AI arrived carrying its own scoreboard, and that is the problem. No enterprise resource planning rollout ever reported 94% accuracy. The number did not exist, and those programs justified themselves in the only language available to them, which was money. AI comes pre-loaded with accuracy, precision, processing speed, and F1 scores. David D. Ellison, Chief Data Scientist and Director of AI and HPC engineering at Lenovo, watches organizations fail on exactly that abundance.
“AI teams measure model accuracy, processing speed, and precision,” he says. “Business leaders measure revenue, cost, and growth. If those two never meet, alignment is impossible.” An AI function can now declare victory without asking the business anything, a capability no technology group has ever held. Ellison spent the past eight years scaling Lenovo’s AI business from $14 million to $1.7 billion, and his verdict is that AI without strategic alignment is expensive experimentation.
Start With the Objective Instead of the Capability
The most common failure Ellison encounters follows a single pattern. An organization buys a platform, hires a team, and then goes looking for problems to solve. The order seems irrational until you notice what the platform provides. A capability arrives with benchmarks attached, letting a team demonstrate progress from week one without identifying a single business outcome.
Ellison reverses it. Start with the top three to five business priorities, whether that means cutting costs, growing revenue, or keeping customers, then ask where AI can genuinely move those numbers. Starting with a clear objective provides the measurement standard alongside the problem, removing the option of grading your own work.
Put AI Under an Executive Owner
“AI strategy doesn’t fail in the data center,” Ellison says. “It fails in the boardroom.” An AI function sitting in IT or a single department unaccountable to senior staff never receives the resources, access, or priority required to scale. Someone at the leadership table has to connect AI outcomes to business outcomes, and without that connection even brilliant technical work stays invisible. An executive owner also introduces what a self-scoring function otherwise avoids entirely, which is an external judge. Technical metrics stop being the final word once someone with budget authority answers for what they produce.
Track Business KPIs From Day One
Every initiative needs key performance indicators (KPIs) that leadership cares about, tracked from the beginning rather than reconstructed afterward. Timing carries the weight. A team that instruments a project for accuracy at the outset cannot retroactively demonstrate revenue impact, especially since nobody captured the data and nobody set the baseline.
Committing from day one to the business metric the initiative should improve will build the business case alongside the model. It also forces an honest conversation about whether the initiative can move that number at all before any budget is moved. Some initiatives cannot move the number, and finding that out early costs considerably less than discovering it at the board review.
Ellison’s summary is to start with the objective, install the right leadership, and measure what the business actually values. Each step removes another way for an AI program to succeed on its own terms, while producing nothing anyone else can see. That is the shift he describes, where AI stops being a line item and starts being a competitive advantage.
To learn more about aligning AI strategy with business objectives, connect with David D. Ellison on LinkedIn.
