Todd Bernson is the Chief AI Officer at BSC Analytics, where he leads AI strategy and cloud transformation for enterprise clients.
Enterprise AI has an accounting problem.
Technology leaders report progress in tools: co-pilots deployed, coding assistants adopted, service agents piloted and hundreds of use cases identified. Boards are measuring something else. They are asking when the company will consolidate its data centers, remove a billion dollars from operating expense, reduce cost-to-serve or materially increase revenue per employee.
Both conversations are legitimate. The problem is that they are rarely connected.
McKinsey’s 2025 State of AI survey found that 88% of organizations regularly use AI in at least one business function, yet only 39% report any enterprise-level EBIT impact and most of those attribute less than 5% of EBIT to AI. Adoption is expanding much faster than measurable enterprise value.
The root cause is not simply weak execution. The use case is the wrong unit of management for enterprise transformation. A use case can improve a task owned by one team. A board mandate spans functions, systems, budgets and months of organizational change. A collection of locally optimized tasks does not become a structural outcome merely because the benefits are added together.
The Use-Case Trap
The standard AI playbook begins with an opportunity inventory. Teams identify dozens or hundreds of ideas, score them on impact and feasibility and fund the most attractive pilots. This feels disciplined, but it creates a predictable bias: The portfolio favors initiatives that are easy to demonstrate, narrowly owned and technically self-contained.
Board-level outcomes have the opposite characteristics. Reducing cost-to-serve may require eliminating avoidable demand, redesigning customer journeys, automating decisions, consolidating platforms and changing workforce capacity. No single chatbot, prediction model or coding assistant owns that result.
BCG’s research illustrates the difference between activity and focus. AI leaders pursue roughly half as many opportunities as less advanced peers, yet scale more than twice as many AI products and services. They also derive 62% of AI value from core business processes rather than limiting the technology to peripheral productivity tools.
The implication is uncomfortable: A larger use-case pipeline can be evidence of weaker strategy. It distributes investment across attractive ideas instead of concentrating it behind the few operating outcomes that matter.
Build An Outcome Architecture
The alternative is not to stop developing use cases. It is to subordinate them to an outcome architecture: a management system that connects an enterprise mandate to value pools, operating-model changes, enabling capabilities and auditable financial results.
Making that shift practical takes four disciplines.
1. Start with a mandate that has an owner, baseline and deadline.
“Deploy generative AI in customer service” is an initiative. “Reduce cost-to-serve by 35% within 18 months without lowering customer satisfaction” is an outcome. The second statement defines the economic target, the constraint, the accountable executive and the period in which value must be realized.
The AI road map should map to a small number of mandates like this, not to a departmental wish list.
2. Decompose the mandate into value pools.
A large target is rarely produced by one breakthrough. Cost-to-serve, for example, may be reduced through demand elimination, straight-through processing, lower rework, greater employee capacity and platform simplification. Each value pool should have a baseline, an operational metric and a credible contribution to the enterprise target.
This prevents teams from attaching unrelated productivity claims to a large financial ambition after the work has already begun.
3. Redesign the operating model before selecting the technology.
Executives should determine which workflows disappear, which decisions become automated, where human judgment remains necessary, which systems can be retired and how roles will change. AI then becomes one enabling capability within a broader transformation, not the transformation itself.
McKinsey’s data bears this out: AI high performers are nearly three times as likely as other organizations to have fundamentally redesigned their workflows. The differentiator is not simply access to better models. It is the willingness to change how work is organized.
4. Create a value ledger that finance can audit.
Every initiative should identify its operational leading indicator, financial lagging indicator, value owner and realization mechanism. A faster process does not automatically create savings. Capacity created by AI becomes financial value only when the company uses it to increase output, avoid future hiring, reduce external spend or remove cost from the budget.
This is where many business cases fail. They count hours saved as cash realized, combine overlapping benefits from multiple projects or claim enterprise impact without reconciling the result to the financial plan. A value ledger forces the portfolio to distinguish technical performance, operational improvement and realized economic value.
From AI Portfolio To Transformation Portfolio
Co-pilots, agents and coding assistants still matter. But they are components of an outcome, not an enterprise strategy. A coding assistant may accelerate application remediation; it does not, by itself, consolidate a data center. That outcome also requires application discovery, migration sequencing, architecture standards, risk controls, infrastructure changes, operating-model decisions and the retirement of legacy assets.
The governance model should reflect that reality. Fund cross-functional outcome portfolios, not isolated AI projects. Assign one executive to the result, sequence the dependencies across technology and the business, and stop work that cannot demonstrate a credible path to the mandate.
Boards no longer need more evidence that AI can draft, summarize, predict or generate code. They need evidence that management can use those capabilities to change the economics of the enterprise.
Stop treating the number of use cases as a measure of ambition. In many organizations, it is a measure of fragmentation. Inventory the outcomes the company must deliver, build the operating and technical capabilities required to produce them, and prove the result in the financial units the board already uses.
Your board already has an outcome inventory. Your AI strategy should start there.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
