Conflicting AI-value surveys need a measurement contract before a strategy
Bain-related reporting says 6% of marketing organisations see significant AI impact, while BCG says nearly half of companies create meaningful value. The gap is a methods question before it is a market conclusion.
What happened
Business Insider reported Bain research in which 6% of 1,397 senior marketing and finance executives said AI delivered significant performance impact in marketing. BCG separately reported that 48.5% of 1,330 companies were “future-built” or scaling AI and outperforming laggards, using broader enterprise and financial definitions.
Why it matters
The results describe different populations, functions, labels and outcomes. Executives cannot use either percentage as an enterprise baseline until they align the unit of analysis, denominator, time window, counterfactual and value definition.
Two survey stories published on 30 September appear to point in opposite directions. Business Insider reported Bain research in which 6% of marketing organisations said AI was delivering significant performance impact. The reported survey covered 1,397 senior marketing and finance executives; 95% said their organisations used AI. The article describes stronger organisations as centralising strategy, redesigning workflows and focusing on customer outcomes.
BCG reported that 7.5% of companies were “future-built” and another 41% were scaling AI and outperforming laggards. Its Applied AI Index survey covered 1,330 CxOs and senior leaders. BCG linked those categories to relative shareholder return, revenue and EBITDA growth and said corporate AI spending had risen to 3.3% of revenue.
The figures are not direct replications. One asks about significant performance impact in a function; the other classifies companies using broader enterprise scaling and relative performance. Respondent roles, samples, definitions, time windows and possible selection effects differ. Neither survey design establishes that AI caused the reported financial outcomes.
Write the denominator first
Before quoting a market percentage, define the unit: campaign, workflow, function, business unit or company. State whether “value” means time saved, avoided cost, incremental margin, revenue, risk reduction or a composite. Record the baseline period, comparison group, attribution rule and confidence interval where available. Separate self-reported adoption from instrumented use and audited financial effect.
For one portfolio, use a common ladder: activity, operational output, business outcome and financial result. A faster content draft is activity. A shorter cycle time with stable quality is an operational output. Higher conversion after an agreed comparison is a business outcome. Incremental contribution after model, data, review and change costs is a financial result.
Reconcile before deciding
Choose ten AI initiatives and rescore them under the same measurement contract. Require an owner, baseline, measurement window, counterfactual, cost boundary and stop rule. Report how conclusions change when the definition moves from self-reported value to observed operational or financial evidence.
The counterargument is that executive surveys identify patterns before audited data mature. They can. The BCG and Bain-related accounts both point toward workflow redesign and coordinated operating models rather than tool purchase alone. That convergence is useful. It remains association and practitioner guidance, not a causal estimate.
The Skills Intelligence Atlas can help describe the capabilities needed to instrument workflows, evaluate models and manage change. It cannot reconcile incompatible survey constructs.
The immediate decision is to approve no new “AI value” target until finance, operations and the business owner sign a one-page measurement contract. Use external percentages as hypotheses for local investigation, not as targets or proof that a programme is ahead or behind.