How agentic AI should reinvent work beyond automation

Agentic AI refers to systems that can plan, act and collaborate across workflows, redesigning how entire processes and resulting outcomes — not individual tasks — can be completed, alongside humans. But many companies are chasing the same incremental productivity gains with AI, through new toolsets rather than confronting the need for new mindsets and skill sets. Truly differentiated AI requires a wider lens that incorporates cultural elements, including people and processes:

• Executives often rely on those closest to a process to determine how best to optimize it with AI. But bolting the technology onto outdated processes will deliver outdated results. Hybrid teams of subject matter experts and technologists — uniting the best of your insiders and outsiders — should start from a blank slate, often in greenfield projects, working backward from the desired outcomes.
• There is a widespread belief that AI will hollow out workforces when human knowledge, experience and creativity are wellsprings of competitive advantage. Accelerating productivity is great, but to truly do different and innovative things, not just the same things differently, human and digital workforces must be seen as two sides of the same coin rather than prioritizing one at the expense of the other.

The way forward is through reinvention. In real estate, slapping on a fresh coat of paint can look like modernization, but it does not address an aging foundation. It is the same with adding AI to workplaces. People, process, data, technology, customers, operating models, governance and more must be assessed and addressed to construct an AI-native organization that achieves growth for tomorrow, not just savings for today.

Processes: from initial steps to giant leaps

Over time, organizations have built up processes over decades. These processes have been steered by people-related limitations, technical landscape debt, regulations (that may no longer exist) and legacy managers requiring transparency into how everything gets done. As you examine these processes, it is easy enough to see where AI can be injected into something like order-to-cash, record-to-report or inspire-to-buy to automate or accelerate select parts of them. What is harder to see — but imperative to figure out — is how those workflows can be entirely rewritten to run natively on AI — built in, not bolted on.

At many companies, product innovation can be an 18-month process of gathering and assessing consumer feedback and then bringing a new product or service to market to fulfill demand. You likely have a defined process for this that you can recite from memory. But AI can constantly monitor customer chatter online, develop options worth testing and then get virtual feedback from a range of personas. Perhaps 56 steps need to be only 14 and 45 hours of work can be 45 minutes and going to market takes mere weeks — but that is not possible just because yesterday’s processes were designed for the speed of yesterday’s technology.

What to do

Start with the input and the output, and then everything in between is fair game to eliminate, replace, enrich or embolden. When you examine the cluster of steps, different agents can be built to complete them — and those agents can talk to each other (with some degree of human oversight, if warranted). And new processes also give rise to new jobs and new controls. AI-native business units can test and validate new ways of working before they are cascaded into the broader organization.

Jobs: use humans better

With AI, you can achieve what was never thought possible with humans. That has prompted doomsday scenarios in which human workers are considered deadweight on balance sheets and therefore expunged. I have seen firsthand how that is not the case: in masterclasses, we get executives to put their hands on the keyboard and use AI solutions to create basketball teams, new snacks, sustainability solutions and more. In these events, we found that most of the outputs are an undifferentiated mass: similar colors, similar logos, similar slogans.

This highlights how creativity and innovative thinking from humans are becoming more important, not less. When workflows are rethought for agentic capabilities, humans gain free time that they should devote to these activities. But companies are instead largely opting for a “human in the loop” model used to validate automated work, which is not creative and is not transformative for bottom lines.

What to do

The EY Agentic AI in the Workplace Survey found that 84% of desk workers are eager to embrace agentic AI in their role because the technology can reorient their workday around what is engaging about their jobs, not the rote tasks they must muddle through. Process redesign codifies that reorientation, meaning that job descriptions will change but headcount does not necessarily need to decline — and could even climb. Many workers also want to embrace the opportunity for upskilling.