What an Engineering Org Actually Spends on AI Tooling

Per-token pricing pages tell you what a model costs. They do not tell you what an engineering organisation ends up paying once seats, APIs and shared infrastructure are all running at once.

This piece frames one calendar month of AI tooling spend across a 212-engineer organisation.

Read this first. Figures are an illustrative cost model built for editorial framing, not audited finance data. Replace with exported billing records, seat counts and token usage before treating any number here as a measurement.

---

The totals

Across 212 engineers over one calendar month, AI tooling spend comes to $18,900, or about $89 per engineer.

Of that total, $17,400 is attributed directly to roles as seat licences and token usage. The remaining $1,500 is shared org-level spend — self-hosted model infrastructure, shared API keys and smaller utilities — that is not allocated to any single role.

That keeps usage high enough for real productivity gains without AI subscriptions becoming a major line item against engineering salary cost.

Figures are an illustrative cost model built for editorial framing, not audited finance data. Replace with exported billing records, seat counts and token usage before treating any number here as a measurement.

---

Spend by role

Role	Headcount	Seat spend	Token/API spend	Total monthly	Per engineer
Product engineers	92	$3,680	$2,760	$6,440	$70
Full-stack engineers	41	$1,640	$1,230	$2,870	$70
Mobile engineers	24	$960	$480	$1,440	$60
Backend engineers	18	$720	$1,260	$1,980	$110
DevOps / platform	9	$540	$630	$1,170	$130
QA / test automation	8	$240	$160	$400	$50
Data / AI engineers	8	$320	$2,080	$2,400	$300
Engineering managers	7	$140	$35	$175	$25
Architects / tech leads	5	$250	$275	$525	$105
Attributed total	212	$8,490	$8,910	$17,400	$82

The average is the least useful number in that table. Data and AI engineers run about $300 a month each; engineering managers run $25. That is a 12x spread, and it is entirely driven by token consumption rather than seat cost — data and AI engineers spend $2,080 on tokens against $320 on seats.

Product engineers are the opposite shape: the largest group, the largest total, and a fairly even split between seats and tokens.

---

Spend by tool

Tool	Spend	Share of total
GitHub Copilot	$5,400	28.6%
Cursor	$4,100	21.7%
Claude Code	$2,700	14.3%
OpenAI API	$2,000	10.6%
ChatGPT Team	$1,500	7.9%
Anthropic API	$1,250	6.6%
Perplexity Enterprise	$850	4.5%
Self-hosted model infra	$700	3.7%
Other AI utilities	$400	2.1%
Total	$18,900	100%

Seat-based tools take the top three positions and 64.6 percent of spend between them. Raw API access accounts for 17.2 percent, and it is concentrated in a small number of heavy users rather than spread across the org.

---

What this changes about budgeting

Do not budget per seat uniformly. A flat per-engineer allocation overspends on managers and QA while capping the roles that actually consume tokens. Allocate seats broadly and token budgets by role.

Watch the API line, not the seat line. Seat cost is predictable and bounded. Token spend is the line that moves with usage, and in this model it is 47 percent of the total while sitting with under 15 percent of headcount.

Shared infrastructure needs an owner. The $1,500 of unattributed org-level spend is the easiest thing to lose track of, because no team sees it on their own budget.

---

Reading this against your own numbers

Role mix drives everything here. An organisation weighted toward data and AI engineering will land well above $89 per head; one weighted toward QA and management will land below it. Compare the per-role figures to your own mix rather than the headline average.

---

Working out what AI tooling should cost across your engineering org, or building the systems that consume those tokens? Ortem Technologies' AI and ML solutions practice works on both sides of that. See our LLM cost optimization guide for reducing the token line, or talk to our team →.