Skip to content
nythrex.
Log inGet started

SaaS teams · AI spend

The AI bill nobody owns: seats, credits, API keys and the corporate card

Ask a founder of a 60-person SaaS company what it spends on AI for engineering and you will usually get a pause. There is the coding-assistant plan. There are a few personal subscriptions on expense reports. There is an API key someone created for an agent experiment, and a cloud invoice with a new AI line. Each one has an owner. The total does not.

Create account
SaaS teams · AI spend. The AI bill nobody owns.
AI arrives through four different doors, and each lands in a different report.

Nythrex team · 10/29/2026

In brief

6-minute read · For founders, CTOs and finance · 29 October 2026

  • Expensed software is exploding. Zylo found expense-based SaaS spend up 267% in a year, with ChatGPT now the most expensed application.
  • Surprises are the norm. 78% of IT leaders reported unexpected charges tied to consumption-based or AI pricing in the past 12 months.
  • Control sits outside IT. Business units control 81% of SaaS spend; IT directly manages 15%.
  • The fix is an owner, a project for every AI cost and one view that puts seats, usage and cloud together.

Four doors AI comes through

  • Seats

    Coding assistant plans per developer. Predictable, until the plan switches to usage-based credits, as GitHub Copilot did in June.

  • Usage and credits

    Agent runs and chat billed by tokens. The cost depends on the model and the task, not on head count.

  • API keys

    Keys created for experiments, scripts and internal tools. Often on one person’s account and one shared invoice.

  • Expense reports

    Personal subscriptions reimbursed monthly. Invisible to engineering, visible to finance only as a vendor name.

By the numbers

+267% year-over-year growth in expense-based SaaS spend. ChatGPT is now the most expensed application.
Source: Zylo, 2026 SaaS Management Index.

Zylo’s index, built on more than $75 billion of spend under management, shows where the growth is. AI-native application spend rose 108% overall and 393% in companies with more than 10,000 employees. Smaller companies feel the same pattern with fewer people to watch it.

Horizontal bar chart of year-over-year growth: AI-native app spend in companies with more than 10,000 employees up 393%, expense-based SaaS spend up 267%, AI-native app spend across all companies up 108%.
Year-over-year growth in software spend. Chart: Nythrex, from Zylo.

  • 78%

    of IT leaders saw unexpected charges from consumption-based or AI pricing in the past year.

  • 61%

    were forced to cut projects because of unplanned SaaS cost increases.

  • 36%

    of SaaS licences are left unused on average.

Why nobody owns it

Who sees which part of the AI bill
RoleWhat they seeWhat they miss
FinanceVendor invoices and expense linesWhich product or project the money served
Engineering leadAdmin consoles of the tools they managePersonal subscriptions and stray API keys
DevelopersTheir own usage, if anythingAny total at all
FounderThe card statementWhether any of it shipped

Every AI cost has an owner. The total doesn’t.

How to take ownership in a month

  1. 1

    Inventory every door

    Pull seats from admin consoles, usage from provider accounts, keys from your secret store and subscriptions from three months of expense reports.
  2. 2

    Name one owner for the total

    Usually the engineering lead, with finance as the second reader. One person answers the question “what do we spend on AI?”
  3. 3

    Give every cost a project

    Products, client projects or internal platforms. Anything without a project stays visible as unassigned, not spread across the rest.
  4. 4

    Cap before you cut

    Usage-based plans offer budgets. Set them per team or project before deciding which tools to drop.
  5. 5

    Review it next to delivery

    A monthly total says little. AI cost next to what each project shipped says whether it is worth it.

Where Nythrex fits

Nythrex brings AI usage from Claude Code, Codex and GitHub Copilot together with CI and cloud, by project. AI spend that has no project stays on its own line. Usage on a subscription is labelled as an estimate from the price list, and a total with a missing price is marked incomplete rather than shown as zero.

How AI cost tracking works

Sources

Figures are from Zylo’s 2026 SaaS Management Index, based on a survey of 218 IT leaders and more than 40 million licences. The four doors and the role table are Nythrex’s summary, not survey data.

Frequently asked questions

What is AI tool sprawl?

AI tool sprawl is when AI reaches a company through many separate channels: seat plans, usage credits, API keys and personal subscriptions on expense reports. Each has an owner, but no one sees the total or what it produced.

How fast is expensed AI spend growing?

Zylo’s 2026 SaaS Management Index found expense-based SaaS spend up 267% year over year, with ChatGPT the most expensed application, and AI-native app spend up 108%.

Who should own AI spend in a software company?

One person should own the total, usually the engineering lead with finance as a second reader. Each cost should belong to a product or project, with anything unassigned kept visible on its own line.

Read next

Stop guessing.
Start seeing.

Create account

No card and no subscription: paid plans have not launched yet.

Create accountNo card, no subscription