Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

Stop treating AI like IT

Successful companies recognize that AI is fundamentally different from conventional IT.

[Photo: Getty Images]

Profile photo of Chuck Reynolds

Managing Director · L.E.K. Consulting

My grandfather and father always told me the same thing: When tackling any job, whether replacing a faucet, crafting a table, or building a fence, make sure you have the right tool. You’ll waste hours trying to make insufficient tools work effectively, while using the right tool will give you the correct angles and leverage to complete the job properly and efficiently. The miter saw, the appropriate wrench—these tools don’t always make the job faster, but they ensure the final product works right.

AI deployments follow the same principle.

The IT tools we are using aren’t sufficient to get AI right. Most AI disappointments start with a governance mistake: treating AI as an IT project. CEOs must architect new systems that take advantage of what AI can deliver, not force AI into existing IT frameworks because it’s easier. That approach leaves value on the table.

AI IS NOT CONVENTIONAL IT

If AI is going to scale and pay off, the CEO must see the whole picture—and understand that AI goes beyond IT. The biggest wins come from growth, revenue, and customer outcomes, not just cost reduction.

Traditional IT approaches fall short because IT excels at automating defined functions while AI creates value by digitizing workflows across functions and silos. AI fundamentally changes how work gets done—it doesn’t just digitize processes, it solves problems differently. This means certain tasks become redundant, requiring organizations to break down decision trees, reconfigure data flows, and rethink entire workflows.

AI needs continuous learning, real-time feedback, and frequent iteration, which is in stark contrast to traditional software development with its quarterly releases. Years of business digitization have created patterns of process re-engineering that focus on minor tweaks rather than the radical shifts that AI enables.

AI OWNERSHIP ISN’T AN IT PROBLEM

When IT departments lead AI initiatives in isolation, three predictable problems emerge.

1. MISMATCHED INCENTIVES AND CADENCE

    Traditional, linear project management approaches that mitigate risk during planning stages—often known as risk-first, waterfall processes—slow the learning loops AI depends on. Task-oriented thinking prevents the systems-level transformation AI requires. The answer is a product model tied to business KPIs—one that bundles platforms, role changes, optimization, security, integration, reliability, and guardrails under a cadence built for iteration, not project completion.

    2. SCOPE IS SILOED

    Department-level deployments underperform because AI’s value comes from connecting decisions across functions. Enterprise value never materializes when AI stays trapped in individual business units. Unlocking it requires data quality and governance, feature pipelines, and insight-to-action workflows that span functions rather than stopping at departmental boundaries.

    3. OUTCOMES ARE INSUFFICIENT

    Projects finish on time and on budget, but deliver weak adoption and minimal KPI movement. Organizations get the wrong kind of success—they achieve process compliance instead of business transformation, checking boxes rather than changing outcomes. The right kind of success means measurable improvements in core business metrics: faster decision-making, reduced operational costs, enhanced customer experiences, and so on. What’s needed instead is a use-case portfolio managed by people who understand where AI creates real value: rapid pilots, workflow embedding, and close partnership with line-of-business owners.

    Above all three sits governance for value. The CEO and CFO set targets, sequence the roadmap, and align incentives across all three legs. They architect the new systems that enable true transformation.

    OPERATING PRINCIPLES THAT WORK

    Organizations that succeed with AI follow four core principles.

    1. PRODUCT, NOT PROJECT

        Each AI solution has an owner, users, and business KPIs—revenue, margin, cost-to-serve, cycle time, or churn reduction. Success gets measured in business terms, not technical metrics. This requires redefining the CIO’s role. Transform your CIO into an enterprise-wide strategist with a seat at the leadership table. Shift metrics and incentives away from efficiency and cost containment toward value creation. Marriott’s Drew Pinto serves as chief revenue and technology officer, overseeing not only technology strategy but revenue generation. This allows Marriott to measure technology ROI in real business terms: employee turnover, guest check-in speed, and partner attraction.

        2. THINK BIG, START SMALL

        Fast pilots connect to enterprise vision. Teams measure rigorously and scale only proven approaches, preventing the scattered efforts that characterize failed AI programs. This requires clear swim lanes for AI and IT teams. Establish cross-functional teams with guidelines, schedules, and incentives that ensure AI projects get the daily, iterative development cycles they need. This prevents AI initiatives from getting trapped in traditional IT governance processes.

        3. FIELD-EMBEDDED DELIVERY

        Cross-functional teams build solutions directly in the business to compress feedback loops and drive adoption. Palantir popularized “forward-deployed” engineers embedded with operators—an illustration of field-based iteration that speeds both learning and adoption. This may require structural change. Consider a specialized AI engineering cadre embedded directly in business operations that drive adoption while accelerating embedded AI solution deployment.

        4. RESPONSIBLE BY DESIGN

        Data governance, risk controls, and human-in-the-loop oversight operate from day one, not as an afterthought.

        TREAT AI LIKE A NEW PARADIGM

        AI requires a new toolbox. Let AI be AI—iterative and embedded—and let IT become strategic by design. CEOs must own the whole system: strategy, data, and delivery moving in lockstep toward measurable value.

        The companies that succeed will be those that recognize AI as fundamentally different from conventional IT. They’ll architect new operating models, deploy the right tools for the job, and measure success in business outcomes rather than technical deliverables.

        The alternative is wasting countless hours trying to make insufficient tools work effectively when the right approach was available all along.

        Chuck Reynolds is managing director at L.E.K. Consulting.