AI margin

You price flat.
You pay per token.

Per seat, per agent, per clinic — while your costs run per token, per minute, per character. Someone has to reconcile the two. Today, that someone is a spreadsheet.

Bring your stack and how you price. We'll show you what your margin picture could look like.

Part of the UnitSense AI margin platform.

The mismatch

Flat price out, variable cost in. Your heaviest customers can be your least profitable — and you find out at the end of the month.

What you see

Margin, at the unit you actually sell.

Not a bill. A per-customer answer to whether the price is working.

01Cost per unit you sell — per call, per minute, per session, per seat
02Margin per customer and per feature
03The customers who break your pricing model
What you do about it

Repricing, rehearsed before you ship it.

What to charge, what to cap, what happens at the next tier — modeled on your own usage, so the pricing conversation runs on numbers instead of nerves.

Why it works

Your whole stack, not one vendor.

Your OpenAI dashboard doesn't know ElevenLabs exists. Your margin lives across every vendor behind the product — so that's where we compute it.

01

Unify

LLMs, voice, speech-to-text, GPU, infrastructure — every cost behind your product lands in one ledger instead of six dashboards.

02

Attribute

Every request tagged to a customer, feature, team, or environment — so cost rolls up to the units your pricing is written in.

03

Reconcile

The ledger is checked line-by-line against your actual provider invoices, so the margin number is one finance can defend.

Integrations

Built for the stack you already run.

ModelsOpenAI·Anthropic·Google Gemini·AWS Bedrock·Azure OpenAI·Mistral
Voice & speechElevenLabs·Cartesia·Hume·Deepgram·AssemblyAI
Voice agents & transportVapi·Retell·LiveKit·Pipecat·Twilio
GPU & cloudAWS·GCP·Azure·CoreWeave·Modal·RunPod
Vector & dataPinecone·Qdrant·Weaviate·Chroma
Coding toolsClaude Code·Codex·Cursor·GitHub Copilot·Windsurf
RevenueStripe·Orb·Lago·Chargebee·QuickBooks

How we connect: LiteLLM · Bifrost · Python SDK · JavaScript SDK · our open-source agent · provider admin APIs · CSV import · invoice ingestion

…and the rest of your stack. Where a vendor exposes usage, we pull it. Where they don't, we ingest the invoice.

Next step

See your numbers.

We'll show you what your margin picture could look like — built on your stack and how you price. That's the whole agenda.

Book 20 minutes

UnitSense is built by AI-infrastructure and platform engineers out of CTDS, previously PLUMgrid.