Salesforce and HubSpot just wrapped their big annual conferences, and both of them made one thing abundantly clear in their keynotes and product announcements:
They’re both all-in on AI agents.
During Dreamforce, Salesforce introduced its Agentforce platform, which it says allows you to “build and customize autonomous AI agents to support your employees and customers 24/7.” Basically, you can create autonomous applications that execute tasks using Salesforce products and data. (As examples, Salesforce touts some pre-built agents you can access in Agenforce like a customer service agent, SDR agent, and sales coach agent.)
During INBOUND, HubSpot announced Breeze, which it calls its “AI engine.” Breeze consists today of over 80 AI-powered features across every HubSpot product. This includes Breeze Copilot, which is an AI sidekick that uses generative AI and your CRM data to help you work better and faster in HubSpot. Breeze Agents are also a core component of this AI engine, which are “AI-powered experts designed to automate workflows,” according to HubSpot. These include a Content Agent, Social Media Agent, Prospecting Agent, and Customer Agent.
HubSpot co-founder and CTO Dharmesh Shah also introduced Agent.ai during his keynote at the event. Agent.ai is “a place where you can find and connect with AI agents,” according to a letter from Dharmesh on the site. The site lists out dozens of AI agents that perform a range of marketing and business tasks. (The site says you can then “hire” these agents to perform these tasks, which means spending credits to use them.)
Make no mistake: The hype around AI agents has now reached fever pitch. And that begs the question:
Where are we actually at with AI agents in the enterprise, once you cut through all the hype? And what does all this mean for marketing and business leaders?
I got the answers from Marketing AI Institute founder and CEO Paul Roetzer on Episode 116 of The Artificial Intelligence Show.
The first question to ask might seem obvious, but really isn’t:
What exactly are AI agents?
“Agents are sort of taking on different definitions that it's not really clear what exactly everyone's talking about,” says Roetzer. In our view here at Marketing AI Institute, we’ve typically thought of AI agents as generally capable AI systems that can take autonomous actions for a user.
But it will probably take several steps over a few years to get to “true” AI agents, a timeline we discussed in Episode 87 of The Artificial Intelligence Show. That starts with narrowly capable agents, not generally capable ones, with varying degrees of autonomy and nothing resembling full autonomy.
In a recent interview on the No Priors podcast, Bret Taylor, board member at OpenAI and cofounder of Sierra, an AI agent startup, shared how he looks at defining AI agents with a little more nuance. He breaks AI agents up into three categories:
Alright, so everyone is defining AI agents in slightly different ways. We’re largely not at the point yet where fully autonomous general agents work (or work well) yet. And there’s nuance to how we can look at the different capabilities of the agents that exist today.
So where are we actually at today with AI agents?
“I think what we’re seeing with HubSpot and others is this very early version of these very specifically tuned agents designed to do very specific tasks,” says Roetzer. “They’re not generally capable of doing many things. They’re trained to be a sales rep bot or a coding bot or a service rep bot.”
In other words, these new products built into sales and marketing platforms don’t yet seem to be the generally capable agents that OpenAI lists as Step 3 of 5 on the road to artificial general intelligence (AGI).
There’s nothing wrong with that. It’s still very, very early and AI agents hold great promise. But it is worth noting as the hype around AI agents grows.
Despite being very early, we need to start thinking about the very real impact AI agents will have on marketing and business now, says Roetzer.
We don’t have generally capable and fully autonomous agents that work well in major software platforms just yet. But all signs point to AI agents growing in both their capabilities and their reliability in the near future.
“They're going to start getting really reliable next year, especially the more narrow they are, the better you train them, the more data they have,” says Roetzer.
As they do, they’re going to change things.
“HubSpot, Salesforce, Google, Microsoft, [and anyone else building agents] none of them are going to talk about these agents being replacements to workers,” says Roetzer. “They're all going to talk about them being unlocks and enhancements and augmentations.”
Yet that’s not the full story—and everyone at these companies knows it.
“Let's get one thing clear: agents will replace people,” says Roetzer. “This is regardless of the marketing messaging that people use.”
As AI agents become increasingly capable at performing tasks within your existing software, you will simply need fewer people to do the work you are doing today. (HubSpot’s page about Breeze on its website begins to hint at this, saying its agents can “Increase your team’s productivity without expanding your staff.”)
But it also may not all be doom and gloom, says Roetzer.
"I'm also highly confident people will have new or evolved jobs training and managing these agents,” he says. “Because who better to build them and to know how they're performing and to improve them than the people who have domain expertise and knowledge in that space?"