ForwardThinking |
|
As AI agents take center stage in 2025, many questions remain |
It’s the time of year where everyone is making predictions, and I’ve got an easy one: Agents are going to dominate in the enterprise in 2025 as companies begin to try and understand how to best incorporate them to improve workflows. |
Just this week, in fact, Salesforce announced Agentforce 2.0, perhaps a bit soon for such a substantial sounding update, just a couple of months after releasing Agentforce 1.0. Regardless, the company is fully committed to an agentic push in the new year, so much that it made this announcement just a week before Christmas, a time when news tends to get lost in a deluge of parties, presents and other holiday distractions. |
Agents are the next step in workflow automation, and a big leap forward from robotic process automation and no-code workflow tools we were talking about just a few years ago. As companies seek greater efficiencies, agents offer great potential to move work across operational boundaries and complete a series of tasks with minimal human involvement, acting as a kind of digital employee, a concept that Salesforce has been fully embracing as of late. |
At Salesforce’s earnings call earlier this month, CEO Marc Benioff was trumpeting his company’s position in the coming agentic world. “Salesforce has become, right out of the gate here, the largest supplier of digital labor, and this is just the beginning.” |
It remains to be seen whether it’s Benioff bluster or the truth. Clearly Salesforce is not alone in this space as just about every company has started talking about agentic AI (including boldstart portfolio companies Kustomer and CrewAI). |
But it’s also worth noting that the coming agentic wave comes with tons of questions. The first step would seem to be a common definition. It feels like everyone has their own idea of precisely what an agent is and what they can do. We could not only use a standard definition, while we’re at it, we could use some interoperability standards, and a basic tech stack on top of which all agents will be built in the future. So far, I’m not aware of any standards body working on these issues. |
Another major issue is how to manage and control multiple agents from multiple vendors operating simultaneously inside an organization. Jon Turow, a partner at Madrona Ventures, who has written extensively about agents, expressed surprise at AWS re:Invent this month that the company didn’t announce a tool to manage multiple agents from multiple vendors. To this point, nobody seems to have done that. |
Speaking at the Axios AI+ Summit this week, Sierra co-founder Clay Bavor said, “The solution to many problems with AI is more AI.” Writer CTO and co-founder Waseem Alshiek expressed a similar sentiment in an interview with me at the Writer AI Leaders Summit earlier this month. |
For the short term, Alshiek says human programmers have to define the agents, which he describes as "a prompt running a bunch of [code] running in the specific models,” but he thinks pretty quickly the AI will begin to do a lot of this work. “We're big believers that at some point, not in five years or six years, I think literally next year, those workflows can be completely suggested and run by AI end-to-end,” he said. |
But he acknowledged that there will need to be some new human roles developing around this too. “I think we're going to start seeing something like an agent supervisor or workflow or human AI supervisor title, someone who has expertise in the business side, deep expertise in those processes, but can also [work with] the AI and make sure the agents are doing the right thing.” |
As agents begin to develop more fully in the coming year, we will need some answers to these questions. Perhaps Bavor and Alshiek are right and that answer is to throw more AI at these issues and they will solve themselves. Perhaps it requires a set of standards for all agents, or a new kind of agent orchestration software to help manage agents as they work inside large organizations. Whatever the answers are going to be, we are about to find out. |
|
Just a heads up that we are closed from December 23-January 3 and won’t be publishing during this time. The FastForward newsletter will resume on Friday, January 10, 2025. |
I want to thank everyone who has subscribed to the newsletter thus far. It’s been so rewarding and gratifying to get this off the ground. I want you to know that I’m just getting started and there’s lots more to come in the new year. Happy Holidays everyone! |
