Enterprises still chase incremental, not transformational, AI gains

Most enterprises have yet to tap into the transformative power of AI, focusing instead on incremental productivity and efficiency gains that don’t lead to competitive advantages, according to a report from research firm Forrester.

Internal productivity gains from AI remain marginal, not material, as organizations haven’t figured out how to drive more meaningful gains through the technology, Forrester says in its recent Accelerate Your AI Voyage report.

The evidence: 43% of AI decision-makers surveyed by the firm measure productivity improvements gained through AI, and 41% measure efficiency gains, but only 32% tie AI outcomes to profits or revenue.

“Saving 10,000 employee hours might look good on paper, but it won’t cover the GPU bill, let alone drive reinvention,” Forrester analysts write in the report. “This incremental thinking forms the basis of a fundamental disconnect in the promise of AI’s transformative potential.”

Only between 5% and 15% of organizations currently have an effective AI strategy, with the percentage likely closer to the low end, estimates Brian Hopkins, vice president for emerging technology at Forrester. By focusing on productivity or efficiency gains, most organizations miss out on the true power of AI, he adds.

“Efficiency is not strategy; it’s project management,” he says. “You’re trying to make your current processes incrementally better.”

Giving employees a copilot to see what they’ll do with it is not a winning approach, Hopkins adds. “This whole idea that we’re going to incrementally invest in productivity and somehow that’s going to capture the potential that AI offers, it’s a fool’s errand,” he says.

Meanwhile, AI productivity improvements often depend on job cuts on the other side of deployment, he adds.

“The problem with incremental productivity improvements is that in order to realize the benefits that your CFO demands, you have to implement a solution, prove that it works, and then you have to fire people,” Hopkins says. “Do you think those people who will be let go are going to help you do that? They’re not. It’s messy, ugly work.”

The Forrester data echoes another recent survey from AI agent platform vendor Decidr, which found that 40% of US businesses are getting the bulk of their AI value from ChatGPT-style tools, instead of agents or custom AI models.

Old ways of thinking

Other IT leaders also see the problems highlighted in the Forrester research. Many companies are focused on horse-and-carriage-level AI strategies in a world that’s moving to self-driving cars, says Christine Park, chief AI transformation officer at mobile link tracking platform provider Branch.

“This is exactly what happens when the market moves faster than the operating model,” she says. “Leaders are optimizing for narrow efficiency inside functions instead of rethinking how the work itself should fundamentally change.”

Productivity and efficiency gains won’t significantly move the needle for most organizations, she adds. True AI transformation won’t enable individual functions, but it requires coordination across all workflows.

“AI for cost efficiency raises the floor, so what?” Park says. “If it’s just an efficiency play, you’re not going to win beyond short-term gains. If you consider cost avoidance versus efficiency, we can grow without proportional headcount growth, but you need true transformation to raise the ceiling.”

Instead, smart organizations will focus on AI as an amplification of both revenue and the experience of people, she adds. The nature of work is changing, with it now happening in multidimensional workflows instead of step-by-step tasks, she says.

“AI is being treated like a feature when it should be treated like a transformation,” she says. “That means taking a human-centered lens and leaders changing how we train people, define roles, and measure success. AI is a human shift, not just a new tool.”

True workflow transformation

Organizations should look for business-wide workflow transformation, adds Mike Flynn, technology sector consulting leader at professional services provider EY. Many organizations are focused on task-level automation instead of redesigning workflows from end to end, he says.

By focusing on task-level gains, companies add AI tool and computing costs without removing significant work from the system, leading to what Flynn calls “trapped work.”

Organizations need to take an AI-centric approach to all their workflows and try to redesign processes to remove repetitive human work as much as possible, Flynn recommends. Organizations should then add in human intervention when it’s needed, he says.

“If you think about bolting AI onto your business problem, as you continue to add AI, the amount of effort that’s going on continues to increase, versus redesigning your processes such that AI is built into your process,” he adds.

Creating an enduring AI strategy goes beyond rolling out a few AI tools for employees, Flynn says, adding that EY runs clients through an AI value blueprint that takes them through potential outcomes of various AI strategies.

“Firms are realizing that this is not as easy as just enabling and giving people tools that they can do something with and kind of bolt onto their existing jobs,” he adds. “To me, the big thing is thinking about re-engineering your operational processes. It’s a process transformation and a people transformation as much as it is around specific AI.”

Are you ready to make the leap?

Most organizations aren’t yet ready to take the next step, however, suggests Thomas Prommer, former president at design, IT, and AI company Huge. Material use cases such as repricing and supply chain decision-marking require model risk management practices and audit trails that most enterprises don’t have yet, he says.

“Internal productivity is the only use case the organization can actually safely test with current governance,” Prommer adds. “They’re doing copilots because copilots don’t need a model risk committee.”

In addition, the transition from incremental to material gains driven by AI requires someone or something to force a change, such as a CEO, activist investor, or competitive shock, he adds. CIOs rarely can drive the change alone, he says.

However, some organizations have moved past productivity savings because they don’t show up on profit-and-loss (P&L) statements, Prommer says.

“If you save an engineer 90 minutes a day, that doesn’t show up on the P&L — it shows up as, ‘We shipped 15% more features,’” he says. “Boards want a line item. Companies that moved to material use cases did so because they had a single P&L owner willing to stake their number on it.”

Forrester’s Hopkins urges organizations to rethink AI strategies and focus on material changes, despite the difficulties. If organizations aim high enough, they can use AI to enable entire business transformation and find AI uses that drive competitive advantages, he says.

Forrester advises IT and business leaders to focus on four key areas:

• Define the business outcomes and success metrics for their AI initiatives.
• Identify specific use cases for AI deployment aligned with those business outcomes.
• Establish a structure to plan, test, and deploy AI applications.
• Scale AI applications using the power of the cloud, frontier models, and embedded agents.

If organizations take the right approach, they can deploy AI in ways that create real competitive advantages, Hopkins says.

“Strategy is where you apply massive force, based on an insight you have, that gives you strength and weakens your competition,” he adds. “You have an insight that your competitors don’t see and establish a capability that your competitors can’t replicate.”