# AI Spend Showback and Chargeback Model

> A free showback and chargeback model for Claude and AI spend: show every cost centre its full cost, hold seats centrally, and recharge only consumption above a set allowance.

Source: https://kowalah.com/resources/templates/ai-spend-showback-chargeback

## Overview

- **Who it's for:** CIOs, CFOs, finance business partners and heads of AI or operations who own a Claude Enterprise seat count and need to allocate AI spend across cost centres
- **When to use it:** Once seats are live and consumption is growing, and before the next budget round or seat purchase. Start with showback; move to chargeback after a quarter of numbers budget holders trust.
- **What you get:** One model that answers 'what did each team spend, and what did they get for it' every month, and recharges only the spend a team controls

## Get it

- **Copy the Google Sheet:** https://docs.google.com/spreadsheets/d/1MXu63-8J61W3rg28rr7S0JNmFb9wQmFLAILiJWP6o50/copy

## How to Use This Model

1. **Make your own copy.** Open the Google Sheet and it creates a copy in your Drive. Use File, then Download, if you work in Excel.
2. **Enter your settings.** Seat prices from your order form, the consumption allowance per seat and any overhead uplift. The example prices are illustrative.
3. **Add your cost centres.** One row per cost centre: headcount, seats by tier, active users and consumption, from Claude usage analytics, the Console or your invoices
4. **Run showback for a quarter.** Share the Showback tab with budget holders each month. Switch on recharges only once they trust the numbers.

***NOTE**: The model is a Google Sheet. Use the button above to make your own copy, then
replace the example figures, including the seat prices, with your own.*

# AI spend showback and chargeback model

Every organisation running Claude Enterprise can say how many seats it has bought. Fewer
can say what each team spends on AI each month, or which teams deserve more. This model
gives finance and the programme lead one sheet that answers both, and recharges only the
spend a team controls.

It follows one rule:

1. **Always show back.** Every cost centre sees its full AI cost every month: seats plus
   consumption.
2. **Charge back consumption only.** Seats are a fixed cost, held centrally.
3. **Each seat comes with a consumption allowance**, also held centrally.
4. **Only consumption above that allowance is recharged** to the cost centre.

## What's in the sheet

| Tab | What it does |
|---|---|
| Start here | The rule, how to use the model, and where each number comes from |
| Settings | Seat price per tier, the consumption allowance per seat, and an optional overhead uplift on recharges |
| Inputs | One row per cost centre: headcount, standard and premium seats, active users in the last 30 days, and consumption this month |
| Showback | Seat cost, consumption, total AI cost, cost per head, cost per active user, seats in active use, and each cost centre's share of the total |
| Chargeback | The allowance per cost centre, consumption within and above it, the uplift, the recharge, and what stays central. A check confirms recharges and central costs add back to total spend |
| Usage segments | How many seat holders are power, middle and idle users, charted against a Kowalah reference split |

Yellow cells are inputs. Every other cell is a formula, so the sheet recalculates as
soon as you enter your own figures.

## A worked example

The sheet ships with six example cost centres and 600 seats. With a £20 allowance per
seat:

| Cost centre | Total AI cost (showback) | Consumption | Allowance | Recharge |
|---|---|---|---|---|
| Finance | £5,750 | £1,400 | £1,900 | £0 |
| HR | £2,100 | £300 | £900 | £0 |
| Legal | £3,050 | £900 | £800 | £100 |
| Sales | £12,300 | £4,800 | £3,200 | £1,600 |
| Operations | £5,700 | £900 | £2,400 | £0 |
| Technology | £24,000 | £9,600 | £2,800 | £6,800 |
| **Total** | **£52,900** | **£17,900** | **£12,000** | **£8,500** |

Every cost centre sees its full cost. Three of the six receive a recharge, and those
recharges come to 16% of total AI spend. The rest, seats and ordinary usage, stays in the
central budget, where the programme decides how to use it. The figures, including the seat
prices, are illustrative.

## Where the numbers come from

- **Seats by tier:** your Claude Enterprise admin console or your order form.
- **Active users and consumption:** Claude Enterprise usage analytics, the Claude Console
  for API and Claude Code spend, or your invoices.
- **Cost centres:** an HR or directory export that maps each user's email address to a
  cost centre.

## Reading the showback

Look at **cost per active user** before cost per head. Cost per head divides spend across
everyone in the team, including people without a seat. Cost per active user shows what
each person using AI costs.

A high cost per active user next to a low share of seats in active use means seats are
going unused. In the example, Operations has 120 seats and 45% of them in use. That is a
case for moving seats or running enablement in that team, not for cutting the AI budget.

## Usage segments

The usage tab splits seat holders into three groups:

| Segment | Definition (adjust to suit) |
|---|---|
| Power | Active on 15 or more of the last 20 working days |
| Middle | Active in the last 30 days, but not a power user |
| Idle | Holds a seat, with no activity in the last 30 days |

The chart compares your split with a 20 / 60 / 20 reference. **That reference is a
Kowalah estimate from our client work, not measured data.** Replace it with your own trend
once you have a quarter of numbers.

The middle group is where to look before you buy more seats. These people use Claude, but
not yet for their repeated work. A shared skill or a better starting point for a common
task moves more of them into regular use than new licences do.

## When to move from showback to chargeback

Run showback for at least one quarter first. Charge back only when:

- budget holders accept the showback figures as accurate,
- each user maps to one cost centre, and
- the spend is metered, so the team can manage it.

Recharging from month one turns every figure into a dispute about the data rather than a
decision about the spend.

## A note for UK groups

Recharges between cost centres inside one legal entity are internal accounting.
Recharges between legal entities in a group raise VAT and transfer-pricing questions.
Involve finance before you recharge across entities.

## Implementation notes

- **Set the allowance from data, not a guess.** Run one or two months with the allowance
  at zero, look at what your middle users consume, and set it at that level.
- **Decide the uplift up front.** If the central team's cost is recovered through
  recharges, set the uplift percentage before the first recharge, not after.
- **Keep one month per copy.** Duplicate the sheet each month, or add a month column to
  the Inputs tab, so the trend is there when the steering group asks for it. Our
  [AI steering group update template](/resources/templates/ai-steering-group-update-template)
  has a place for it.
- **Tie spend to results.** Showback tells you what AI costs. To show what it returns, pair
  it with the measures in your
  [AI business case](/resources/templates/ai-business-case-template).

## Frequently Asked Questions

### What is the difference between showback and chargeback?

Showback reports each cost centre's share of AI spend without moving any money: it makes the cost visible. Chargeback recharges that spend to the cost centre's budget. This model always shows back the full cost, and charges back only consumption above a central allowance.

### Why hold seats centrally instead of charging them back?

A seat is a fixed cost the organisation commits to for the contract term, and a team can't reduce it month to month. Charging it back pushes budget holders to give seats up, which cuts the use the programme is paying for. Consumption is different: it moves with what a team builds and runs, so the team can manage it.

### What is a consumption allowance, and how do we set it?

A monthly amount of metered usage per seat that the centre pays for, so ordinary use never triggers a recharge. Set it from your first months of data: a level that covers your middle users' normal usage. Anything above it is heavy API, Claude Code or agent spend that a team chose to run.

### What counts as consumption?

Metered spend that scales with use: usage billed on top of Claude Enterprise seats, API calls, Claude Code, and agents running on Anthropic's platform. Take the figures from Claude usage analytics, the Claude Console or your invoices, and map users to cost centres with a directory export.

### When should we move from showback to chargeback?

After at least one quarter of showback numbers that budget holders accept as accurate, and only for metered spend. Recharging from month one turns every figure into a dispute about the data rather than a decision about the spend.

### Why does the model show cost per active user?

Cost per head spreads spend across everyone, including people with no seat. Cost per active user shows what each person using AI costs. A high figure next to a low share of seats in active use points to idle seats, which is a reallocation question before it is a budget one.

### Where does the 20 / 60 / 20 usage split come from?

It is a Kowalah estimate from our client work, not measured data, and the sheet labels it that way. Use it as a reference point and replace it with your own trend. The middle group, people who use Claude but not for their repeated work, is where to look before buying more seats.

### Are there tax issues with recharging AI costs in the UK?

Recharges between cost centres in one legal entity are internal accounting. Recharges between legal entities in a group raise VAT and transfer-pricing questions, so involve finance before you recharge across entities.

---

Free and ungated, from Kowalah, a UK Anthropic implementation specialist. More templates and skills: https://kowalah.com/resources/templates. Talk to us: https://kowalah.com/contact