Cost Gets You In. Usage Keeps You.

Glean announced $300M ARR on May 28, fifteen months after crossing $100M. Fortune 500 customer count nearly doubled year over year. That happened while Microsoft, Google, OpenAI, Anthropic, Salesforce, and Atlassian were all shipping something aimed at the same job.

Everyone read that as a growth story. I think the growth is the least interesting part.

The interesting part is what Glean was saying at the register. The pitch that carried them through 2025 and into this year was not a feature comparison against Copilot. It was closer to: your AI bill is out of control, and we make it smaller. Arvind Jain has been explicit that the context graph reduces how many tokens a query burns, because the model gets the right internal context up front instead of grinding through everything. TechCrunch covered that repositioning when the number landed.

That works. It also has a shelf life, and I want to be honest about which part of it I'd copy.

The problem with selling savings

If your value is denominated in someone else's cost, your ceiling moves when their cost moves. Token prices have fallen roughly an order of magnitude a year for three years running. Sell a 30% reduction on a line item that keeps shrinking and you are running up a down escalator. You have to win more of a smaller thing every single renewal.

I wrote the Sierra piece about the opposite move: price against the outcome you produce, not the spend you avoid. Resolution pricing captures value created. Savings pricing captures a fraction of value destroyed. Those two posts sit next to each other on this site and the tension is deliberate, so let me resolve it.

The savings pitch isn't the value. It's the permission.

Look at the other numbers in the same press release

Glean also said more than 85% of customers use it across five or more departments, and that it holds a 45% wDAU/wMAU ratio, more than double the enterprise SaaS norm. Nearly half of monthly actives come back on a given day.

Nobody logs in daily because they're pleased with the procurement math.

That's the tell. Two different things are being measured and they've quietly come apart. The cost argument is what gets the purchase order signed, because a CFO in 2026 will approve a subtraction faster than an addition. The usage behavior is what makes anyone renew. Cost got them in the door. Return behavior is what kept the expansion compounding across departments, and department five is where a $60K land turns into a seven-figure account.

Most builders I talk to are instrumenting exactly one of these. Guess which.

The dashboard ends up matching the sales deck. Cost per query, tokens saved, hours reclaimed, all of it modeled, all of it defensible, none of it predictive of whether the thing survives the next budget cycle. Meanwhile the number that actually forecasts renewal, does a person come back tomorrow without being told to, sits uninstrumented because it never made it into the business case.

What I can't prove

I can't tell you the pitch caused the growth. Jain has said the company had essentially no competition for its first four or five years, and a category monopoly explains a lot of curve on its own. Run that as the alternate hypothesis and the numbers still work.

What the cost framing explains better is the acceleration after the giants arrived. Going from $208M at the end of 2025 to $300M by May, with Copilot bundled into licenses these accounts already own, is not something a features argument wins. "It's included in what you're already paying Microsoft" beats "ours is better" every time. It does not beat "your AI spend is $4M and nobody can tell you what it bought."

One more honest note on the number itself. TechCrunch flagged that Glean bills partly on consumption, so a chunk of that $300M is an annualized run rate rather than contracted recurring revenue. That's not a knock, it's a different machine. Consumption revenue expands with usage and contracts with it too, which is the whole argument here in accounting form. The usage layer isn't a soft metric for these guys. It's the invoice.

There's also a real tension inside their own pitch that I'd fix if I were shipping it: you can't fully sell "we lower your AI bill" while metering the customer per query. The claim that survives is consolidation. One context layer instead of six overlapping tools nobody fully adopted. That one doesn't deflate when inference gets cheap.

Where this doesn't transfer

Glean sits at the context layer, which is one of the few places in the stack that can measurably move inference cost. Most of us have no comparable line item to point at. If you're building a design tool or a vertical workflow app, there is no embarrassing invoice sitting on a CFO's desk with your name adjacent to it, and copying the savings pitch will get you a polite meeting and no PO.

What transfers is the split, not the wedge. Find whatever your buyer's actual permission structure is, and stop assuming it's the same thing as your value. Sometimes permission is cost. Sometimes it's risk, or a compliance deadline, or the fact that a competitor's logo showed up in a board deck. It is almost never the thing your users love you for.

The two-layer version

Run them as separate instruments with separate owners.

The buyer layer answers "why was this approved," and it needs to be re-provable at renewal in the same units the original case was made in. If you sold savings, you owe a savings number in twelve months, and if you don't produce one, someone else will produce a worse one for you.

The user layer answers "would these people be upset if it disappeared." Unprompted return rate is the cleanest proxy I know. Daily-over-monthly, cohorted, no email nudges counted. Departmental spread is the leading indicator underneath it, because a tool that escapes its original team has crossed from budget line to infrastructure, and infrastructure doesn't get cut in Q4.

Pick one thing this week: open your metrics review and mark every number as buyer-layer or user-layer. If everything falls in one column, you've been telling one story to two audiences, and only one of them is going to keep paying you.