyed platform professionals using it daily. But adoption isn’t the issue. Scaling is.

Most AI efforts live in silos, individual experiments instead of an org-wide strategy. To dig deeper, Platform Engineering surveyed more than 240 platform professionals to uncover what’s working, what’s not, and what it takes to move from experimentation to enterprise impact.

The result? The State of AI in Platform Engineering, an in-depth report packed with insights, benchmarks, and strategies to turn AI into a competitive advantage.

Get your free copy today.

PRODUCTS

Agents take root in front-end processes

Rubio-Licht: What kinds of agentic use cases are actually seeing adoption? 

Thattai: If I were to give you a few examples, at Dreamforce, we talked a lot about William Sonoma using Agentforce for customer service kinds of use cases, and that's a very popular use case. But William Sonoma is also using Agentforce at the front end of their experience to introduce people to their products.

Adecco interacts with 600 million candidates who apply for jobs every year. They're using Agentforce in the candidate qualification and matching process. So that's kind of at the front end of their cycle. 

We really think of these as the transformation of the customer experience, whether it is in marketing, whether it's in sales, in commerce, in service. We think every touch point in that customer experience is going to evolve quite a bit.

Rubio-Licht: Which ones aren’t getting as much traction? 

Thattai: I don't think there are significant anti-patterns yet, but let me give you examples of where we think technology has to evolve to unlock more capability. Let's take the Adecco example. That agent is not just answering questions based on some content. That agent is actually guiding somebody through a qualification process. That example is really important, because there is a mix of deterministic steps, but there's also rich natural language interaction. 

You want to marry those two things. LLMs give you incredible experiences, they are flexible in their understanding of language. However, they're not good at deterministic, declarative process execution. You need both in order to do that right. 

Rubio-Licht: How are employees themselves taking to agentic, digital coworkers? 

Thattai: We started with the employee experiences, and there were a lot of lessons that we learned along the way. Our customer service experiences, where we give employees the ability to answer questions while they're with a customer much more rapidly, to be able to summarize and contextualize the learnings from that interaction more quickly, those are some of our highest adopted experiences right now. 

The really important thing for the employee experience is it has to be in the flow of work. If you create an agent that someone is working in service cloud or in sales cloud, and they have to go somewhere else to engage with that agent to get information, it's probably not going to be that effective.

TOGETHER WITH CODER

Why Do Most AI Efforts Plateau?

It’s the trillion-dollar question, and solving it is the key to unlocking serious competitive advantage. But where do you start?

Start here. In The State of AI in Platform Engineering, over 240+ platform professionals were surveyed to understand where and why most AI efforts stall and how to break through.

Inside, you’ll find what separates high-performing teams, what an AI-native platform really looks like, and how to turn AI from a one-off experiment into a strategic engine.

Download your copy now.

BIG TECH

Inside Salesforce’s agentic strategy

Rubio-Licht: How is Salesforce adopting agents internally? What sorts of returns or gains are you seeing? 

Thattai: Our own agentic transformation is a top-level company priority for a couple of reasons. One, we believe this is how organizations will be constructed in the future, which is humans and agents working together. Two, we believe it's actually transformative to the company in terms of how we execute towards our customers today. And three, doing this internally is an incredibly valuable lesson in how to build these products. 

Take help.salesforce.com. I think they just crossed 2 million agentic interactions on that website since it was launched. Putting that experience out there, and there were some ups and downs, there were some bumps along the way, but we got a lot of feedback from our customers, which was really, really helpful. That taught us a lot about how to build agentic experiences. 

Rubio-Licht: Has Salesforce faced any challenges in adopting agents? 

Thattai: We face the same challenge that our customers have faced, which is, if you don't have a clear goal and a KPI and an outcome in mind, then you're going to do AI for the sake of AI. That is, I would argue, the most critical learning over the last 12 months. And that was the same for us.

We used to have this slide that had all of our Salesforce internal agents, in customer service and sales and technology and HR, across the company. That slide used to say, “We have 120 agents that we are experimenting with.” What I love is that, if you look at that slide today, it will say we have 40 agents we're experimenting with. But those 40 are really, really valuable. Those are the ones where we've determined are actually going to be further.

LINKS

• Microsoft signs agreement disavowing religious bias following conservative pressure

• China, US companies partner on UK robotaxis

• Privacy startup OneTrust is exploring a private equity sale 

• Maryland to deploy Anthropic’s Claude to help residents apply for Medicaid, food aid

• SourcePilot: An AI-powered personal text editor. 

• Sendr: An AI sales assistant to personalize outreach at scale.

• Snowflake: AI Applied Scientist, Code Intelligence

• Google DeepMind: Research Scientist/Engineer, Security Classifier

A QUICK POLL BEFORE YOU GO

Please give us feedback on what you thought of today’s weekend edition! This is a new format for us, so we really appreciate you taking the time to share your thoughts. Use the poll right below!