Ghost Seats: The $420/Month AI Waste Pattern Nobody's Tracking

The number your vendor dashboard hides from you

Enterprise AI licensing follows the same pattern as cloud over-provisioning: IT buys seats for everyone. Most employees use them occasionally. Power users hit the ceiling while others sit idle.

The industry average is 35% seat waste. You are paying for licences that generate no value.

We call these Ghost Seats: AI licences with less than 10% capacity utilisation over a rolling 30-day window, still billing at full plan rate.

In our demo dataset of 10 enterprise AI seats, 4 qualify as ghost seats. Monthly waste: $180. Annual waste: $2,160. For a single 10-seat deployment.

Scale that to 500 seats across a mid-market organisation and the recoverable spend becomes a board conversation.

Why this is invisible in your current tooling

Your vendor dashboard reports active licences — seats that exist in the system, regardless of whether anyone used them. PromptKing reports active usage — actual token consumption as a percentage of plan capacity, per seat, rolling 30 days.

Those are two very different numbers. The gap between them is your ghost seat exposure.

The five-step ghost seat audit

Step 1: Pull seat utilisation rate per licence Actual token consumption divided by plan capacity, measured over the last 30 days. Any seat below 10% is a ghost seat candidate.

Step 2: Confirm over two consecutive periods A single low-utilisation month could be holiday coverage or project downtime. Two consecutive periods below 10% is a reliable signal.

Step 3: Cross-reference against the seat owner Who owns this licence? Are they still at the company? Have they switched to a different AI tool? Is the seat shared informally across a team?

Step 4: Calculate the rightsize impact A ghost seat on a Max 5× plan ($100/month) can be downgraded to Pro ($20/month) or deprovisioned entirely. No capability loss for the user. Immediate cost reduction.

Step 5: Build the audit cadence Ghost seats accumulate over time as teams grow, projects end, and employees leave. A monthly utilisation review prevents the problem from compounding.

What PromptKing shows

The Seat Fleet dashboard surfaces ghost seat count and recoverable spend in real time across all six AI vendors — Anthropic, Microsoft 365 Copilot, GitHub Copilot, Google Gemini, AWS Bedrock, and IBM Watsonx.

The savings page shows exactly which seats are overpaying and by how much, with one-click downgrade recommendations and confidence scores.

In the demo org: 4 ghost seats, $420/month recoverable, $5,040/year. The FinOps program pays for itself before the second invoice.

The terminology that matters

Seat utilisation rate: Actual token consumption as a percentage of plan capacity, measured per seat over a rolling 30-day window. The primary rightsizing signal.

Recoverable spend: AI budget that can be freed through rightsizing without any reduction in capability for active users.

Ghost seat threshold: Less than 10% utilisation for two consecutive 30-day periods. High-confidence downgrade candidate.

---

PromptKing connects to your AI vendor accounts, pulls per-seat usage data nightly, and surfaces ghost seat recommendations with confidence scores. Free to start at promptking32.com.