The word on agentic AI’s ability to deliver on its promises is: so far, so good. With caveats.
A majority of 300 senior executives adopting AI agents, 66%, say they’re delivering positive results in terms of productivity, a recent PwC survey suggests. But, let’s face it — all systems deliver some degree of productivity. What executives need is that extra edge that delivers extreme competitive differentiation.
At this point, few AI agents are ”transforming how work gets done,” the PwC report’s authors state. “Many employees are using agentic features built into enterprise apps to speed up routine tasks — surfacing insights, updating records, answering questions. It’s a meaningful boost in productivity, but it stops short of transformation.”
The biggest barrier isn’t the technology; “it’s mindset, change readiness and workforce engagement,” the PwC authors conclude,
Mahe Bayireddi, CEO and co-founder of Phenom, which offers agents for HR tasks, agrees this is where the challenge lies. I had the opportunity to sit down with Bayireddi at Phenom’s recent user conference in Philadelphia, where he pointed out that for AI agents, context is everything.
“I think there’s a lot of learning in this whole process,” he said. “There are no experts dynamically saying how they can handle AI agents effectively.”
Agents can bring up productivity almost by 20-30%, if they use it in the right format, do the change management effectively, and use the data in an engagement format,” Bayireddi continued. “The point is how do they make it fly, how do they manage the change management.”
AI agents and the data they consume need to be domain-specific, and will vary industry to industry, company to company. “The data at the universal level is actually complex,” he said. “The nuance of a context and that nuance of personalization is very critical for AI to work. It can’t be too general.”
The rise of agents advances generative AI to more practical levels. Once put into place, agents can be “baked into the workflows,” he said. “Up to now, everybody has had to go to ChatGPT and ask a question and get an answer. It’s not the way how people work.”
The emphasis needs to be on addressing the nuances of functions and processes to be automated with agents. “That has to manifest in an effective format with a context,” he said. “That can only happen with an agent being effective in a department.”
Bayireddi doesn’t see agents as a threat to jobs, but they will change the nature of jobs. “There are new jobs which are going to come up because of agents. There is a new work which is also going to pop up because of agents. Skills is one thing, but also the work will change and the jobs will change.”
Don’t settle for too little when it comes to AI agents, the PwC authors advised. “Companies that stop at pilot projects will soon find themselves outpaced by competitors willing to redesign how work gets done. We see few companies moving early to define the future, building new operating models that integrate and orchestrate multiple AI agents. Fewer than half are fundamentally rethinking operating models and how work gets done (45%) or redesigning processes around AI agents (42%).”