How AI Agents Are Quietly Rewriting Who Buys Software Inside the Enterprise

Table of Contents

• The Old Buying Committee Was Built for Seats
• Who Gains Power When You Sell Outcomes
• The New Personas at the Table
• The Line-of-Business Operator
• The Finance Reclassifier
• The Risk and Security Gatekeeper
• Procurement Has to Learn a New Question
• The Champion Is No Longer the User
• What This Means for How You Sell GaaS
• Insights Most People Overlook
• References

The Old Buying Committee Was Built for Seats

For two decades, enterprise software sold to a predictable room. A department head felt a pain, an IT or admin team vetted the tool, security ran a questionnaire, procurement squeezed the per-seat price, and a VP signed. Everybody understood the unit being bought: a license, attached to a human, renewed annually. Gartner has spent years documenting how this committee swelled to six, ten, sometimes more stakeholders, and how that bloat is the single biggest reason big software deals stall. The average B2B buying group now runs to a dozen-plus people, each with veto power and almost none with full authority.

That whole apparatus assumed one thing: you were buying capacity for people to do work. Seats were a proxy for headcount. When a vendor said "we have 4,000 seats deployed at this account," everyone in the room knew what that meant operationally and financially.

Agents break the proxy. When the unit you buy is a completed task, a resolved ticket, a reconciled invoice, a booked meeting, the seat count stops describing anything real. You might have three humans supervising an agent that does the work of thirty. Or zero humans in the loop for an entire workflow. The buying committee that was optimized to negotiate seat counts suddenly has no idea what it's negotiating. That confusion is the story of enterprise buying right now, and it is reshaping who sits at the table.

Who Gains Power When You Sell Outcomes

The shift from seats to outcomes redistributes authority, and the redistribution is uneven in ways that matter for anyone selling GaaS.

The clearest winner is the line-of-business owner, the person who owns the P&L for a function rather than the IT budget. When an agent is priced per resolved support ticket, the head of customer support can reason about it the same way she reasons about a BPO contract or an extra headcount: cost per unit of work, against a known volume. She doesn't need IT to translate "27 seats" into business value because the value is already denominated in her own terms. This is a profound change. The economic buyer and the user are collapsing back into the same person, after fifteen years of SaaS pulling them apart.

The clearest loser, at least in the short term, is the IT procurement function that built its leverage on standardization and seat consolidation. "We'll move everyone to one vendor and negotiate volume" is a powerful play when the unit is seats. It's a much weaker play when each agent is bought against a specific outcome with its own ROI math, often by a different department, sometimes on a different cadence. IT doesn't disappear, but its veto softens from "you may not buy this" to "you must buy this safely."

McKinsey's work on the economic potential of generative AI makes the underlying force explicit: the value is concentrated in a handful of high-volume operational functions, customer ops, software engineering, marketing, sales, where work is repetitive enough to delegate to an agent and measurable enough to price by the outcome. Those are exactly the functions whose leaders now have a credible claim on the buying decision.

The New Personas at the Table

The committee doesn't shrink so much as it recomposes. Three personas become disproportionately important.

The Line-of-Business Operator

This is your new primary buyer, and you should sell to them directly. They think in unit economics, not feature lists. A VP of revenue operations evaluating a sales-development agent isn't comparing it to another piece of software; she's comparing it to the fully loaded cost of an SDR, roughly $80K to $120K a year in many US markets, against what the agent costs per qualified meeting. If your agent books meetings at a third of that cost with comparable quality, the math sells itself, and she can run the pilot on departmental budget before anyone in central IT hears about it.

The strategic consequence: bottoms-up adoption, which the a16z thesis on the future of enterprise sales has been calling for years, now happens with an even shorter fuse because the buyer can self-justify the spend with a number rather than a vibe. The land-and-expand motion that defined the best SaaS companies gets faster and more measurable in the agent era, see our companion piece on #298 - The "agent does the workflow the SaaS used to host" pattern for how this changes what you actually deliver.

The Finance Reclassifier

When an agent is sold per outcome and replaces labor, the CFO's office stops treating the spend as software opex and starts treating it as labor spend or cost-of-goods. That sounds like an accounting footnote. It is not. It changes which budget the money comes from, which approval thresholds apply, and how the purchase is scrutinized.

Software budgets are usually capped and fought over annually. Labor budgets flex with volume. If your GaaS offering can be funded out of the much larger labor or operations line rather than the constrained IT line, your addressable budget at that account just multiplied. The flip side: labor spend gets scrutinized differently. A CFO who would happily renew a $200K SaaS contract without blinking will ask hard questions about a $200K agent contract that's framed as replacing two FTEs, because now it's competing with headcount decisions, severance optics, and workforce planning. Smart GaaS vendors are learning to arm their champion with the framing that lands in finance: this is variable-cost capacity, not a fixed software commitment.

The Risk and Security Gatekeeper

Here's the persona that gains the most net-new power. When software was passive, it displayed data, a human acted, the security review was about data handling. When the software acts autonomously, taking actions in production systems, moving money, sending customer communications, the review is about agency itself. What can this thing do without a human? What's the blast radius if it does the wrong thing? Who's liable?

This is why agent reliability and agent security have become first-class buying criteria rather than afterthoughts, and why a CISO can now kill a deal that the business owner is desperate to close. The gatekeeper's questions are different too: not "where does the data live" but "what are the guardrails, what's the audit trail, can we set spend and action limits, what's your incident track record." Vendors who treat this as a checkbox lose. Vendors who lead with observability, permissions scoping, and a clean liability story get through the door faster.

Procurement Has to Learn a New Question

Traditional procurement is exceptionally good at one question: how do we pay less per unit for the same thing? That muscle is nearly useless when the unit is an outcome whose quality varies.

If you buy 1,000 resolved tickets a month, the resolved-ticket price is only half the story. The other half is: resolved to what standard? An agent that closes tickets cheaply by marking hard ones "resolved" and dumping them back on humans is worse than useless. So procurement has to develop competence it never needed before, defining the acceptance criteria for an outcome, building SLAs around quality and not just uptime, and structuring contracts where payment ties to verified results. This is closer to how you'd contract a BPO or a marketing agency than how you'd buy a software license, which is exactly why the outsourcing and BPO playbook is becoming relevant again (covered in #302 - BPO and outsourcing firms vs. agent platforms).

The mature version of agent procurement looks like outcome contracts with clawbacks, quality thresholds, and human-escalation guarantees baked in. The immature version, which is most of the market today, is a line-of-business owner expensing an agent on a credit card with no SLA at all, then discovering the quality problem in production. Procurement's real job in the agent era isn't to drive the price down. It's to make the outcome legible enough to hold the vendor accountable. That's a genuinely new function, and the companies that build it first will buy agents far more effectively than their competitors.

The Champion Is No Longer the User

One of the quieter shifts: in classic SaaS, your champion was usually a power user who loved the product and lobbied internally to expand it. Adoption and advocacy came from the people using the tool every day. With agents, the heaviest "user" might be no human at all, the agent runs the workflow, and the humans supervise exceptions.

So who champions the purchase? It's the manager whose metrics improved. The support lead whose backlog dropped, the AR manager whose days-sales-outstanding fell, the marketer whose campaign throughput tripled. They champion the result, not the interface. This means the demo that wins is not a slick UI walkthrough, it's a before-and-after on a metric the manager is measured on. It also means your expansion path runs through operational dashboards and QBRs framed around business outcomes, not through seat-expansion nudges and feature adoption emails.

This is one reason the disappearing dashboard phenomenon matters to the buyer relationship (explored in our broader cluster): when the agent acts instead of displaying, the proof of value has to be delivered as a reported outcome, because nobody is staring at a screen all day to feel the product working. The vendor that controls the outcome reporting controls the renewal narrative.

What This Means for How You Sell GaaS

Pull these threads together and a concrete go-to-market emerges for anyone selling agents as a service.

Sell to the P&L owner first, because they can self-fund a pilot in business terms and don't need IT to translate value. Lead the pilot with a single metric they're already accountable for, and instrument it obsessively so the before-and-after is undeniable. Arm that champion early with the finance framing, variable labor capacity, not fixed software, so the spend lands in the right budget and survives CFO scrutiny. Bring security in proactively rather than waiting to be ambushed; lead with guardrails, audit trails, action limits, and a real answer on liability, because the risk gatekeeper now has a kill switch they didn't have before. And structure the contract like an outcome agreement with quality thresholds, not a seat license, because that's the only structure procurement can eventually hold you to and the only one that survives the seat-count question being meaningless.

The vendors who keep selling agents the way they sold seats, to IT, by license count, with a UI demo, are pitching a buying committee that no longer holds the power. The room reorganized. The sooner your motion matches the room, the faster you close.

Insights Most People Overlook

The "shadow agent" problem will dwarf shadow IT. Shadow IT meant employees signing up for SaaS tools without approval, annoying but mostly contained to data and spend leakage. Shadow agents are worse because they take autonomous actions in real systems. A line-of-business owner who expenses an agent on a card isn't just creating a data-governance gap; they're deploying an autonomous actor with no audit trail, no action limits, and no liability framework. The very bottoms-up buying motion that makes GaaS spread fast is the same motion that creates ungoverned autonomous agents inside the enterprise. The companies that win long-term will be the ones whose central function gets ahead of this with a sanctioned-agent catalog before the unsanctioned ones cause an incident.

Outcome pricing quietly transfers risk to the vendor, and many vendors haven't priced it in. Per-seat pricing is beautiful for a vendor because revenue is decoupled from whether the product actually works, you pay for access, not results. Per-outcome pricing re-couples them. If your agent's quality dips, you don't just churn, you literally earn less per unit this month. Most GaaS startups racing to outcome pricing to win deals haven't modeled what a reliability regression does to their revenue, not just their renewals. Outcome pricing is a reliability tax, and the vendors selling it are underwriting their own quality.

Procurement's loss of power is temporary and will reverse hard. Right now procurement looks sidelined because business owners are buying agents directly. But the moment a few high-profile agent failures hit, a pricing agent that mispriced, a comms agent that emailed the wrong customers, procurement and risk will come roaring back with agent-specific governance frameworks, and they'll be stricter than anything seat-based software ever faced. The current buyer-power vacuum is a window, not a new equilibrium. Vendors who exploit the window without building the governance story will get locked out when it closes.

Finance reclassification cuts both ways on deal size. Everyone notices that moving spend from the capped IT budget to the larger labor budget expands your addressable dollars. Fewer people notice that it also subjects you to workforce-planning politics. When your agent is framed as replacing FTEs, your renewal now depends on HR dynamics, union considerations, and executive optics about layoffs, variables that have nothing to do with your product quality. Some smart vendors are deliberately positioning as "augmentation that lets the team do more" specifically to keep their spend out of the headcount-reduction conversation, even when the economics are really about displacement.

The buyer who benefits most is often invisible during the sale. The agent's real beneficiary is frequently a frontline manager two levels below the signer, whose backlog quietly evaporated. They have no formal role in procurement, so vendors ignore them, but they generate the internal testimonials that drive expansion. Mapping and cultivating these invisible beneficiaries, the people whose Monday got better, is becoming a more reliable expansion lever than courting the executive sponsor who signed but never feels the product.

References

• Gartner: The B2B Buying Journey
• McKinsey: The economic potential of generative AI: The next productivity frontier
• Andreessen Horowitz (a16z): The Economic Case for Generative AI and Foundation Models