How to Track, Allocate and Control AI Spend by Team

How do you track and control AI spend?

To track AI spend, use Cledara: it keeps every AI tool on one list with an owner, a cost centre and a budget, connects to AI providers such as OpenAI, Anthropic and Cursor, tracks usage daily against the budget you set, and warns finance and the owner before spend runs over. The method behind it: one list of every AI cost, an owner, a cost centre and a monthly budget for each line, and daily tracking for anything billed by usage.

AI spend management is the work of tracking, allocating and budgeting everything your company pays for AI, so finance can answer three questions every month: what are we spending on AI, who owns each cost, and where will it land this year?

AI now takes a large and growing part of the software bill. 27% of software spend on Cledara now goes to AI-native tools, up from about 3% two years ago, and AI-native spend from August to October 2026 was up 25% in the UK, 23% in Europe and 16% in the US on the previous three months (AI software spend winners and losers). Much of it is billed by usage, bought on personal cards or hidden inside price rises on existing tools, so a normal software budget process misses a lot of it.

This guide covers how to track AI spend in five steps, track it by team, allocate AI costs and budget for usage-based AI bills. To compare tools, see our roundup of AI spend management tools; for what other companies spend, see our AI spend benchmarks for 2026.

Key takeaways

• There are four types of AI cost: seat-based tools, usage-based APIs, AI features inside existing tools and AI agents.
• Put every AI cost on one list with an owner, cost centre and monthly budget.
• To track AI spend by team, code each AI tool to its department and cost centre once, so every payment lands against the right team.
• Track usage-based AI daily: Cledara does this for OpenAI, Anthropic and Cursor and alerts the owner and finance before spend runs over.
• Allocate seat plans by seat and APIs by usage share. Start with showback, then move to chargeback.
• Budget usage-based AI from three to six months of actuals, a 20 to 30% buffer and three scenarios, and review every month.

The four types of AI spend, and how each is billed

Every AI cost falls into one of four types, and each shows up somewhere different: seat-based tools on card statements, usage-based APIs in developer billing consoles, AI features inside existing tools as add-ons and renewal price rises, and AI agents in a mix of all three. Most finance teams only track the first.

A seat plan with a usage allowance is a hybrid: track it as a seat-based tool, with a usage line for any overage charges.

Note: AI features inside existing tools are the easiest to miss, because they create no new vendor. Our guide to finding the AI tools already inside your SaaS stack shows where to look.

How to track AI spend in five steps

To track AI spend, build one list of every AI cost, give each line an owner and a cost centre, move payments onto company cards and accounts, connect usage data for anything billed by usage, then review the list every month.

1. Build one list of every AI cost.
2. Give every line an owner and a cost centre.
3. Centralise AI payments.
4. Connect usage data for usage-based AI costs.
5. Review every month and consolidate.

1. Build one list of every AI cost

Start with 12 months of payment data and flag every AI vendor. Then check where card statements don't reach:

• Expense claims: AI tools bought on personal cards and claimed back, one seat at a time.
• Developer billing consoles: ask engineering leads for every API account, its billing email, the project it serves and its expected monthly spend.
• Invoices and renewals: compare each with the previous price and note AI add-ons and increases tied to AI features.
• Browser usage: Cledara's Engage browser extension shows which of 8,000+ tools your team actually uses, including AI tools nobody approved.

For each line, record the vendor, type, billing model, monthly cost, seats or usage, renewal date and the card or account that pays it. Repeat the sweep every quarter. Tools bought outside the approval process are covered in our shadow AI guide for finance teams.

2. Give every line an owner and a cost centre

An AI cost without an owner is a cost nobody will cut. Name one person per line who approves seats, answers usage questions and decides at renewal. Then map each tool to its expense account, department and cost centre once, so every payment is coded the same way at month-end. Shared tools need a rule, covered in the allocation section below.

3. Centralise AI payments

Scattered payment methods are the main reason AI costs go untracked: when ten engineers each expense a coding assistant, nobody sees the total. Move every AI subscription onto company payment methods and every API account onto a company email address, and give finance access to each billing console.

The cleanest set-up is one virtual card per AI subscription, so every charge belongs to one tool, one owner and one budget line. When someone leaves, the tool keeps working, because the card belongs to the subscription rather than the person.

4. Connect usage data for usage-based AI costs

Usage-based AI costs need daily data, not a monthly invoice. A misconfigured integration or a large batch job on an expensive model can use up a month's budget in days.

• Create separate API keys, projects or workspaces for each team or product feature, so usage can be split later.
• Use each provider's own limits as a safety net: monthly spend limits per organisation and project at OpenAI, and per workspace in the Claude Console.
• Check whether automatic credit top-ups are switched on, and who receives the receipts.
• Track usage every day against the monthly budget, and warn the owner when the run-rate points over it.

Tip: If an API account normally costs £50 a day and suddenly costs £300, you want to know that week, not when the invoice arrives. Daily tracking against budget is what makes that possible.

5. Review every month and consolidate

Review the AI list every month with the owners: new tools, seat changes, usage against budget and renewals in the next 90 days.

How to track AI tool spend by team

To track AI tool spend by team, code each AI tool to its department and cost centre once in Cledara, so every payment lands against the right team. Each tool gets an owner; Cledara tracks OpenAI, Anthropic and Cursor usage daily against budget and alerts the owner and finance before spend runs over. Your accounting then shows AI spend by team without month-end recoding.

AI tools are no longer just for engineering. Of the 872 tools Cledara's data classifies as AI-native, 258 are marketing and growth tools, 83 sales tools, 57 engineering tools and 44 customer support tools.

1. Code each AI tool once. Map every AI tool to its expense account, department and cost centre in Cledara; every payment for it is then coded the same way and syncs with its invoice to your accounting system.
2. Name an owner in the team that uses it. The owner approves seats and decides at renewal; Cledara alerts them and finance before each renewal.
3. Split shared API accounts before the month starts. Give each team its own OpenAI project or Claude Console workspace, so usage can be attributed.
4. Set a budget per team and tool. Cledara tracks usage every day for OpenAI, Anthropic and Cursor against the budget you set for each.
5. Act on alerts. When a team's run-rate points over budget, Cledara warns the owner and finance, with Slack notifications they can act on.
6. Report monthly by team. Send each budget holder their AI spend by tool, against budget, with the change from last month.

For example, at a 200-person company:

Tools everyone uses, such as a company-wide AI assistant, need a rule: hold them centrally or split them by seat or headcount, as the next section explains.

AI cost allocation: how to split AI spend across teams

AI cost allocation assigns each AI cost to the team, product or project that uses it, so every budget holder sees and answers for their own share. Match the method to the billing model: allocate seat plans by seat, usage-based APIs by each team's share of usage, and shared company-wide tools centrally or by headcount.

For example, if your company pays for 40 seats on an AI assistant and marketing holds 12 of them, marketing carries 30% of the cost. If the data team used 60% of last month's API tokens and the product team used 40%, the API bill splits 60:40.

Showback or chargeback?

Showback reports each team's AI costs to its budget holder without moving them. Chargeback moves the cost into the team's budget in the ledger, through cost centre coding or a monthly journal. Start with showback, which builds trust in the numbers, and move to chargeback once owners agree the split is fair.

For most growing companies, department budgets with central rules work best: each department buys within its own AI budget, while finance or IT sets the total and keeps the approved vendor list.

How to budget for AI tools and usage-based AI costs

Once you track AI spend daily, budgeting gets easier. Treat seat-based tools like any other SaaS: seats multiplied by price, plus room for new seats. Budget usage-based AI costs from three to six months of actual spend, a growth assumption and a buffer. Then set budgets by tool, team and use case, and review them every month.

Start from three to six months of actuals

Pull at least three months of history for each AI line and look at the trend, not just the latest bill. For APIs, track tokens or calls as well as cost: model prices change often, so cost alone can hide how fast usage is growing.

Add a buffer that matches the billing model

Usage-based AI costs will not land on an exact number, so plan a range. These rules of thumb are a sensible start:

• Seat-based AI tools: 5 to 10% above the current run-rate, for unplanned seats.
• Hybrid plans: budget at the tier that covers expected usage, plus 15% for overages.
• Usage-based APIs and agents: 20 to 30% above projected spend.

Pair every buffer with an alert that tells you early when you will need more.

Plan three scenarios

• Conservative: current tools and seats, with usage growing at the trailing three-month rate. This is your floor.
• Base case: 10 to 20% more seats and one or two new tools, with usage growing at 1.5 times the trailing rate. This is your planning number.
• Aggressive: a company-wide rollout or a new AI feature in your product, with adoption up 30 to 50%. Your buffer should cover most of this gap.

How to reduce AI spend without slowing teams down

Most AI savings come from four places: individual plans that should be one team plan, paid seats nobody uses, duplicate tools and renewals accepted without a conversation. None of them takes AI away from the people who use it.

• Consolidate individual plans. 15 developers each paying £20 a month on individual plans cost £3,600 a year, with no admin controls. A team plan adds admin and security controls, and the seat count gives you something to negotiate with.
• Take back paid seats nobody uses. Compare paid seats with active users every month. Cledara Engage shows who last used each tool, and Cledara Agent removes the seats you approve.
• Cut duplicate tools. Three AI writing assistants or two meeting note-takers is common after a year of team-by-team buying. Of companies on Cledara that pay for Anthropic, 88% also pay for OpenAI, so check whether both are needed for the same job.
• Negotiate renewals early. Vendors raise renewal prices 8 to 12% a year on average. Companies that start renewal negotiations 90 days out average 49% savings, against 19% when they start 30 to 90 days out (SaaS renewal benchmarks 2026).

Cledara's Spend Optimisation module flags duplicate tools, unused seats and price increases, benchmarks what you pay against similar companies and drafts renewal negotiations with its Negotiation and Renewal Copilot.

AI spend checklist

Use this checklist to track AI spend from the first sweep onwards, and rerun the monthly items at each review. If you can tick every line, you can say what you spend on AI, who owns it, which team it belongs to and where it will land.

• ✓ Every AI tool, API account and agent is on one list with its billing model, cost and renewal date.
• ✓ Each line has a named owner, an expense account, a department and a cost centre.
• ✓ AI subscriptions are paid from company cards, ideally one card per subscription.
• ✓ API and agent accounts use company emails, are split by team and have a monthly budget.
• ✓ Usage-based AI is tracked daily, with alerts to the owner and finance before spend runs over.
• ✓ AI add-ons and AI-driven price rises are recorded at each renewal.
• ✓ Each team sees its AI costs monthly, with an agreed date to move from showback to chargeback.
• ✓ The AI budget has a buffer by billing model and three scenarios.
• ✓ Paid seats are compared with active users monthly, and renewals reviewed 90 days out.

How Cledara helps you track AI spend and control it

Cledara is software subscription management for finance teams. It keeps every SaaS and AI tool you pay for on one list, with its owner, users, cost, renewal date and card, so AI tools sit with the rest of your software, not in a separate spreadsheet.

• Every AI tool on one list: built from every payment made with Cledara and the tools you add.
• Daily AI usage tracking: Cledara connects to AI providers such as OpenAI, Anthropic and Cursor, tracks usage every day against the budget you set, and warns finance and the subscription owner before spend runs over.
• Spend by team: map each tool to its expense account, department and cost centre once. Payments sync to Xero or QuickBooks on every plan, or NetSuite on Premium and Pro, which also add advanced reporting and forecasting.
• AI tools nobody approved: Cledara Engage shows which tools people actually use, including AI tools bought outside the approval process.
• A virtual card per subscription: every charge has an owner; freeze the card and the vendor can't charge it again.

Cledara Agent, an add-on for any plan, signs in to your SaaS and AI tools to collect the invoices nobody emails you and take back paid seats nobody uses, once you approve. See how Cledara Agent works.

Customers save 23% on software spend on average. See how Cledara manages every software subscription, or compare Cledara plans.