What should executives answer when asked about the impact of their AI initiatives?
The short answer
Executives usually answer the impact question with activity: project status, milestones and budget consumption. A robust answer would compare expected and realized value contribution for each initiative. Where this comparison is missing, continuation depends less on impact than on the persuasive power of the person presenting it.
The question arrives no later than year two
In the first year of an AI transformation program, an advisory or supervisory board asks about progress. In the second, it asks about impact. The difference between the two questions is the difference between a presentation and an extension.
The answers that usually follow describe activity: completed milestones, funds spent, active initiatives. These are valid facts, but they do not answer the question. Impact is the comparison between the contribution expected and the contribution realized.
Pressure on this question is rising quickly. 91 percent of German companies now regard generative AI as business-critical, up from 55 percent in the previous year; 82 percent plan higher budgets, more than half of them by at least 40 percent (KPMG Germany, 2025).
The body asking the question can barely assess the answer
Almost 60 percent of German executive and supervisory board members say they have little or no AI knowledge. Two percent consider themselves very knowledgeable and experienced. Internationally, 66 percent of surveyed board members say their board has little to no knowledge or experience with AI, and in 31 percent of boards the topic is not on the agenda (Deloitte Global Boardroom Program, 2025).
That creates an uncomfortable situation. The impact question is asked by a body that can assess the technical answer only to a limited degree. In that setting, continuation is decided by the persuasiveness of the presenter unless the evidence is structured beforehand.
The proof horizon is longer than the reporting cycle
More than two thirds of AI leaders say return must be viewed over longer periods of two years or more. At the same time, only about one third have redesigned complete workflows, and 5 percent use AI agents for a fundamental redesign of delivery (Deloitte ROI of AI, German cut, 2026).
Even where AI is already productive, the period required for proof is longer than the rhythm in which boards receive reports. If impact evidence begins in the report, it begins too late: baselines are missing and assignment from initiative to goal can only be asserted after the fact.
This pattern is not new and not specific to AI. It appears in every reorganization, system rollout and market entry whose effect appears after the reporting cycle. AI merely removes the escape route: where a board cannot assess the subject matter deeply, only the evidence remains.
The obvious solution creates a new problem
The common reflex is one metric per goal, reported quarterly and perhaps tied to compensation. The less a board can judge the content, the more attractive this reduction becomes.
Surrogation is the confusion of a metric with the goal it is meant to represent. In two experiments, the effect was strongest when compensation was tied to a single metric of a goal and much weaker with several measures of the same goal (Choi, Hecht & Tayler, The Accounting Review, 2012).
The well-known Wells Fargo case shows the cost. Cross-selling became the proxy for customer relationship quality. The result was millions of unauthorized accounts (Harris & Tayler, Harvard Business Review, 2019).
For board reporting, a single number per goal is therefore not a solution but a risk. It becomes robust only with several measures that read the same goal from different directions.
What belongs into the board paper before it is needed
Four entries per initiative, recorded at the decision rather than at review: the strategic goal it contributes to, the expected value contribution, the stop condition and the responsible person. With those four entries, the impact question can be answered in one meeting.
The stop condition is the most uncomfortable and valuable part. It states upfront how the organization would recognize that it was wrong. Written down early, it allows an initiative to be stopped twelve months later without turning the stop into a loss of face.
BizzPlAI is built for this. The strategy comes from the company; the system keeps what was expected at the decision. Target, expected value contribution and stop condition remain findable from that moment onward.
Start with the three largest AI initiatives. For each, write down the contribution expected at approval. Where that cannot be reconstructed today, the impact question will become uncomfortable in year two, regardless of how well the program has actually run.
Common follow-up questions
Which four questions should an advisory board be able to ask about transformation?
What did it bring, what did it cost, what did we leave for it, and how will we recognize progress in the next 90 days? The first two are usually answered. The third and fourth require assignment of measures to goals from the decision point.
Why does a status traffic light not prove impact?
A traffic light shows whether an initiative is on plan. It says nothing about whether it contributes to a strategic goal or by how much. A program can be green throughout and still create no measurable contribution.
What is surrogation in relation to metrics?
Surrogation means confusing a metric with the goal it is meant to represent. Executives then treat the measure as if it were the strategic construct itself. The effect is strongest when compensation is tied to a single metric.
Sources
- Governance of AI: A critical imperative for today's boards, 2nd editionDeloitte Global Boardroom Program, 2025
- KI im Aufsichtsrat (German cut of the Global Boardroom Program)Deloitte Germany, 2025
- Aus Kuer wird Pflicht: 91 Prozent der deutschen Unternehmen sehen KI als geschaeftskritisch anKPMG Germany, 2025
- The ROI of AI - German CutDeloitte, 2026
- Lost in Translation: The Effects of Incentive Compensation on Strategy SurrogationThe Accounting Review (Jongwoon Choi, Gary Hecht, William Tayler), 2012
- Don't Let Metrics Undermine Your BusinessHarvard Business Review (Michael Harris, Bill Tayler), 2019
See how strategy arrives.
In 30 minutes, we show you how to steer your initiatives instead of merely managing them.