NCR Atleos

Jaideep Agrawal

Enterprise Transformation Authority

Almost every executive team I talk to is asking some version of the same question: how do we use AI and automation to move faster, cut cost-to-serve and improve customer experience? It's a fair question, but it gets asked too early. Before scaling AI, someone should ask what kind of operating model AI is being asked to scale.

Technology rarely fails on its own. It struggles when layered on work that's fragmented, over-governed, or missing clear ownership. AI can speed up a process, but it doesn't automatically make that process better. If the operating model underneath is complicated, AI just helps that complexity move faster.

Plenty of organizations have spent years digitizing workflows and deploying bots, producing real, local wins: less manual effort here, cleaner reporting there. Yet the enterprise can still feel slow to change. It's rarely the technology that's the bottleneck; it's the seams between teams, where work quietly stalls. A decision sits idle because nobody's accountable. An approval survives for years because nobody revisits it. Controls pile up, rarely removed.

I've seen this pattern most clearly in field service organizations, where dispatch, diagnostics and technician decisions often sit with different teams and no single owner can see or fix the full flow. The instinct is to add a bot at each handoff. The better fix is to redesign who owns the decision end to end, then let AI support that owner. I've watched this approach, applied across a field network supporting several hundred thousand devices, push prescriptive repair accuracy past 95 percent, a result that came from rethinking ownership, not from the technology alone.

CIOs, CTOs and COOs need to change this conversation. AI shouldn't be treated only as a productivity tool; it's a chance to modernize the operating model itself. Before deciding where to apply AI, ask which handoffs exist because the work needs different expertise, and which exist because nobody owns the outcome. Once that's visible, simplification becomes a discipline: some steps deserve automation; others should be removed, standardized or redesigned. The best opportunities sit where volume, risk, cost and speed intersect.

  Ask yourself: are you applying AI to make a complex process repeatable, or are you simplifying the operation first and then applying AI? The first can make a broken process run faster. Only the second lowers what it costs to run.  

None of this holds without real engagement from business process owners, not just technology leadership. Simplification redraws who's accountable, and that draws resistance unless it's led from the top with genuine buy-in and a clear RACI: who's responsible, accountable, consulted and informed.

Skip that, and old handoffs quietly creep back in no matter how well the process was redesigned.

AI does far more good once placed inside a simplified decision flow: prioritization, exception routing, spotting patterns, supporting a decision rather than replacing it. Task automation asks how to make an activity faster. Enterprise transformation asks why it exists, who owns the result and how the work should flow. That question is where technology leadership earns its seat at the table.

Governance has to be part of this from the start, not bolted on later. As AI shapes more decisions, organizations need clarity on data ownership, escalation paths and who's accountable when something goes wrong.

One piece still gets skipped: explainability.

If a model flags a transaction or recommends an action, someone has to explain why, not just confirm it happened. Legal exposure deserves the same early attention: privacy rules, compliance obligations and a growing body of AI regulation are moving faster than most governance frameworks were built for. Bringing legal and compliance in early is cheaper than finding the lines after the model is already live.

Metrics need the same discipline. Counting bots or AI use cases in flight makes a program look busy, but it doesn't prove the enterprise is better off. What matters more: decision cycle time, manual handoffs, SLA reliability, exception rates and the numbers that show whether AI is creating real leverage or just digital noise.

The organizations that get this right won't have the most sophisticated models. They'll have a clear operating model, sound decision-making and the discipline to connect technology investment to real outcomes. Before your next AI investment review, ask  which handoffs exist because of real expertise, and which because no one owns the outcome; could we explain a model's recommendation to a regulator tomorrow; and are decisions getting faster and better, or are we just counting AI use cases?

The future of enterprise AI isn't more automation stacked on automation. It’s simpler operations, clearer ownership and execution that bends without breaking. So ask yourself: are you applying AI to make a complex process repeatable, or are you simplifying the operation first and then applying AI? The first can make a broken process run faster. Only the second lowers what it costs to run.