Imran Aftab, CEO and Co-founder, 10Pearls—driving AI innovation and creating meaningful opportunities that make an impact.
Hitting an AI plateau isn’t unheard of, but few organizations can break through it easily. To move past it, or avoid the plateau altogether, leaders must look beyond the surface metrics and dig deeper into the very blueprint of process design.
The plateau appears when tools are deployed and usage is real, but ROI and P&L impact aren’t growing. This is where leaders need to look beyond blaming the model and to the root cause: adding AI to a preexisting workflow instead of asking whether that workflow should still exist in its current form.
Breaking The Ceiling
There are organizations that simply add AI as a productivity layer on top of existing workflows, and those that redesign the underlying processes around what AI makes possible. Boston Consulting Group’s data shows the two groups don’t just perform differently; one is miles ahead. Roughly 60% of organizations using co-pilots, bots and automation layers have yet to capture material value at scale, while firms committed to end-to-end process redesign see 80% reductions in cycle time and cost reductions exceeding 60%.
The primary constraint is not AI usage, but whether redesign happens. And, according to McKinsey’s latest State of AI report, high-performing organizations are three times more likely to redesign workflows in depth rather than simply layering AI onto them. Moreover, organizations with a clear redesign strategy see measurable business impact rise by 25 percentage points, compared to about five points from better tools alone.
Defining The Workflows That Should Exist
This is where most transformation conversations start off wrong. Just asking which tasks AI can take over makes space for the assumption that the task is actually worth preserving. The better point to probe is whether a workflow needs to happen the way it’s happening, if at all.
This changes the meaning of AI transformation and moves away from a technology rollout laid over the existing organizational processes. Instead, it paves the way for a redesign exercise that may or may not lead to automation, a faster version of the same process, or cutting out the step entirely because it’s no longer needed amid everything else that has changed. Leaders who skip this part can fall into the trap of automating waste faster, and with a bigger AI spend to show for it.
The Vanity Metrics Trap
Once an organization deploys AI, it’s easy to report on deployment: the number of agents that are live, how many queries get processed and in how short a time frame, the total hours nominally saved. Those numbers move in one direction and look appealing, but they don’t answer all the questions that senior executives care about.
According to more than half of CEOs, AI hype is actively distorting boardroom decision-making. This is a sign that the metrics being reported and those that the board actually cares about are not always aligned. It builds on my earlier point that CEOs must look past pure financial framing to really tap into AI’s value. In fact, firms that capture both revenue and cost benefits from AI are the ones going beyond efficiency metrics to build the right foundation first. Vanity metrics are what’s left behind when leaders skip the foundation-building step.
Grounding To A Named Business Metric Before Launch
Forget more dashboards. The key pillar is deciding which business metric a use case is supposed to drive, whether that’s cost per claim, cycle time, retention, error rate or whatever the underlying process actually gets measured on, and reporting against that metric specifically.
This single discipline resolves two of the most common failure modes simultaneously, stopping “we deployed 40 agents at once” from standing in for “we don’t yet know if this worked.” It also forces the redesign conversation to happen up front, because a business metric moving an initiative can’t be named without first deciding what the underlying process is actually supposed to accomplish.
From Usage Data To Business Outcome Metrics
With redesign, reporting moves from how much AI is running to what changed in the number the business already tracks. Usage and adoption don’t disappear as metrics because they’re still useful as a leading indicator of whether an initiative is even being used enough to have an impact, but they do stop being presented as the ultimate outcome.
This is also where executive fluency matters just as much as technical delivery. A board or leadership team that can’t identify which metrics an AI initiative should move, and over what time frame, will keep approving AI initiatives without full visibility, and keep getting metrics that don’t give the complete picture of AI optimization. Closing that gap is a leadership and process discipline, not simply a data science one, which is the capability organizations are aiming to build when they invest in structured AI education for the people setting investment priorities in the first place.
A plateau doesn’t necessarily mean that AI has stopped working, but it is a sign that an organization has optimized the wrong layer: the tool instead of the process, and the deployment total rather than the business impact. The firms making the most of their AI strategies are able to pinpoint when a workflow should exist or not, and measure what’s left against the numbers that make the most sense for the business.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
