SolutionsToken IQ (spend attribution)

See who spent your AI budget, on what, and whether it was normal.

Token IQ attributes AI agent spend to people, teams, models and kinds of work, and flags the days that fall outside your own pattern.

VP Engineering
Book a demo ↗︎

AI unit prices are falling and AI bills are rising.

83%

of enterprises have AI coding assistants or agents in production.

a16z CISO Survey, 2026
5–30×

more tokens per task for agentic models than for a standard chatbot.

Gartner
10% → 65%

About 10% of users account for roughly 65% of token use.

McKinsey, internal usage data
52%

of organizations have no dedicated owner for AI cost.

Harness
Who spent it

Spend attributed to agents, models, people and teams.

Pick a dimension to see who and what drives the total. Every figure is priced from observed tokens at your negotiated rates.

claude-code81.6%$152,072
codex14.0%$26,141↑232%
26 others4.4%$8,057

Codex grew 232% on the prior month.

On what

Spend by the kind of work it paid for.

Each session is classified by the work it did, so you can see what each kind of work costs and how that is changing.

Feature33.6%$62,611↑60%
Bug fix13.6%$25,301↓6%
Ops & deploy13.2%$24,555↑123%
Code review12.5%$23,227↑28%
Investigation11.7%$21,763↓58%
9 more kinds of work15.4%$28,813
Was it normal

Unusual spend, flagged before the invoice.

The detector compares each day with your own pattern and flags only the days outside it. A meter tracks spend against your budget and alerts the owner in Slack or Teams.

Typical day $6,010
Sep 9Sep 16Sep 23Sep 30Oct 8
Sep 21Needs verification$15,393

The only day outside this organization’s pattern.

Sep 17Within pattern$13,000

High, but inside the normal range.

Sep 18Within pattern$10,100

Above a typical day, inside the normal range.

Get started

15 minute installation. No blind spots.

See Token IQ on your own AI spend: who spent it, on what, and what to act on this week.

Runs in your environment. Your usage data never leaves the perimeter you govern.