Let AI agents do all the scoping?
getty
Not too long ago, the way we bought software was, shall we say, Netflixed. Instead of receiving disks from a provider to load and keep on a shelf, it was transmitted and updated over the wire, part of a cloud-type delivery mechanism. Now, the industry is going through another transition, beyond Software as a Service.
“AI software is now writing more software,” tech entrepreneur and investor Anthony Pompliano recently pointed out on an X/Twitter post. “Humanoid robots are going to eventually manufacture and assemble more humanoid robots. This idea of exponential productivity is something we have never really seen in our lifetime. The impact will be much larger than we all expect.”
So, will software also be deciding, purchasing, and installing new software as well? That day is already here, and may also help deliver some of that exponential productivity, as well as change the composition of an industry.
AI agents are starting to size up software specifications and purchases for us, as explained in a McKinsey report issued in October, suggesting that traditional SaaS licensing models may go the way of installation disks. Per-seat pricing models are being supplanted by agent-to-agent interactions that base software value on outcome- or usage-based models.
“We’re entering a new era where AI agents become the primary users of SaaS,” according to Vara Kumar Namburu, co-founder and head of R&D and solutions at Whatfix, which helps enterprises manage software lifecycles. “AI agents won’t replace SaaS or humans; they’ll redefine both—automating the routine, amplifying decision-making, and ensuring simplicity and control remain at the center of digital work.”
AI agents will “autonomously execute workflows, interpret intent, and collaborate across systems to deliver outcomes. Much like self-driving cars now share the same roads as human drivers, we’ll soon see AI agents and humans sharing the same digital interfaces, working side by side,” Namburu believes.
This transformation is reshaping how software is bought and valued, he continued. “The per-seat model – built for a human-user world – will gradually give way to usage-based pricing that mirrors real activity and measurable value creation.”
This emerging approach is inverting the economics of software purchasing, “Traditional SaaS is expensive to build but has near-zero marginal costs, while AI is inexpensive to develop but incurs high, variable operational costs,” said Michael Mansard, principal director at Zuora Subscribed Institute, a think tank for the subscription economy. “These shifts challenge seat-based or feature-based models, since they fail when value is tied to what an AI agent accomplishes, not how many people log in.”
Within such a framework, vendor revenues are derived from “outcome-based models, such as paying for a percentage of savings or gains,” Mansard added. “Outcome-based offers a fairer model, aligning incentives and transferring performance risk to the software provider. For example, Zendesk charges its customer service agentic AI $1.50 per actual case resolution.”
At this time, fewer than 10% of AI services are monetized this way, Mansard continued, “given how complex it is to both demonstrate and operate. For the foreseeable future, we’ll see a spectrum of models, such as by activity – AI work hours — or by output – AI deliverables.”
The ramifications of post-SaaS go well beyond pricing, said Dor Sasson, CEO and cofounder of Stigg.io, a software billing service. “It’s a permissioning problem. In the SaaS world, humans bought access: seats, credits, outcomes. In the agentic world, software buys autonomy. AI agents don’t care about seat limits or usage tiers; they care about the license to act, to read data, execute workflows, and make decisions inside systems.”
For that reason, per-seat or even per-usage pricing won’t work in this new era. “AI agents don’t sign up. They transact,” said Sasson. “The replacement isn’t just another monetization hack; it’s a new contract between vendors and software itself – agentic licensing.”
However, agentic licensing won’t be appearing overnight, warned Bryan Murphy, CEO of Smartling, a content translation and globalization service. “The transition is messier than the frameworks suggest. The real challenge isn’t the pricing model itself, it’s that most SaaS companies bolted AI onto products designed for a different era.”
Murphy added that he’s been through enough technology waves – from on-premise to cloud to SaaS – “to recognize when the industry is hitting an inflection point, and this one’s different. Most enterprises are stuck in this awkward middle ground: they know per-seat pricing doesn’t work when agents are doing the work humans used to do, but switching to consumption or outcome-based models requires rethinking your entire cost structure and ROI calculations.”
Preparing for the post-SaaS agentic era “means rearchitecting your platform from the ground up, retiring legacy features that don’t make sense anymore, and accepting that your product roadmap from 12 months ago is now irrelevant,” Murphy explained. “For customers, the near-term reality means evaluating vendors on whether they’ve genuinely rebuilt for an agentic world or just repackaged legacy tools with a chatbot interface.”
To get a sense of vendor preparation, it’s important to ask vendors “how their pricing aligns with the actual value you’re getting,” Murphy advised. “If you’re paying per seat but agents are handling 60% of the volume, something’s broken. Push for transparency on how consumption is measured and what you’re actually paying for.”
The winners in this transition “will be customers who demand outcomes-based contracts and urge vendors to prove ROI in concrete terms, not theoretical productivity gains,” said Murphy.
