For decades, the formula was simple: sell software to more people, grow revenue. Artificial intelligence (AI) breaks that formula.
For investors and market observers, AI is changing the way software companies define their value proposition. Software usage historically expanded in line with headcount; as businesses grew and hired more employees, software revenue increased accordingly.
However, in the new agentic era, the push is to amplify the productivity of each individual, not to add more users. Now, one finance professional can potentially run an AI agent to automate profit and loss statements, cash flow analysis, and bookkeeping tasks that used to occupy an entire team. The human is still there, but the leverage is fundamentally different.
While the shift is logically straightforward, quantifying that value creation is another matter entirely.
To understand why software metrics are changing, it helps to examine how each generation of software created value and how the metrics used to evaluate performance evolved alongside it.
In the On-Premise era, value came from depth of integration. Enterprise software was hard to install, hard to customize, and even harder to rip out. The investor story was simple: We are deeply embedded, and switching costs are high. Stickiness was the signal.
SaaS shifted software from a product you bought to a subscription you renewed. Every employee received a login, making revenue more predictable and compounding, with churn, expansion, and net revenue retention becoming metrics that could be modeled with reasonable confidence. The investor story became: Our revenue is forecastable, and it grows automatically as customers add seats and upgrade tiers.
The AI/Agentic era is something different altogether. In prior eras, value scaled with users. Now, value could increasingly scale with the amount of work the software actually performs. As a result, both business models and measurement frameworks are beginning to evolve.
In fact, some companies, such as Datadog, Confluent, Snowflake, and Sumo Logic, helped normalize consumption-based pricing and reporting frameworks in recent years, making the evolution toward AI-driven value metrics a more natural extension of trends already underway. As of 2025, Metronome reported that 85% of surveyed SaaS companies were experimenting with consumption-based pricing models, according to Metronome in 2025, reflecting athis broader shift toward pricing software based on usage and outcomes rather than seats alone.
This is not the first time a technological shift has forced a reckoning over measurement. When the commercial internet emerged, several metrics persisted because they directly linked user activity to economic outcomes, including customer acquisition cost, customer lifetime value, conversion rates and unique visitors. Others faded once they proved to be attention dressed up as value (remember “eyeballs”?).
Every era measured value through the mechanism that drove growth.
If AI creates value through work performed rather than seats added, how we evaluate software companies may need to evolve as well.
New Metrics Are Emerging, But the Market Hasn’t Converged Yet
As AI moves from experimentation to operational deployment, companies are placing greater emphasis on how the technology generates measurable business value. References to AI have broadened significantly across the Nasdaq-100®, with nearly four out of five constituents now discussing AI on earnings calls, up from one in five before ChatGPT was released.
It’s not just companies grappling with this question either. In his October 2025 note, RBC Capital Markets analyst Rishi Jaluria observed that traditional ARR may be an imperfect measure of AI business quality, suggesting investors may need new metrics to better assess consumption patterns, retention, and the durability of revenue growth.
Companies are increasingly reporting metrics unique to their business in attempts to capture the strength of their AI products. As such, these disclosures are not directly comparable across firms. Broadly, they fall into four categories: work performed, platform activity, customer productivity and AI consumption.
The examples below illustrate how thought leaders in the SaaS space are coming up with new and unique ways to communicate AI-driven value creation to investors.
-
Salesforce introduced an internally developed metric called the Agentic Work Unit (AWU), a platform-level measure designed to capture the total volume of work performed across its agentic products, from Agentforce to Slack. Alongside it, the company reported AI-specific revenue, total token usage, large language model (LLM) calls, and records ingestion in its most recent earnings.
-
ServiceNow disclosed in its earnings calls platform metrics, including 80 billion workflows and 6.4 trillion platform transactions. The company also noted that its Level 1 AI Specialist agent is capable of closing 80%-85% of service requests without human intervention. For financials, it reported AI-specific Now Assist Annual Contract Value.
-
Intuit reported that 2.8 million customers are actively using its AI agents. It noted customer time savings, with its accounting AI agent saving customers an average of 12 hours per month, while the finance agent saves users approximately 17 hours per week.
-
Nasdaq focused on demonstrating AI value through customer productivity gains and adoption. Nasdaq’s Verafin products, Agentic Enhanced Due Diligence Analyst and Digital Sanctions Analyst, can reduce time spent on high-volume customer monitoring tasks by up to 50% and cut alert review workloads requiring human intervention by more than 80%, respectively. Nasdaq has also begun disclosing adoption metrics – 74% of active IR Insight users and 65% of Boardvantage users are now leveraging embedded AI capabilities.
-
Adobe detailed cumulative digital output through metrics such as API calls, automated customer sessions, and “generative credit” consumption from their products.
Some of these metrics measure inputs, including token usage, API calls, and LLM requests. Others measure outputs, such as hours saved, tasks completed without human intervention, and alerts resolved autonomously. The distinction matters, and the market hasn’t decided yet which matters more.
For investors, the opportunity is identifying which metrics reflect durable platform adoption and real economic value, rather than short-term experimentation. Until the market converges on a common framework, if that’s even possible, comparisons across companies will remain difficult.
The Road Ahead
The metrics used to evaluate software have always reflected how software created value. Licenses reflected lock-in. Seats reflected recurring users. ARR reflected compounding subscriptions. Each measure made sense for its moment.
AI introduces a different model. As software increasingly performs work rather than simply enabling it, value creation may become more closely tied to outputs, productivity gains, and economic outcomes than to user growth alone.
The metrics that win won’t just measure AI adoption – they’ll define what we think productivity is worth. That’s a bigger question than it sounds.
