Will Salesforce disclose evidence of per-seat subscription revenue declining while AgentForce grows by Q2 FY2027?

Q4 FY26 earnings comprehensively undercut every YES pathway. Pathway (1) management acknowledgment: management did the opposite -- explicitly confirmed seats growing YoY and QoQ, and framed AgentForce as driving PREMIUM seat upgrades (tripled QoQ). Pathway (2) sub-5% subscription growth: subscription growth ACCELERATED to 10%+ with FY27 guided at ~12% CC -- the trajectory is moving away from the 5% threshold, not toward it. Pathway (3) seat count declines: 7 of top 10 deals added seats, and Salesforce may eliminate per-cloud disclosure entirely, removing the detection mechanism. The ILA model bundles seats + consumption credits, making them complementary. Two remaining earnings calls (Q1 FY27 June, Q2 FY27 August) offer fewer opportunities than the three previously assumed. The prior 14% was appropriate pre-earnings but this data requires a substantial downward revision.

Evaluating the updated math: AgentForce at $800M ARR is ~2% of revenue. Even if it grows 50%+ by Q2 FY27 (reaching ~$1.2-1.5B, ~3-4% of revenue), it cannot mathematically depress subscription growth from 12% guided to below 5% -- that would require AgentForce to cannibalize $2.5B+ of subscription revenue, which is 2-3x its total ARR. The premium seat upgrade pathway (tripled QoQ) is the key new data point: AgentForce is driving MORE seat revenue, not less. The remaining risk is a careless management comment during Q&A, but Q4 FY26 transcript showed extremely disciplined messaging -- Milano and Benioff reinforced the additive narrative multiple times. Revenue by cloud disclosure elimination further reduces pathway (3). Setting at 9% to account for residual risk of analyst probing across two remaining calls.

The strongest remaining YES pathway is criterion (1): an analyst corners management on per-seat economics at Q1 or Q2 FY27 earnings. The AELA mechanism (customers converting headcount savings into AI credits) is real and accelerating. But Q4 FY26 shows management has now developed a polished counter-narrative: seats are growing because AI makes each seat more valuable, driving premium upgrades. This is a sophisticated reframe that makes accidental acknowledgment less likely. However, if AgentForce reaches $1.5B+ ARR by mid-2026 and macro conditions pressure enterprise spending, the narrative could crack. Two calls give ~3-5% trigger probability each. I weight this slightly higher than other opus runs because the disclosure elimination may paradoxically INCREASE analyst scrutiny on the metrics that remain disclosed.

Q4 FY26 data is a near-complete refutation of the seat cannibalization thesis within the resolution timeframe. Every metric moved in the wrong direction for YES: subscription growth accelerated, seat counts grew, premium upgrades tripled, and management guided FY27 subscription growth to 12%. The resolution criteria require EITHER explicit management acknowledgment, OR sub-5% subscription growth, OR disclosed seat declines. None of these pathways are plausible given the current trajectory. Management would need to completely reverse their narrative in two earnings calls, which would be value-destructive and is not in their interest. Setting at 6% as a floor for residual uncertainty.

The previous prediction was 14% with three earnings calls and uncertain trajectory. Now we have: (a) one fewer earnings call in window, (b) subscription growth accelerating not decelerating, (c) explicit confirmation that seats are growing, (d) data showing AgentForce drives premium seat upgrades. This is a fundamental shift in the evidence base. The question asks about DISCLOSURE of evidence -- even if some pockets of seat pressure exist in specific clouds, the aggregate numbers are so strong that management can truthfully claim overall seat growth. The only realistic YES scenario is a macro shock causing enterprise spending cuts that force management to explain subscription deceleration, combined with AgentForce growth making it obvious the substitution is happening. This is a compound low-probability scenario.

I want to avoid anchoring too heavily on the positive Q4 data and consider what could still trigger YES. Scenario: macro downturn hits in H1 2026, enterprise customers freeze hiring, seat counts plateau even as AgentForce substitutes for incremental seats that WOULD have been purchased. In this scenario, management might acknowledge that 'AI capabilities are helping customers do more with existing seat counts' -- which is borderline criterion (1). The revenue by cloud elimination could also signal that management anticipates cloud-level weakness they want to obscure. These are speculative but not impossible. Still, guided 12% subscription growth and confirmed seat increases make even this scenario unlikely to produce the specific resolution criteria within 6 months.

The data is overwhelming. Subscription growth accelerating to 12% guided. Seats growing YoY and QoQ. 7 of top 10 deals added seats. Premium upgrades tripled. AgentForce at 2% of revenue is too small to move the needle. Only 2 earnings calls left. Management has no incentive to disclose negative seat data. Strong NO.

All three resolution pathways face even higher barriers post-Q4 earnings. Management acknowledgment: they explicitly said the opposite. Sub-5% growth: guided to 12%, would need to miss by 7pp. Seat count declines: 7/10 top deals added seats, and cloud disclosure may be eliminated. Prior was 14% but every new data point pushed toward NO. Calibration check: sibling markets showed we were too bearish on revenue growth and cRPO -- the business is stronger than pre-earnings expectations, making cannibalization disclosure even less likely.

Mostly aligned with other haiku runs but giving slightly more weight to tail risk. AgentForce ARR grew 48% QoQ -- if it maintains that pace it could reach $1.5-2B by Q2 FY27 (~4% of revenue). At that scale, analysts will increasingly probe the substitution question. But management has now established a strong counter-narrative with actual data (seat growth, premium upgrades) that will be hard for analysts to crack. Residual risk mostly from unforeseen macro events or a management slip.