Track software subscription growth

# Software spend growth

## Purpose

Put a figure on software and subscription spend, spot tools whose cost is rising and those that appeared recently.

State the exact period (the last 12 months by default; 366 days at most per call).

## Two ways to identify software spend

Neither is complete on its own; use them together and say which one was used.

**A. By payment type** — spend made with a subscription card:

```json
{
  "tool": "spendesk_analyze_spend",
  "params": {
    "fromDate": "YYYY-MM-DD", "toDate": "YYYY-MM-DD",
    "groupBy": ["supplier", "month"], "limit": 50,
    "filters": { "payableTypes": ["subscription"] }
  }
}
```

Misses: software paid by invoice.

**B. By expense category** — a category cannot be filtered by name. Get its ID first:

```json
{ "tool": "get_expense_categories", "params": {} }
```

Pick the software category or categories (for example "Software & Subscriptions") and ask the user to confirm when the choice is not obvious. Then:

```json
{
  "tool": "spendesk_analyze_spend",
  "params": {
    "fromDate": "YYYY-MM-DD", "toDate": "YYYY-MM-DD",
    "groupBy": ["analyticalCode", "supplier", "month"], "limit": 50,
    "filters": { "analyticalCodes": ["<category id>"] }
  }
}
```

With this filter, put `analyticalCode` **first** in `groupBy` and read suppliers under the category's row: `summary.totalAmount` counts the matching expenses in full, including their lines on other categories. Misses: miscategorised spend.

With `limit: 50`, check that no significant tool fell into the "Others" row.

## Step 2 — Annual cost per tool

For each supplier, from its monthly series:

- regular payments every month: monthly average × 12;
- a single payment in the year: it is an annual renewal, take that amount;
- irregular: the 12-month total.

Say which rule was applied to each tool. Rank by annual cost.

## Step 3 — What is rising

For each tool, compare the monthly average of the last 6 months with that of the 6 months before.

- 🔴 more than 30%: rise to look into
- 🟡 10 to 30%: to watch
- 🟢 stable or falling

A rise can come from a price change or from a change in the number of licences: Spendesk does not say which. An annual renewal falling in one of the two halves distorts the comparison: flag it rather than announcing a rise.

## Step 4 — Tools that appeared recently

Suppliers whose first month of spend falls in the last 3 months. Present them as "new subscriptions": the data does not say whether they were approved or not.

## Report

```
Software spend — from <date> to <date> (method: payment type / category)

Annualised total: X EUR — N active tools
| Tool | Annual cost | Rule applied | 6-month change |
Rises to look into: tool, before, after, change
New subscriptions (last 3 months): tool, first month, monthly cost
```

Amounts in functional currency, including VAT, on the document date. Card purchases are included in these figures, once the transaction has appeared on the statement.

## Observed result (demo account, 5 October 2026)

No subscription-type spend over the last 12 months: the account's spend was invoices and card purchases. The "Software & Subscriptions" category exists but carried no spend over the period. The method could therefore not be tried on a real data set.

## Example prompts
- "What is our total software spend and how has it changed?"
- "Are there subscriptions whose cost has gone up?"
- "List the new subscriptions that appeared in the last 3 months"