GitLab, Palo Alto Networks Fold AI Agent Usage Into Seat Contracts

GitLab Inc. (GTLB) and Palo Alto Networks, Inc. (PANW) both reported on September 1, 2026 that they are folding AI agent usage into contracts customers already hold, rather than pricing agents on a separate meter. GitLab replaced its seat-only subscription with a fixed dollar commitment customers can reallocate monthly between seats and consumption credits. Palo Alto Networks is selling agent identity control as a paid upgrade to its installed privileged-access-management (PAM) base rather than as a standalone per-agent product. Both companies say their human seat counts are still growing, not shrinking.

Why Seat Contracts Can't Track Agent Usage

The standard enterprise-software subscription has customers forecast a number of human seats a year in advance and sign a contract around that number, consumed steadily over twelve months. That process assumes the population needing system access changes slowly and that usage growth can be planned on an annual cycle. AI agents break that assumption: the work agents generate can scale week to week, and buying agent capacity as its own line item would force customers through a fresh annual procurement cycle — one vendors risk losing if it moves too slowly.

GitLab and Palo Alto Networks both chose to keep the existing contract container and expand what it covers instead of creating a new one. GitLab replaced its subscription with a fixed-dollar commitment called Flex, letting customers redeploy budget that used to be lost to renewal contraction — because a customer had over-forecast its seat count — into consumption products instead, without a new procurement round. At the same time, GitLab says more non-engineering employees are requesting seats, and each of those new human users becomes a future buyer of consumption credits, so the human seat count and machine usage expand inside the same contract. Palo Alto Networks' approach is similar: its new Modern PAM product is packaged as an upgrade for customers who already own the traditional PAM system it acquired with CyberArk, priced under the existing PAM relationship rather than sold as a new per-agent purchase. In both cases, what changed is the shape of the contract, not the underlying customer relationship.

Flex Adoption Numbers and Agent Traffic Growth

Three data points support this reading. GitLab CEO Bill Staples described the shift in direction on the earnings call: the business started the year monetized entirely through seat subscriptions, and management now expects to exit the year with seats continuing to grow alongside a meaningful and scaling consumption business [1]. CFO Jessica Ross disclosed the pace of adoption: in the first six weeks after launch, more than 130 customers committed over $20 million to Flex, and paid consumption run rate ended the quarter above $40 million, up from $15 million in the prior quarter [1]. At Palo Alto Networks, management disclosed that agent-driven traffic on its SASE platform has surged 9x over the past nine months, while the company launched Modern PAM as a paid upgrade to the traditional PAM base it acquired through CyberArk, extending an existing customer relationship rather than opening a new one [2].

Reported Revenue Lags Operating Metrics

This shift changes vendors' internal operating metrics before it shows up in reported revenue. GitLab disclosed that roughly 15% of a traditional self-managed license is recognized upfront, while Flex licenses are no longer recognized upfront under the old rule — management expects this timing change to reduce reported FY2027 revenue by up to $13 million, and it is part of why the Q3 guide sits below the Q2 print [1]. Meanwhile, operating metrics such as paid consumption run rate and net retention already reflect the effect of the expanded contract: dollar-based net retention rose sequentially to 117% this quarter [1]. At Palo Alto Networks, agent identity management revenue is reported inside the existing access-management product line rather than broken out as a separate agent-security business, which will make it harder to see the incremental agent contribution directly from the financial statements. Tracking paid consumption run rate, net retention, and revenue growth in Palo Alto Networks' access-management line that now includes Modern PAM will be needed to judge whether this contract structure is actually expanding vendors' share of customer spending.

Companies this change could affect:

• Okta, Inc. (OKTA): An identity-management vendor facing the same agent-pricing question. Management said on its own earnings call that agent governance features are still sold as an uplift on the per-user charge — the same logic of folding new functionality into an existing contract [3].
• Atlassian Corporation (TEAM): A collaboration-software vendor that, 26 days before GitLab's call, had already bundled AI usage credits into its existing Teamwork Collection and was building a dollar-based commitment model usable across multiple product lines — an approach close to GitLab's Flex [4].
• JFrog Ltd. (FROG): A software-delivery tools vendor that charges based on the binary traffic customers generate. Usage from agents writing code and running pipelines shows up directly in JFrog's cloud revenue growth and net dollar retention [5].

Sources

[1] Drillr · GitLab Inc. (GTLB) · 2026-09-01 · FY2027 Q2 earnings call

"We entered this year with a business monetized entirely through Seed subscriptions. and we now expect to exit it with seats continuing to grow alongside a meaningful and scaling consumption business."

(A transcription error renders "seat" as "Seed" in the source transcript.)

[2] Drillr · Palo Alto Networks, Inc. (PANW) · 2026-09-01 · FY2026 Q4 earnings call

[3] Drillr · Okta, Inc. (OKTA) · 2026-08-26 · FY2027 Q2 earnings call

[4] Drillr · Atlassian Corporation (TEAM) · 2026-08-06 · FY2026 Q4 earnings call

[5] Drillr · JFrog Ltd. (FROG) · 2026-08-06 · Q2 2026 earnings call

This is only meant to help you notice industry changes and companies that may have been overlooked - not a stock recommendation.