How do I chargeback AI spend to the right GL account?

Tag every AI call with the X-Cost-Department header, then map each department to a GL account code on the GL Mappings tab. Cognocient produces a monthly department breakdown you can export as a generic CSV or a FOCUS-aligned CSV for your ERP.

Can I set the department header without touching every call site in my codebase?

You still need to send X-Cost-Department per call (or set it once via default_headers on your AI client) — proxy keys themselves don't carry a default department today.

What export formats are available for chargeback reports?

A generic CSV (date, department, project_code, gl_account, gl_account_name, amount_usd, description) and a FOCUS-aligned CSV. There is no PDF or JSON chargeback export today.

Can I chargeback AI costs to enterprise clients, not just internal departments?

Yes. Use the client's account ID as the X-Cost-Department value, and each client account appears as its own card in the Department Breakdown tab.

What happens to calls that arrive without a department header?

They're excluded from the department breakdown and export rather than allocated to an explicit 'Unattributed' line — tag consistently to avoid gaps in your chargeback data.

How do I chargeback AI spend to the right department or GL account?

AI spend attributed to the department that generated it, mapped to your GL chart of accounts, and exported as CSV — ready for your ERP or your board deck. No spreadsheet. No manual allocation.

Cognocient attributes every AI call to the department that generated it, using the X-Cost-Department header you set. Map departments to your GL chart of accounts once, and every month's chargeback export is ready to hand to finance without spreadsheet work.

What AI chargeback solves

Most organisations with multiple teams using AI face the same problem: a single AI provider invoice arrives with one number. Finance has to allocate it manually, usually by asking engineering to estimate, usually wrong.

Cognocient solves this at the proxy layer. Every API call is tagged with a department at the moment it is made — not allocated after the fact. By month-end you have a department-level breakdown you can export directly.

Step 1 — Tag calls with a department

The only required chargeback header is X-Cost-Department. Add X-Cost-Feature too if you want project-level detail in the export.

X-Cost-Department: customer-success (required) The business department this call belongs to. Use consistent slugs — these are exactly what you'll map to GL codes in Step 2.

from openai import OpenAI
 
client = OpenAI(
    api_key="sk-cog-YOUR-PROXY-KEY",
    base_url="https://api.cognocient.com/v1",
)
 
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": user_message}],
    extra_headers={
        "X-Cost-Feature":    "ticket-resolver",
        "X-Cost-Department": "customer-success",
    },
)

Set it once at the client level

If every call from a given service belongs to the same department, set X-Cost-Department in default_headers when you create the client, instead of passing it on every call. There's no key-level default today — the header still has to come from your application (directly or via default headers), not from proxy key configuration.

Step 2 — Map departments to GL accounts

Go to Chargeback → GL Mappings. Click Add Mapping, enter the department name exactly as it appears in your X-Cost-Department header, and give it a GL account code and name (e.g. customer-success → 6420 / "Software & SaaS"). Every call tagged with that department now resolves to this GL code in your exports.

Step 3 — Review the Department Breakdown

Go to Chargeback → Department Breakdown and pick a month. Each department appears as its own card showing total spend, call count, and its top features. There's no per-department efficiency score or approval-workflow step on this page — it's a straightforward spend breakdown.

Step 4 — Export for finance

On the Export for Finance tab, download either:

Format	Contains
Generic CSV	`date, department, project_code, gl_account, gl_account_name, amount_usd, description`
FOCUS CSV	FinOps Open Cost and Usage Specification columns (best-effort alignment — see FOCUS Export for current conformance status)

There is no PDF or JSON chargeback export today. If your ERP needs a different shape, the generic CSV's columns map cleanly onto most GL import tools.

Step 5 — Chargeback to enterprise clients

For B2B SaaS products that want to pass AI costs through to enterprise clients, use the client's account ID as the X-Cost-Department value. Each client account then appears as its own card in the Department Breakdown tab and its own row in the export.

def handle_request(request, account_id: str):
    response = client.chat.completions.create(
        model="gpt-4o",
        messages=request.messages,
        extra_headers={
            "X-Cost-Feature":    "ai-assistant",
            "X-Cost-Department": f"client-{account_id}",   # client as department
            "X-Cost-User":       request.user_id,
            "X-Cost-Session":    request.session_id,
        },
    )
    return response

Scheduled delivery to finance

From Reports → Scheduled Delivery, enable automatic email delivery of an AI ROI Summary or Full Detail report to your finance team on a monthly, weekly, or end-of-cycle cadence. This is a general spend/ROI report rather than the GL-coded chargeback CSV — for the department-and-GL-coded export itself, download it from the Chargeback page each period.