Usage-based and hybrid pricing models are changing how B2B companies generate revenue — and creating new headaches for the finance teams behind them.

Tabs co-founder Rebecca Schwartz and PwC Partner Amit Dhir sat down to unpack exactly what that means in practice: how pricing model decisions ripple into revenue recognition, forecasting, and financial ops — and what it takes to scale without piling on manual work.

Watch the on-demand recording to get practical frameworks, real-world examples, and a clear path to operationalizing usage-based revenue — including a forward-looking take on how AI will reshape financial workflows. If your team is navigating pricing complexity heading into the back half of the year, this is worth an hour.

Watch the recording now

Each edition delivers one clear, evidence-backed idea you can use. This week: I identify five key metrics for measuring and reporting on AI adoption and operation

Many management teams tracking AI are measuring the wrong things. They're watching adoption dashboards light up, counting licenses deployed, and celebrating model accuracy scores — none of which tell them whether AI is actually working for the business. The research is blunt on this: companies that enhance their KPIs with AI-specific measurement are three times more likely to see greater financial benefit than those relying on traditional approaches. Yet the majority of organizations still haven't built the measurement infrastructure to get there.

Here are the five metrics that actually close that gap.

This is your north star. Not "are people using AI?" but "did AI move a number that matters to the business?"

IBV measures the quantifiable improvement in key business outcomes directly attributable to AI — increased revenue, reduced costs, faster cycle times, or improved customer outcomes. The hard part is attribution. You need a baseline before deployment and a controlled comparison after. McKinsey's five-layer AI measurement framework puts financial impact at the top of the stack: revenue uplift, cost-to-serve reduction, margin improvement, and total cost of ownership including cloud and token spend. Everything else feeds into this number.

A critical caution from recent research: AI adoption frequently follows a J-curve, with measurable productivity declines in the short term before gains materialize. American Economic Association research using U.S. Census Bureau data found that a one standard-deviation increase in AI use was initially associated with roughly 2% lower labor productivity — followed by outperformance over peers at the four-year mark. If management only sees the dip without context, they pull the plug on initiatives that were about to pay off. IBV, tracked with a timeline, gives them that context.

What to track: Revenue uplift or cost reduction attributable to specific AI use cases, measured against a documented pre-deployment baseline. Build in a 12–24 month horizon before declaring an initiative dead.

AI's primary purpose in most organizations is not automation — it's improving judgment. Decision Impact Rate measures the percentage of AI-influenced decisions that result in materially different (and better) outcomes than would have occurred without AI.

Peer-reviewed research from the journal Management Decision (2025) confirms that AI enhances forecasting accuracy, accelerates decision speed, and improves managerial decision quality — but decision speed and quality are only improved if the AI is actually influencing the decision, not just decorating the workflow. Giachino et al. (2025) and Neiroukh et al. (2025) empirically validate a positive relationship between AI-driven decision-making capabilities and firm performance, with decision speed and quality identified as the key mediating mechanism.

This metric surfaces a failure mode that adoption metrics completely miss: AI that gets deployed, runs in the background, and is quietly ignored by the people it's supposed to help.

What to track: Percentage of AI recommendations reviewed vs. acted upon by function; downstream outcomes for AI-influenced decisions vs. unassisted decisions. In healthcare contexts, this has been tracked by measuring physician treatment plan changes based on AI suggestions and corresponding patient outcome improvements.

Technically excellent AI that users circumvent creates zero value. Adoption velocity measures both the speed and depth at which AI tools are actually used across the organization — and it's one of the most predictive early indicators of whether an initiative will scale.

Federal Reserve research on AI uptake in the workplace found that firm-level AI adoption grew at a 73–78% annualized rate between 2023 and 2024, but actual worker utilization ranges from 20–40% across occupations — with wide variance. That gap between license deployment and genuine usage is where most SMB AI investments go to die.

The MITRE AI Maturity Model identifies adoption and scaling as one of six pillars critical to AI success, alongside strategy, talent, operating model, technology, and data. A systematic literature review of AI maturity frameworks (2025) similarly surfaced organization and culture — not just technology — as a primary dimension of AI maturity. Adoption velocity is the leading indicator for both.

What to track: Weekly active users by tool as a percentage of target population; feature utilization depth (not just logins); departmental adoption patterns to identify where AI is stalling and why.

Most organizations budget for the license. That's 30–50% of the real cost. The rest — infrastructure, data engineering, integration, training, governance, and ongoing model maintenance — gets hidden in other budget lines or simply not tracked at all.

Most AI project evaluations report only one side of the ledger. Kunz & Hess (2025) find that AI systems are “rarely looked at from a TCO-based angle with regard to the whole lifecycle,” and that cost dimensions are routinely omitted from AI business case calculations. Their framework identifies nine lifecycle cost categories: system design and development, customizing, integration into controlling, deployment and documentation, initial user training, ongoing operations (including per-token inference costs and subscription fees), recurring user training cycles, risk management, and regulatory certification. Miss any one of these and your ROI denominator is wrong before you start.

The basic ROI formula is straightforward:

AI ROI=((Total Benefits−Total AI Costs) / Total AI Costs)​×100

But the denominator requires all nine cost categories to mean anything. McKinsey’s five-layer measurement framework similarly recommends tracking total cost of ownership alongside benefits on a rolling basis, with fixed review cadence and stage gates that prevent unproven initiatives from scaling.

What to track: All seven cost categories — software licenses, compute, data infrastructure, talent, integration, training/change management, and governance/compliance. Compare against realized business value on a rolling basis. Require a living business case for every initiative above a defined spend threshold.

Production AI degrades silently. Models trained on last year's data are making decisions with this year's reality, and nobody gets an alert. Drift rate tracks how much a model's real-world performance has diverged from its deployment baseline — and it's the metric that keeps operational AI from becoming a liability.

There are two failure modes management needs to understand: data drift (input distributions shift, making the model's training less applicable) and concept drift (the underlying relationship between inputs and outputs changes). Healthcare AI research uses metrics like AUROC, precision, recall, and F1-score to monitor for drift against predefined benchmarks, with automated alerts when performance falls below threshold. Enterprise AI governance frameworks recommend statistical distribution monitoring using Population Stability Index (PSI) and Kullback-Leibler divergence to detect when models need retraining.

Deloitte's 2026 research on AI and decision-making argues that AI evaluation is now a distinct organizational capability — one that requires not just technical monitoring but accountability structures, audit trails, and risk thresholds that trigger human review when model behavior deviates. The PwC 2025 Responsible AI Survey frames ongoing model monitoring as a shift from static governance to continuous control — and identifies it as the defining characteristic of AI programs that generate sustained value.

What to track: Model accuracy and output distribution in production vs. deployment baseline; retraining frequency and triggers; percentage of high-risk AI outputs reviewed by humans; mean time to detect significant degradation.

None of these five metrics work in isolation, and none of them work if they're only reviewed once a quarter. McKinsey's measurement framework explicitly recommends building AI as a managed investment with fixed review cadence, clear stage gates, and a single evidence pack tracking both benefits and total cost of ownership — so only initiatives that prove value advance to scale.

The organizations pulling ahead aren't those with the most AI tools deployed. They're the ones with measurement discipline: they defined value before implementation, built attribution into their rollout, and report to management with evidence — not enthusiasm.

Crane, L., Green, M., & Soto, P. (2025). Measuring AI uptake in the workplace. FEDS Notes. Board of Governors of the Federal Reserve System. https://doi.org/10.17016/2380-7172.3724

Hoang Thanh, N., & Truong Cong, B. (2026). Harnessing artificial intelligence (AI) for enhanced organizational performance in public sectors. Humanities and Social Sciences Communications, 13(1), 1–17. https://doi.org/10.1057/s41599-026-06571-y

Khairullina, A., & Baikalova, A. (Eds.). (2025). Guest editorial: Artificial intelligence (AI) for management decision-making. Management Decision, 63(10), 3213–3220. https://doi.org/10.1108/MD-10-2025-377

Kunz, S., & Hess, C. (2025). A systematic framework to determine AI-induced productivity gains. In DIGITAL 2025: Advances on Societal Digital Transformation (pp. 1–6). IARIA Press. ISBN 978-1-68558-285-2. https://www.thinkmind.org/articles/digital_2025_1_10_18001.pdf

McKinsey & Company. (2026, April 24). From promise to impact: How companies can measure—and realize—the full value of AI [Five-layer AI measurement framework]. QuantumBlack, AI by McKinsey. https://www.mckinsey.com/capabilities/quantumblack/our-insights/from-promise-to-impact-how-companies-can-measure-and-realize-the-full-value-of-ai

McElheran, K., Brynjolfsson, E., Yang, M.-J., & Kroff, Z. (2025). The rise of industrial AI in America: Microfoundations of the productivity J-curve(s) (CES Working Paper No. 25-27). U.S. Census Bureau, Center for Economic Studies. https://doi.org/10.2139/ssrn.5036270

MITRE Corporation. (2023). The MITRE AI maturity model and organizational assessment tool guide. https://www.mitre.org/news-insights/publication/mitre-ai-maturity-model-and-organizational-assessment-tool-guide

Neiroukh, S., & Emeagwali, O. L. (2025). Artificial intelligence capability and organizational performance: Unraveling the mediating mechanisms of decision-making processes. Management Decision, 63(10). https://doi.org/10.1108/MD-08-2024-1789

Sen, R., Golbin Blumenfeld, I., Kosar, J., & Gürdeniz, E. (2025). PwC’s 2025 responsible AI survey: From policy to practice. PricewaterhouseCoopers. https://www.pwc.com/us/en/tech-effect/ai-analytics/responsible-ai-survey.html

Schrage, M., Kiron, D., Candelon, F., Khodabandeh, S., & Chu, M. (2024). The future of strategic measurement: Enhancing KPIs with AI. MIT Sloan Management Review & Boston Consulting Group. https://sloanreview.mit.edu/projects/scholars/the-future-of-strategic-measurement-enhancing-kpis-with-ai/