The deal doesn't
close in the meeting.
It closes before it.

Snowflake reported $6.9 billion in remaining performance obligations. That number gets cited as proof of product-market fit. I read it differently. That's $6.9 billion in enterprises that are now locked into a negotiation they can't win at renewal.

"Finance can negotiate better deals. Operations determines whether you capture the value."

Here's the trap. A big cloud commitment looks like strategic alignment. It's actually a slow transfer of pricing power from you to the vendor. You've pre-committed the budget. You've eliminated your ability to walk away. You've told them, in writing, exactly how much you need them.

How it happens

Year one: the champion gets a great deal. Vendor gives aggressive discounts to win the commitment. Everyone celebrates. The press release goes out.

Year two: usage is up. The champion got promoted. The new team doesn't know the original negotiation history. They're managing the contract, not the relationship.

Year three: renewal. The vendor knows your switching costs better than you do. They know your engineers are trained on their stack. They know your data is in their warehouse. The discount evaporates. The price goes up 30%. What are you going to do—migrate?

What I've seen work

The enterprises that maintain leverage do three things. First, they never pre-commit more than 60% of anticipated usage. The uncommitted 40% is negotiating power, not slack. Second, they maintain a live competitive option—an actual alternative, not a threat. Vendors can tell the difference. Third, they separate the technical champion from the commercial relationship. The person who loves the product should never be the person who signs the renewal.

The commitment trap is a feature, not a bug, of the enterprise SaaS business model. Your job is to know you're in it and manage accordingly.

Your operations team gathers every morning to stare at dashboards. Red, yellow, green. Numbers that describe what happened yesterday. Decisions that will affect tomorrow, made on data that's already stale.

"Graphs at 8am are pointless if the software showing them didn't stop the cash fire at 7pm."

The dashboard was the right tool for the information age. You couldn't act on data in real time, so you batched it, visualized it, and made humans interpret it in meetings. That was the best available option.

It is no longer the best available option.

What autonomous agents actually do

An agent doesn't show you that your churn rate increased 4% last month. An agent notices at 11pm on a Tuesday that three enterprise accounts haven't logged in for eight days, cross-references that with open support tickets, identifies the one that has a renewal in 47 days, and sends a flagged summary to the account manager at 7am Wednesday with a suggested intervention.

By the time your team is looking at the dashboard Friday morning, that account has already been called.

The control plane shift

The executives who understand this are quietly rebuilding their operations stack. Not replacing humans—repositioning them. The humans set the rules. They define what "good" looks like. They make the judgment calls agents can't make. The agents handle the surveillance, the routing, the first response.

The dashboard doesn't disappear. It moves to the exception layer. You look at it when something the agent couldn't handle escalates to you. That's it.

The enterprises still building dashboards as their primary intelligence layer in 2026 are building yesterday's solution. The control plane has already moved.

The seat-based model built the last decade of enterprise software. Salesforce, Workday, ServiceNow—all of it priced per user, per month. It was clean. Predictable. Easy to budget. Easy to sell.

CIOs aren't buying it anymore. Not for AI. And eventually, not for anything.

"You're not selling software. You're selling what the software does to my business."

The four gaps vendors are still missing

Gap one: usage ≠ value. Two hundred seats doesn't mean two hundred people getting value. It means two hundred licenses. The CIO knows this. The vendor pretends not to.

Gap two: AI work isn't human work. An AI agent doesn't have a seat. It does the work of twelve people. Pricing it per seat is like charging per employee for electricity.

Gap three: outcomes are measurable now. For the first time in enterprise software history, you can actually measure what the software did to the business. Revenue influenced. Time saved. Errors caught. CIOs want to pay for that. Vendors don't know how to sell it.

Gap four: procurement has changed. CFOs who lived through three years of SaaS sprawl are done with open-ended subscriptions. They want proof of ROI before renewal. Vendors who can't show it are going to lose renewals they thought were automatic.

The vendors who figure out outcome-based pricing in the next eighteen months will own the next decade. The ones who defend seat pricing will defend it right up until they can't.

Marc Benioff didn't build a CRM. He built a platform you can't leave. The data model, the integrations, the custom objects, the trained-on-Salesforce administrators—all of it is infrastructure for a cage. A very comfortable, very expensive cage.

"You should have a data strategy that ensures you don't have to trust anyone."

The enterprises I've watched navigate vendor concentration well share one trait: they treat their data layer as sovereign territory. The application sits on top. The data belongs to them.

What lock-in actually costs

Not the migration cost. That's visible. The invisible cost is the deals you don't negotiate because you can't walk away. Every renewal you enter without a credible alternative is a negotiation you've already lost. The vendor knows it. You know it. The question is who blinks first—and it won't be them.

The three moves that preserve optionality

First: own your data in a format you control. Before you sign any major platform contract, answer this question: if we had to leave in ninety days, what would we lose that we couldn't reconstruct? If the answer is "everything," you're already locked in.

Second: maintain at least one credible alternative in each major technology category. Not as a threat. As a real, evaluated, partially-piloted option. Vendors have internal notes on which clients can actually leave. Be one of them.

Third: separate the technical relationship from the commercial one. Your engineers shouldn't be negotiating renewals. They love the product. That's their job. Get someone in the room who doesn't.

Every enterprise I've walked into in thirty years has had a version of the same belief: our people are better than the market. Our engineers are sharper. Our strategists are deeper. Our process is more rigorous.

Some of them were right about their people. Almost none of them were right about the outcome.

"Intelligence inside an organization without intelligence about what's happening outside it is just expensive confusion."

The best deals I've been part of—on either side of the table—happened when one party had asymmetric information. They knew something the other didn't. Not because they were smarter. Because they were watching different things.

What this means for enterprise buying

When you're evaluating a vendor, the person sitting across the table has spent the last three years talking to every one of your competitors about this exact problem. They know what your competitors tried. They know what failed. They know what they're about to try next.

You have one perspective. They have hundreds.

The smart enterprise buyer uses that. They stop treating vendor conversations as pitches to be deflected and start treating them as competitive intelligence to be extracted. Ask the right questions. The answers are worth more than any RFP response.

What this means for enterprise selling

Your smartest salesperson isn't as smart as they think they are about the buyer's internal politics. Stop guessing. Ask. The buyers who will tell you what's really happening inside their organization are the ones worth spending time on. They're also usually the ones who close.

The number is down. The board is asking questions. The CEO is in the fourth meeting this week about pipeline. Someone in that room is going to say it: maybe we need a new head of sales.

I've watched this movie play out more times than I can count. New head of sales, same product, same market, same result. Six months later, a slightly different framing of the same problem.

"You can't hire your way out of a product-market fit problem. You can only delay figuring out what the problem actually is."

How to tell the difference

Three questions. First: are deals starting but not closing, or are they not starting at all? If buyers won't take the first meeting, the problem is upstream of sales. If they take meetings but don't buy, you might have a sales problem.

Second: what do the closed-lost deals say in the exit interviews? Not "too expensive" or "went with a competitor"—those are the polite answers. What's the real one? If it's "we couldn't get budget approved internally," that's a positioning problem. The product didn't map to a line item anyone controlled.

Third: what do your best customers love about the product? Is what they love the thing you're selling? Often it isn't. The feature that closes deals is different from the feature that retains customers. Enterprise sales that builds around retention features instead of conversion features is always going to underperform.

Solve the product-market fit problem first. Then optimize the sales motion. In that order. Never the other way around.