The consensus on agentic AI’s ability to fulfill its promises is cautiously optimistic: so far, so good, but with important caveats. A recent PwC survey of 300 senior executives adopting AI agents found that 66% report positive results regarding productivity. However, it’s worth noting that all systems typically yield some productivity gains. What executives truly seek is a significant advantage that creates extreme competitive differentiation. Currently, only a few AI agents are “transforming how work gets done, ” according to the PwC report’s authors. “Many employees utilize agentic features embedded in enterprise apps to accelerate routine tasks — surfacing insights, updating records, answering questions. This represents a meaningful productivity boost, but it falls short of true transformation. ” The main obstacle isn’t the technology itself; rather, it is “mindset, change readiness, and workforce engagement, ” the PwC authors conclude. Mahe Bayireddi, CEO and co-founder of Phenom, a company providing AI agents for HR functions, concurs that challenges lie primarily in these areas. Speaking at Phenom’s recent user conference in Philadelphia, Bayireddi emphasized that context is crucial for AI agents. “There is a lot to learn throughout this process, ” he said. “Currently, there aren’t experts who can dynamically guide how to effectively manage AI agents. ” Bayireddi elaborated, “Agents can increase productivity by about 20-30%, provided they are used correctly, change management is handled well, and data is engaged with appropriately. The key question is how to make it succeed and manage the change effectively. ” AI agents and the data they handle must be domain-specific, varying from industry to industry and company to company. “Data at a universal level is quite complex, ” he explained. “The nuances of context and personalization are critical for AI to function properly. It cannot be overly generic. ” The emergence of AI agents advances generative AI toward practical application.

When integrated, agents can be “embedded into workflows, ” Bayireddi said. “Until now, people have had to go to ChatGPT, ask a question, and wait for an answer. That’s not how work naturally happens. ” Focus must be placed on addressing the nuances of the specific functions and processes targeted for automation by agents. “This needs to be realized in an effective format that includes context, ” he added. “That can only be achieved if the agent operates effectively within a department. ” Bayireddi does not view AI agents as a threat to jobs but acknowledges they will alter the nature of work. “New jobs will emerge because of agents, and new types of work will also develop. Skills are one part of this; the work itself and the roles will evolve. ” The PwC authors advise not settling for minimal gains with AI agents. “Companies halting at pilot stages risk being outpaced by competitors willing to redesign work fundamentally. Only a few firms are proactively defining the future by building new operating models that integrate and orchestrate multiple AI agents. Less than half are fundamentally rethinking operating models and workflows (45%) or redesigning processes around AI agents (42%). ”