AI invoices show what organizations are charged, but those amounts can differ from the estimated cost of AI use. Larridin Spark data shows how large the gap can get and why finance teams need both views.
The Short Answer
Billed AI spend and usage-based cost estimates can tell different stories. In one Larridin customer, recorded billed spend totaled $324,400 over 12 weeks, compared with $164,000 in estimated usage costs. One tool showed $0 in the billed ledger despite $162,255 in usage-based costs because its flat-fee licenses weren't captured there. Finance teams need both views to understand AI spending and usage.
Key Wins
What’s a Spark Insight?
A Spark Insight is a data-backed finding Larridin surfaces automatically from how your teams actually use AI, flagging the patterns worth a leadership conversation.
Who This Matters To
CFO · CIO · Procurement
| If you’re asking… | The answer | Go deeper |
|---|---|---|
| What is the difference between billed and observed AI spend? | Billed spend is the cost recorded in the AI spending ledger. Observed spend estimates the cost of measured AI activity, including usage associated with flat-fee tools. The two figures aren't interchangeable. | Definitions |
| Can AI tools have usage but no visible cost? | Yes. In one org, a tool showed $0 on the billed ledger despite $162,255 in observed usage over 12 weeks. | The data |
| Do billed and used AI spend move together? | Not always. In one org over four weeks, billed spend rose 22% while observed usage fell 65%. | The data |
| How do I find AI spend I can’t see? | Compare invoices and ledgers with usage data across every AI tool, including ones billed through other contracts. | How to measure |
Finance teams see AI spend through invoices, but billed charges don't always align with estimated usage costs. Flat-fee licenses, bundled contracts, and unused seats can all contribute to the gap.
| What you’re struggling with | What you actually need | How Larridin helps | Proof from Spark data |
|---|---|---|---|
| “Our AI costs don’t match what teams tell us they use.” | Billed and observed spend side by side | Compares actual (billed) spend with observed usage by tool | One org: $324,400 billed vs. $164,000 observed over 12 weeks |
| “Some tools look free, but people use them constantly.” | Visibility into usage the ledger doesn’t show | Measures observed usage for every tool, regardless of billing | One tool: $0 billed vs. $162,255 observed |
| “Spend went up, but usage seems down.” | Trends for both numbers | Tracks billed and observed spend over time |
The question you’re asked: “Is our AI spend under control?”
What to do: Review billed and observed spend together each month. Investigate any tool where the two diverge sharply.
For more on financial oversight, see AI Monitoring for CFOs: Turn Untracked Spend Into Measurable ROI.
The question you’re asked: “Which AI tools are we actually using?”
What to do: Use observed usage to decide renewals, not invoice size. A $0 line on the ledger can hide heavy use and risk.
For more on shared IT and finance accountability, see Why CIO and CFO Cannot Agree on AI Accountability.
The question you’re asked: “Are we buying the right AI licenses?”
What to do: Match contract terms to observed usage before renewal. Right-size seats where billed spend has no matching use.
For more on evaluating AI contracts before renewal, see How to Decide Which AI Tools Survive Your Budget Review.
Estimated usage costs are not additional invoices and should not be added to billed spend. A $0 ledger entry means a billing source is missing from that ledger, not that the tool is free.
Because billing structures, bundled contracts, and unused licenses separate what you pay from what people use.
Some AI tools are billed through flat-fee agreements or other vendors’ contracts, so they show little or nothing on the AI ledger even when usage is heavy. Others are billed for capacity no one uses. Either way, the invoice describes the contract, not the behavior.
Spark callout: In one org, a tool showed $0 on the billed ledger over 12 weeks while observed usage for the same tool was valued at $162,255.
Myth: Your AI spending ledger captures the full cost picture.
Reality: Flat-fee licenses and bundled contracts can create large gaps between billed charges and estimated usage costs.
Decisions based on either number alone can go wrong.
In one org over four weeks, billed spend rose 22% to $82,000, while estimated usage costs fell 65% to $112,300. Looking only at billed spend, finance would see an increase. Looking only at observed spend, IT would see a decline. Neither view alone explains what happened.
With two numbers: billed and observed by tool over time.
Billed spend shows charges recorded in the ledger. Observed spend estimates costs from AI activity. Comparing them helps finance teams identify discrepancies worth investigating before making budget or renewal decisions.
Observed spend estimates the value of AI usage from actual activity, so it can be compared with what was billed.
Usually because it’s billed elsewhere, such as through a flat-fee agreement, a bundle, or another vendor’s contract, so it doesn’t appear on the AI ledger.
It may reflect unused capacity, contract terms, missing telemetry, or a mismatch between estimation and billing. Reconcile those factors before treating the gap as waste or using it in a renewal decision.
See billed and observed AI spend side by side for every tool, and bring it to your next finance review.