Who Owns Results When AI Workflows Span Teams?

· Decision Intelligence
China Strategic Signals

Before expanding an AI rollout, COOs should define what counts as a finished job and give someone the authority to fix the handoffs that can delay it.

An AI pilot can cut the time one team spends preparing a file and leave the team receiving it with more exceptions to resolve. Both teams can meet their targets while the customer waits just as long. Before approving a wider rollout on the strength of that pilot, a COO needs two answers. When does the work count as done? Who can change the process if the gains disappear at a handoff?

In a November 2026 survey by HFS Research and Rackspace, 33% of 502 senior AI decision-makers said their most significant AI program had met or exceeded its original business case. The share was 37% among respondents with AI in production in one function, 28% among those with deployments across several functions, and 56% among those with AI embedded in day-to-day operations. The survey compared different groups at a single point in time. The differences could reflect program complexity, the targets companies set, or which companies made it to more mature stages.

The survey also found that 98% of respondents reported having both an executive sponsor and someone accountable for the impact on profit and loss. Among respondents whose AI programs had reached production or beyond, differences in reported success by sponsor type or operating arrangement did not reach statistical significance. The survey did not measure whether the people accountable for the results had the authority to change how teams handed off work. A manager can own the result and still need three other managers to agree on who will handle a growing queue of exceptions.

Decide where the clock stops

Northern Trust and Microsoft reported that an AI agent helped cut the time needed to prepare historical transaction data from about two and a half weeks to three days. The measure covers data conversion, analyst review, exception handling, and final validation. Northern Trust also described a regular reconciliation process before going live. The companies reported preparation time and gave no comparable figure for the full client onboarding cycle.

If the investment case promises faster preparation of validated data, measure the time to final validation. If it promises faster client onboarding, keep the clock running until the client can use the service. Suppose validated files arrive sooner but spend longer waiting for reconciliation. Measuring both intervals shows where the delay occurs and which team needs to be involved.

In a randomized field experiment at Procter & Gamble, participants using AI produced product proposals that better integrated technical and commercial perspectives. The study measured performance on a defined innovation task. In an operational rollout, check whether the receiving team can use the output, how much review or rework it needs, and whether the gains carry through to delivery.

Give the person responsible room to act

Assign the workflow to a business leader who will remain responsible after the project team moves on. Australia’s Digital Transformation Agency takes a similar approach in its benefits management guidance, assigning responsibility for realizing benefits to an ongoing business role. If one department sends too many incomplete cases downstream, the workflow owner should have the authority to change what the receiving team will accept, reroute exceptions, or coordinate temporary staffing within agreed limits. The COO can resolve decisions beyond those limits, such as moving budget between departments. An existing manager can fill the role if they already have the necessary authority.

The workflow owner can show how many staff hours the process saves. Finance can assess whether those savings free up capacity, improve service, or reduce spending. Any extra time absorbed by the review team belongs in the same calculation. If the hours saved upstream reappear downstream, Finance needs both figures to calculate what happened to the cost of delivery.

What should justify a wider rollout?

Before approving expansion, agree on the improvement that would justify it and the limits the process must stay within. If faster delivery is the goal, set a target for elapsed time. Track the share of cases that pass review, the hours each team spends on review and rework, and the total cost per acceptable delivery. Compare periods with similar case volumes and a similar mix of case complexity. Record changes to staffing or the AI model so they can be taken into account when interpreting the results.

If one team gets faster and the customer still waits, use the handoff data to find where work is piling up. When the workflow owner has the authority to change that handoff, test the change at the current scale. Resolve any outstanding decisions about staffing, budget, or controls owned by another team before expanding. The COO can approve expansion when the evidence shows that the workflow delivers the promised result and a named owner has the authority to intervene if performance falls short.

Section image

Turning Global AI Signals into Business Decisions

InfoAI helps decision-makers identify what is changing in AI, evaluate the evidence behind those changes, understand what they mean, and determine which business assumptions or decisions may need to be reconsidered.