The last few years have seen generative AI capture the hearts and minds of the tech community and the broader world. Still, the next few years may be defined by the next generation of artificial intelligence, as “agentic” AI is increasingly catching stream.

The exact definition of agentic AI seems to depend on who you ask, but it primarily revolves around AI gaining the ability to access more sources, cobble together workflows, and automate to move beyond presenting a simple report or generated image to the point where it can take action on behalf of the humans providing the prompt without their further involvement.

“It’s a framework of agents talking to computer systems, agents talking to other agents, interpreting data and taking action on it,” said Rob Barton, CTO of Cisco Canada. “Agentic is going to rise very rapidly in the next couple of years.”

The user prompts their agent, who gathers information from various sources to solve a problem and presents its findings based on the data and information it encounters. Ultimately, in some situations, the agent should be able to act on its conclusions and solve problems independently.

Barton sees the technology gaining use in four distinct steps:

1. Knowledge base information;
2. Asking questions about systems;
3. Troubleshooting and solving problems, and finally
4. Automation.

He said customers are already actively planning solutions in the first two categories, with the latter two likely to follow behind.

The goal would be to ask the agent something like “Why is my app so slow right now,” and have the agent go out and query the application, the network, and a variety of other places that might have clues — service management or IT management software, DNS, systems logs, etc., and present its findings from the information gathered from all of those sources via API calls.

“Even in the early days, it will be able to solve problems and pass the data back to us intelligently,” Barton said. “Customers want to have this kind of solution to complicated problems.”

But as we’ve seen in the generative AI wave, there are concerns. The security of these agents and the guardrails around who can access them and what data the agents can work with will be even more critical than in the generative wave. These bots will need access to at least systems-level information and possibly key data that businesses want to mine and analyze for as much value as possible. 

“These bots are going to have to have access to backend systems, largely unfettered,” Barton said. “And they may be using very sensitive information about systems and people. We need to understand how it’s handling data, where it goes, how it’s encrypted, and who is asking.”

As solutions become more able to take action themselves—say, going beyond presenting a report on why an app is underperforming and making the changes needed to make the app experience smoother—they will need to be watched even more closely, particularly for those Rumfeldian “unknown unknowns” that may arise.

“We may find situations where there are dependencies elsewhere, and the bots do things we don’t even realize, and that can create unwanted disruptions,” Barton said.

Barton said that solution providers should expect to see companies like Cisco bring agentic frameworks to market. These frameworks will help companies manage their offerings in this kind of environment and ask the big questions about their data. Of course, crafting those questions is likely to become a big services opportunity for today’s and tomorrow’s solutions providers.

“It’s a very solid opportunity for professional services,” he said. “Whole new industries will be built around this.”

This technology is likely to impact solutions providers’ businesses. For as long as managed services have existed, “automate as much as possible” has been the rallying cry, and the number of systems a single admin can manage has been a key metric for profitability. A smarter automation layer should allow MSPs to offer new and more services to their customers.

“As it helps me manage new things and solve problems much faster, I’ve got a carrot to show [customers] how much more I can do without much uplift,” Barton said. “The opportunity for them to grow their footprint with their clients is massive.”