Feature
Research Report
Mind The Agentic Action Gap: Stop Losing Money With AI Agents
Once upon a time, if you wanted software, you bought it. If two or three people wanted it, you bought two or three copies.
Then enterprises realized that while they might have a hundred people who used a particular software package, they seldom all wanted to use it at once. This begat per-seat pricing: A company with 100 users bought 50 copies that users shared. This especially happened with Software as a Service (SaaS) vendors.
Now (with apologies to Taylor Swift), we’re entering a new era. Software users could include agentic AI systems, which might run dozens of instances of the product at once, or might run it continuously, doing more work than three human users. This is likely to price software differently. How will it work?
Table of Contents
- AI Agents Break the Per-Seat Pricing Model
- What Will New Software Pricing Look Like?
- Outcome-Based Pricing Creates a New Budgeting Problem
- As Pricing Shifts, Software Comparisons Get Murkier
- To Prepare for Agent Pricing, Start Measuring Now
AI Agents Break the Per-Seat Pricing Model
Software pricing under agentic AI is a vendor issue as well, said Bo Lykkegaard, associate vice president at IDC.
“A mid-sized 1,000-person company can buy a financial suite with only three users,” he explained. “These three named human users also let agents carry out transactions, and all transactions are channelled through these three. As a result, what used to be a 50-user deal is now a three-user deal. That is, of course, not acceptable for the software vendor.”
Faram Medhora, principal analyst at Forrester, added that AI agents are breaking the economic logic of per-seat software pricing. “When a single agent performs the work of multiple employees, pricing software by user count no longer reflects how value is created, and CFOs are already challenging that math.”
Consequently, vendors are looking at different ways of pricing software likely to get used by agentic AI, though they’re considering multiple approaches.
Related Article: Best Free AI Tools in 2026: Top Picks for Writing, Design, Productivity and More
What Will New Software Pricing Look Like?
These pricing approaches, according to Lykkegaard, could be based on:
- Consumption, such as the number of AI units or CPU cycles used
- Outcomes, such as an improvement in key performance indicators (KPIs)
- Organizational capability, such as the number of employees served or the number of invoices submitted
While KPI improvement is somewhat further out, Lykkegaard noted, some financial application vendors are experimenting with it. Another possibility is that more capable subscription layers will have a higher minimum user count.
Vendors could combine these approaches with per-seat pricing in a hybrid approach, said Medhora. This could be a fixed platform fee combined with variable charges tied to what agents actually do.
Some vendors use such pricing models now, he added, pointing to Salesforce's Agentic Work Units and Workday's consumption-based Flex Credits. These approaches "reflect a market testing new economics while trying to preserve buyer confidence and revenue predictability,” he said.
As of October 2025, around 35% of the 30 SaaS companies surveyed had already increased their per-seat pricing, bundling AI features into existing tiers (such as Zoom). Around 65% introduced a hybrid approach, layering an AI meter of usage or feature access on top of seat-based pricing (such as Adobe and Salesforce). None either charge separately for AI nor have shifted to outcome-based pricing.
Vendors that don’t turn to outcome-based pricing? According to Lykkegaard, they'll face growing misalignment between price and delivered value as customers purchase fewer and fewer seats and freeride instead, resulting in margin pressure and general market misalignment.
Outcome-Based Pricing Creates a New Budgeting Problem
One of the major challenges for enterprises in outcome-based pricing, noted Medhora: How do you budget for it?
Any organization that found an AI system or agent imposed more costs than expected has encountered the problem. But as multiple software products get priced that way, it’s tough for an organization to know how much it’s going to cost.
It’s not going to be an immediate problem, though, said Lykkegaard. “It will be a bit harder in the future, but the change will be more gradual than you might think. Subscription models will continue, and even for consumption-based pricing, customers will buy SaaS access via subscription tiers with different consumption levels.”
With a caveat.
“Of course, organizations could exceed that threshold and face significant additional subscription changes,” Lykkegaard warned. “This is why enterprises will have to adopt some level of scenario-based budgeting, anchored in expected volumes, productivity gains and agreed consumption guardrails.”
“Enterprises will not accept variable, outcome‑linked AI pricing unless they can trust that agents are operating safely, predictably and within defined bounds,” agreed Medhora. “Before CFOs pay for outcomes, they need confidence that agents are producing the right outcomes, without introducing compliance risk, operational errors or hidden downstream costs. As a result, agent pricing adoption is inseparable from guardrails, auditability and value attribution.”
As Pricing Shifts, Software Comparisons Get Murkier
Outcome-based pricing may also make it more difficult for enterprises to compare offerings from different vendors, Lykkegaard said.
“In the near term, lack of standardization will limit transparency. Comparison will evolve toward price-to-value ratios, such as cost per outcome, per process or per productivity gain, or simply the total annual cost of a given capability for the organization.”
The transition could also temporarily add costs for enterprises, wrote Eric Maltiel and Josh Sandberg, partners at Bain & Co.
“The toughest challenge could be asking customers to spend more before they see savings,” they noted. “Take the case of a SaaS vendor pitching a $40,000 AI agent that could eventually replace an $80,000 sales development rep. In the short term, the customer still needs both the employee and the AI agent while it evaluates outcomes. The customer must raise its cost by 50% for an undefined period.
"Unless the vendor helps its customer bridge that gap — perhaps with incentives such as delayed payments, flexible contracts or performance guarantees linked directly to profit and loss outcomes — the sale may stall.”
Webinar
Apr
23
AI for Your DXP: Connect What You Have, Transform How You Work
Most AI strategies stop at the platform—but work happens elsewhere. Bring intelligence into Teams, email, tickets and CRM.
Register
Webinar
Apr
27
Where Contact Center AI Is Actually Headed
A straight read on where the contact center platform market is going and what it means for your next technology decision.
Register
Webinar
May
12
How Marketing Teams Can Take Control of the Web Experience
See how Veo and Kvalifik put marketing in the driver's seat, shipping faster and growing qualified bookings by 150%.
Register
Webinar
On demand
The State of Enterprise Site Search: From Basic Utility to Strategic Advantage
Join CMSWire and SearchStax for a conversation about how enterprise IT and marketing leaders are moving beyond basic site search.
Watch Now
Webinar
On demand
Content Leaders Collective: Navigating Content Decisions at Scale
Discover how content leaders are modernizing content operations, avoiding costly missteps and preparing for scale and AI.
Watch Now
Webinar
On demand
Do More with Less: Modernizing the Cloud Contact Center for 2026
Learn how to leverage cloud platforms without adding a single hire to personalize every customer interaction.
Watch Now
Related Article: Enterprises Are Handing Over Authority and Calling It a Software Purchase
To Prepare for Agent Pricing, Start Measuring Now
How should enterprises prepare? It’s going to involve a lot of metrics, and organizations should start tracking them now to ensure they're ready.
“Most enterprises are not yet equipped for this shift,” Medhora said. “Their budgets, controls and governance models assume fixed per‑seat spend and human accountability. Enterprises that win will treat AI agents as economic actors: measuring what they produce, what they cost, what risks they introduce and what value they safely deliver.”
Lykkegaard said enterprises should:
- Establish cost and value governance for SaaS and AI-powered SaaS, using FinOps for agentic SaaS apps and AI agents
- Establish workflows to track usage, outcomes and unit economics
- Pilot and baseline productivity and cost benchmarks, including the total cost per year for a given capability for the organization
Organizations that don’t do this ahead of time will run into problems, Medhora predicted. “Those that bolt agents onto existing contracts without re‑instrumenting governance and finance will face cost volatility, risk exposure and value disputes they never modeled.”
A future software bill may look more like an electric bill, with some fixed and some per-use charges.
“Agent‑based pricing will scale only as fast as enterprises can prove, govern and trust the value AI agents deliver,” according to Medhora. “The future software bill will look more like a utility statement, and only buyers who can measure both cost and safety will be ready for it.”