LLMs
Industry News
Reddit Is Using LLMs to Fight a Problem LLMs Largely Created
Reddit removes 100,000 bot accounts daily, yet licenses data to the AI firms whose models generate that same synthetic content back onto the platform.
•11 min read
Per-seat SaaS pricing assumes humans do the work. In the agent era, that costs you revenue. Here is the hybrid pricing playbook winning 2026 renewals.
Per-seat SaaS has a quiet assumption baked into the price card: a human is doing the work. As of 2026, that assumption is wrong often enough that hybrid SaaS pricing is the fastest-growing model on the renewal table. Gartner's 2026 forecast pegs hybrid pricing adoption at 40% by year-end, up from under 5% in 2025. Companies that moved to hybrid report roughly 38% higher net revenue retention than those still on pure subscription. Zendesk now bills $1.50 per Automated Resolution. Intercom is at $0.99 per resolved ticket. Microsoft just shipped Agent 365 at $15/user/month on top of an E7 suite that lists at $99. The category leaders have all priced agent output. The question is whether your product has — and whether your renewals are slowly leaking the difference.
This post covers why per-seat broke, what the three live pricing modes look like, and how to repackage in 90 days without spooking procurement.
Per-seat pricing has an implicit unit of work — one human, one workstation, one set of decisions per day. That worked when SaaS was a productivity multiplier for human users. It breaks the moment software does the work itself.
Two things shifted in parallel this year. Agents got good enough to run multi-step workflows without human approval at every step — Deloitte's 2026 prediction has enterprise applications with embedded task-specific agents rising from under 5% today to 40% by year-end. And the cost side flipped: pure AI-first SaaS gross margins are running 50–60%, against 60–70% for the rest of public SaaS, because every interaction has a token bill attached.
Net effect: per-seat structurally undercharges power users and overcharges light ones. Agents widen the gap, because a power user's agent now triggers 50x the value-delivering work without buying 50x the seats.
Three modes are live in the market right now:
Per API call, per token, per action. Predictable cost-to-revenue mapping, hard for procurement to budget. Best fit: developer tools and infrastructure SaaS where customers already think in volumes.
Per result delivered. Zendesk's Automated Resolution model is the cleanest live example. Best fit: support, sales, and anywhere the outcome is unambiguous enough to bill cleanly.
Fixed platform fee plus variable consumption or outcomes. Futurum's 2026 survey shows 43% of buyers prefer consumption-based and 27% prefer outcome-based — and most want a fixed floor.
This is no longer theoretical. The category leaders have published price cards:
The pattern: the floor is a platform fee procurement can plan; the ceiling is consumption or outcomes that scale with delivered value. The highest-valued SaaS companies in 2026 share three properties — real agent capability, consumption or outcome pricing, and NRR above 110%.
Before any team rewrites their price card, walk through these honestly:
Most teams blow up renewals because they try to flip the whole book at once. Don't.
Days 1–30: Instrument. Before changing a price, get clean telemetry on the unit of value you plan to bill on. If you can't measure it today, you can't bill on it in 90 days.
Days 31–60: Pilot a usage rider on new logos. Sell new customers a fixed platform fee plus a metered consumption rider. Watch deal velocity and the procurement conversation. Don't touch existing renewals yet.
Days 61–90: Migrate renewals selectively. Start with customers materially over-consuming on per-seat and ones who explicitly asked for usage-aligned pricing. Leave the rest for one more cycle.
Two mistakes to avoid: leading with outcome pricing before you can measure the outcome, and waiting one more cycle while competitors quietly repackage around you.
Build a SaaS pricing model that scales with usage and outcomes.
Continue exploring these related topics
Reddit removes 100,000 bot accounts daily, yet licenses data to the AI firms whose models generate that same synthetic content back onto the platform.
H100 on-demand rental rates are back down to roughly $2 to $2.53 an hour as of June 2026 That's a level last seen before the AI boom priced the same hardware at $8 an hour.
June 30, 2026, Anthropic released Claude Sonnet 5 and pitched it as the model that finally makes cheap, reliable agents the default. The claim underneath the launch is simple: near-Opus intelligence at a Sonnet price.