From Dashboard to Skill Hire: The Death of Per-Seat Software
Layer Scoring
Sublayer Impact Map
Which of the 50 sublayers this case actually touches, and at what magnitude.
Intelligence Cube · 2D
Footprint across Functions × Verticals × Layers, the three axes that determine structural fate.
Layers × Verticals
5 cells · 5×1
Layers × Functions
25 cells · 5×5
Two 2D projections of the Intelligence Cube (Functions × Verticals × Layers). Filled cells = this move occupies that intersection.
Timeline
1999–2015
Era 1, The Dashboard. Salesforce, Tableau, Google Analytics. Per-seat is the natural pricing unit.
2015–2023
Era 2, The Workflow. Asana, Notion, HubSpot. Per-seat still aligns with humans-doing-work.
Nov 2022 → now
Era 3, The Dialogue begins with ChatGPT. Per-seat starts to crack as one human + AI replaces three humans.
2024–2025
Klarna's 700-agent disclosure and Sierra's L8-architected agent platform mark the first production Era 4 deployments.
2026–2028
Era 4, The Workspace becomes default in CX, support, sales ops, software engineering. Per-resolution and per-outcome pricing become standard.
2028+
Era 5, The Skill Hire. Agents are 'hired' for roles. Per-seat is structurally over. Pricing resembles labor contracts more than software.
- Who Wins
- Outcome-priced AI-native vendors. Per-resolution (Sierra), per-action (Klarna's internal model), per-deployment (Glean), pricing units that grow with usage as humans compress.
- L4 incumbents that re-price aggressively. Microsoft moving Copilot to per-message and per-agent rather than per-seat 365 add-ons. Adapting the pricing model is the harder act than building the product.
- Vertically-stacked plays with L1+L5+L8. Harvey, Bloomberg, Sierra, Glean, the architecture is exactly what Era 4 and 5 buyers want to pay for.
- Who Loses
- Pure per-seat SaaS with bolt-on AI. Pricing model self-defeats. The harder the AI works, the faster seat count falls. Every 2015-vintage workflow SaaS faces this.
- L7-only AI products. Era 4 absorbs them. The surface is the cheapest layer to replicate, every L4 owner ships it for free as the era turns.
- Workforce categories whose work is structured + reviewable. Customer support tier 1, software bug fixes, scheduling, invoice processing, basic legal review, content moderation. Era 4 hits these first.
- Steelman: The Counter-Thesis
The strongest counter is that AI capability gains will plateau or hit a regulatory ceiling before Era 4 truly arrives, leaving the world in an extended Era 3, humans clearly augmented but still in the loop, per-seat pricing still working because seat counts compress slowly rather than collapse. There is real evidence for this: model improvement is becoming more incremental, regulated industries (healthcare, legal, finance) require human-in-the-loop by law, and most enterprises have integration constraints that make full agent deployment a multi-year program. Honest read: Era 4 arrives unevenly. Customer support, software engineering, and L&D land in Era 4 by 2027. Legal, healthcare, and most regulated B2B remain in Era 3 well into the 2030s. The era thesis is directionally right; the dates are illustrative.
Most product organizations are still building Era 2 software in an Era 3 market and pricing it for an Era 1 buyer. That mismatch is what's actually breaking SaaS roadmaps right now, not "AI."
The five eras of software.
Era 1 (1999–2015), The Dashboard. Software shows data; the human decides everything. Salesforce reports, Tableau dashboards, Google Analytics. Value created by visibility. L7 (the surface) is the entire product. Per-seat pricing makes perfect sense, each seat is a pair of human eyes interpreting the data.
Era 2 (2015–2023), The Workflow. Software guides the human through a sequence; the human still executes. Asana, Notion, HubSpot sequences, modern CRM workflows. Value created by structured guidance. L5 (workflow) joins L7 as a product layer. Per-seat still works, each seat is a human moving through the workflow.
Era 3 (2023–now), The Dialogue. Human directs, AI builds. ChatGPT, Cursor, Copilot, Claude. The user states intent in natural language; the system generates the artifact. L2 (model) becomes a first-class product layer for the first time. Per-seat starts to crack, one human with AI can do the work of three without it, so seat counts compress even as the user base grows. We are here.
Era 4 (2026–2028), The Workspace. AI orchestrates a multi-step task across systems; the human supervises and approves. Klarna's CX agent. Sierra. Devin (in theory). Salesforce Agentforce. L5 + L6 + L8 become the dominant layers. Per-seat pricing collapses, the unit is the resolution, the task, the outcome, not the human seat. Pricing models migrate to per-action, per-success, per-outcome.
Era 5 (2028+), The Skill Hire. The agent IS the worker. You "hire" an agent for a role the way you hire a contractor, with a job spec, an outcome, an SLA. The human role moves up: strategy, governance, edge-case judgment, supervision of a fleet of agents. L1 + L5 + L8 dominate (proprietary data, real execution, compounding memory). Per-seat pricing is structurally over. The Skill Hire is priced like labor, by output, by retainer, by guaranteed result.
Why per-seat is dead. The math is brutal and simple. Per-seat economics assume more humans use the product → more revenue. AI makes the inverse true: better AI → fewer humans needed → fewer seats → less revenue. Every SaaS that's "AI-powered" with per-seat pricing is structurally exposed its own roadmap. The harder the AI works, the faster the customer's seat count falls. Pricing has to migrate to per-action, per-outcome, or per-deployment, any unit that grows with usage instead of compressing with productivity.
Why both axes matter. The Cube projection is the right lens here. The customer axis (which functions × which verticals you serve) determines TAM. The depth axis (which layers you own, L1, L4, L5, L6, L8) determines defensibility as the era shifts. Era 3–5 products that own only L7 + a rented L2 are surface plays, they will be absorbed by L4 owners. Era 3–5 products that own L1+L5+L8 are durable across the transition.
The structural read for product leaders.
• If you are building for Era 2 (workflow guidance) and your buyer is buying for Era 4 (outcome delivery), you will lose the contract to a startup that is.
• If you are pricing per seat and your customer's seat count is dropping because of AI, you are pricing your own decline.
• If your moat is L7 (a polished surface), an L4 owner will ship the same surface for free as a feature within 12–18 months.
• If your moat is L1+L5+L8 (data + execution + memory) in a focused vertical, the era transition makes you more valuable, not less.
The investor read. The per-seat-SaaS multiple compression isn't a sentiment shift; it's an architectural one. The companies trading at a premium across the Era 3→4→5 transition are the ones with proprietary data, real execution permissions, and compounding memory loops, Sierra, Glean, Cursor, Harvey, Bloomberg, Adobe Firefly. The companies under pressure are the ones with thin surfaces on rented models, most of the GPT-wrapper class and any 2018-vintage SaaS that bolted "AI" onto a per-seat workflow product without re-architecting.
Framework synthesis; era boundaries are illustrative, not strict dates. The transitions are gradients, not step functions.
What This Means for You
Product Leader
Audit your pricing today against the era you're actually building for. If your roadmap is Era 4 (AI orchestrates) and your pricing is Era 2 (per seat), you're pricing your own decline.
Investor
The SaaS multiple compression is structural, not sentiment. Underwrite which era a company is *architected* for, not which era they market in. Per-seat AI is structurally exposed.
Operator
When negotiating an AI vendor contract, push for per-outcome or per-resolution pricing. If they refuse, ask why, the answer reveals whether they are Era 3 or Era 4 architecture.
Anand Arivukkarasu
Ex-Meta product leader. Creator of Supply Chain of Intelligence™. Writes about where AI value accrues, and who can fire your product. LinkedIn
Read next
Harvey AI Through the LayersGet the next teardown in your inbox.
One issue when something structurally important happens, usually weekly. No spam, no filler, unsubscribe anytime.
Worth sharing? Pull-quote: "Per-seat pricing assumes more humans means more revenue. AI makes the inverse true. Every per-seat SaaS with AI features is structurally exposed its own roadmap."