At a recent executive event, on the heels of a Salesforce Agentforce World Tour event in London, I was struck by how quickly executives could test new business ideas, customer experiences, and automations. workflow. The problem is that you need to import the data into the Salesforce data cloud to derive insights. And you must use the different Salesforce services to act automatically and responsibly.

This struck me as catnip for busy executives wondering why it takes weeks or months to create a proof of concept or years to deploy it. Different types of agents seem to be all the rage today, with different levels of maturity or promise. Salesforce’s first-mover advantage with a practical solution creates a compelling value proposition for consolidating more workloads onto the Salesforce ecosystem.

For example, he played a key role in Tottenham Hotspur CTO Rob Pickering’s decision to consolidate its IT infrastructure on Salesforce. He was tired of telling executives that it would take twelve months to try a new thing. He explains:

I’m a simple farm boy and grew up on a farm. For me, it’s simple, it’s durable, reliable, secure, scalable and supportable. It’s everything you want as a technologist. So the more you can simplify your business systems and the more you can standardize those systems into fewer systems, the better your technology environment will be. I don’t think many technologists would disagree with me on this point, although some may be limited by five, ten, twenty, or fifty enterprise systems they have to work with. I think if you ask them all, they would like to have fewer and fewer of these systems, not more and more.

When you have your data in fewer systems, you can use the power of some of these tools to take action on it. So if I only had a third of my image in one of these systems and I wanted to act on it, I would have to create a really sophisticated integration, have a series of developers do it, pass that data over , make something happen, and maybe return the result. It’s hard work. To be fair, this involves things breaking as it happens. When data lives in fewer systems, you can get the most out of it and do it in a way that delivers the most value as quickly as possible.

The data aspect behind Salesforce’s strategy is also quite compelling. As teams are pushed to try AI everywhere, it is essential to consolidate and improve the quality of their data systems to integrate them into new AI approaches and use cases. And caution is advised, as teams may not want to pay large AI vendors to train better models for their competitors.

This is where Salesforce Data Cloud comes in. It allows businesses to consolidate their unstructured data in one place and organize it appropriately in different formats for more accurate retrieval augmented generation (RAG) queries. This can allow a large language model to execute the query and then forget the data after an interaction. It also connects to the broader Agentforce infrastructure to allow an agent to query a customer’s data without inadvertently revealing information about others.

In a press roundtable, Zahra Bahrololoumi CBE, CEO of Salesforce UK and Ireland, said that Data Cloud is their fastest growing product. It was an almost offhand comment during a longer discussion about something else. But it seemed to me to be a pretty compelling example of the value of helping companies get their data in order.

I dug deeper afterward, as a new service developing with a smaller revenue base would surely grow faster than more established services. The team said Data Cloud is the fastest growing organic product in Salesforce history, with 130% year-over-year growth in paying customers. It also boasts impressive scale, processing over two quadrillion records per quarter.

Data Cloud was also designed to be open to support other data infrastructures. It uses a Zero Copy feature to help customers connect infrastructure investments they have already made with Amazon, Google, Microsoft, IBM, Snowflake and Databricks.

My inbox is currently flooded with vendor news about agents or agents this, that and the other, suggesting that the concept can mean very different things. In the long run, this may promise that multiple agents will negotiate and act on our behalf across many applications and services.

However, the industry isn’t there yet when it comes to accountable and governable agents in the enterprise, says Patrick Stokes, global vice president, products and industries at Salesforce. Today, their goal is to help teams create responsibly managed agents to perform individual tasks and pass them to humans when necessary. When a business is already using the Salesforce platform, this might include presenting all relevant context for a human agent to continue the conversation.

Stokes also explains a little more about how it works today and his vision for the future:

Our goal has been to build the infrastructure or system required to enable agents, within the front office, and in some cases beyond, to enable them to communicate with data and, in many cases, with humans too. , and interact with them in ways that accelerate the way we work within these systems. However, there is some kind of clear future, not necessarily present, but clear in which we expect agents to start talking with other agents. I expect things like protocols to be developed that will be standardized in the same way that the web is standardized around certain protocols, like HTTP and beyond, so that all of this can work. I think we will be at the forefront of that. Many of our partners, like Workday for example, build an HR agent on our platform. We probably want some of our agents to be able to talk to this agent. But I think at the moment we are a little bit in the future, and probably in the next wave. We’ll probably talk about this a lot at next year’s Dreamforce.

Back in April, I wrote about how Rabbit’s big action patterns foreshadow the future of RPA. The CES demo was compelling, and the company achieved ten times expected sales for its agent gadget, which promised to automate interactions with various apps and websites. I was so impressed that I even bought one.

The problem was that it didn’t work. I returned mine a few weeks after reflecting on my lost hope in the colorful clipboard on my desk. I suspect that many agentic things will suffer the same result.

Salesforce’s more cautious approach to building an agent library on top of its platform seems more pragmatic and could evolve over time. And it seems simple enough to get the ball rolling quickly.

Stokes said they recently scheduled a twenty-minute meeting with busy executives from a large company already running numerous workloads on various Salesforce clouds. Executives asked how Agentforce could help them implement an idea they had. Stokes was able to spin up an agent in three minutes, and they spent time confirming that he was indeed working on the company’s data. To be clear, it would probably be wise to hand such a proof of concept to a subject matter expert to resolve any issues.

The fact that such rapid proof of concept is now possible makes me wonder how many of these executives might start to wonder why they can’t yet do the same thing on their other applications and infrastructure.