Resource tiers & sizing
Every PointOfContactAI deployment runs entirely in your own Azure subscription. The purchase wizard pre-sets each resource to the tier recommended for your plan, and you can adjust the cost levers under Tiers & cost before you deploy.
You own the infrastructure. Every resource runs in your Azure subscription and resource group. Your conversation data never leaves your tenant, PointOfContactAI holds zero standing access to it, and the only thing that ever leaves your stack is a content-free seat count for Marketplace billing. You can audit, export, and delete all of it.
Recommended tiers by plan
| Resource (in your subscription) | Start — $49 / 3 seats | Growth — $99 / 7 seats | Professional — $199 / 15 seats |
|---|---|---|---|
| Azure SignalR | Free F1 ($0, 20 connections) | Free F1 ($0, 20 connections) | Standard S1 (~$49.73/mo, 1,000 connections + SLA) |
| Azure OpenAI ceiling (gpt-5-mini) | 10K TPM | 30K TPM | 60K TPM |
| Function App (Flex Consumption) | Scale-to-zero | Scale-to-zero | + 1 always-ready instance |
| Cosmos DB | Serverless | Serverless | Serverless (provisioned optional) |
| Storage / Key Vault / App Insights | Minimal | Minimal | Minimal |
| Estimated Azure infrastructure | ~$3–15/mo | ~$10–30/mo | ~$60–120/mo |
Plus pay-as-you-go AI tokens (gpt-5-mini ≈ $0.25 per 1M input, $2.00 per 1M output) and the per-seat plan license ($49 / $99 / $199) — all billed separately. See cost structure below.
Cost structure
Your monthly cost has two parts that are billed independently:
- The per-seat plan license — Start $49 / Growth $99 / Professional $199 (plus metered overage if you exceed your seat quota past the 7-day grace), billed through Azure Marketplace on your Azure invoice.
- The Azure infrastructure in your subscription — billed to you directly by Microsoft at standard rates (the table above), plus pay-as-you-go AI tokens.
The plan license does not include the Azure infrastructure cost. Running your own isolated stack is what keeps your data private — and that stack carries its own Azure bill.
The cost levers (adjustable in the wizard)
Azure SignalR — real-time messaging
Each active agent dashboard and each open visitor widget holds one connection.
- Free F1 ($0/mo) — 20 concurrent connections, no SLA. Fits small teams (Start, and most Growth deployments).
- Standard S1 (~$49.73/mo) — 1,000 connections plus an availability SLA. Recommended for Professional or any deployment that may exceed 20 concurrent connections.
- Minimize for Starter: Free F1 — the single biggest standing-cost item, kept at $0.
Azure OpenAI throughput (gpt-5-mini)
A tokens-per-minute ceiling, not a reservation — you pay only for the tokens you consume, so a higher ceiling has no idle cost. Raise it only to avoid throttling during concurrent spikes.
- 10K TPM (Start) · 30K TPM (Growth) · 60K TPM (Professional) · up to 120K available.
Function App warm instances (Flex Consumption)
- Scale-to-zero — no standing compute cost; a few seconds of cold-start latency after idle. (Start, Growth.)
- 1+ always-ready — stays warm for lower latency, at a small standing cost. (Professional.)
Cosmos DB throughput
- Serverless — pay per request, no idle cost; cheapest at low/medium volume. The default on every plan.
- Provisioned autoscale — predictable cost for sustained high volume. Choose it at initial deploy — a Serverless account cannot switch to provisioned in place later.
Changing tiers later
Switching plans in the Marketplace updates your seat quota immediately, but it does not re-provision the resource tiers you picked at deploy (SignalR SKU, OpenAI ceiling, always-ready instances). Those are standard Azure resources in your own subscription — scale them yourself whenever your traffic grows. (Cosmos Serverless ↔ provisioned is the one exception that must be set at initial deploy.)
What stays minimal on every plan
Storage (Standard_LRS), Key Vault (Standard), and Application Insights are already near-zero cost and identical on every plan — there’s nothing to tune.
Estimating your bill
- Use the in-dashboard cost projection for AI token spend.
- SignalR Free, Cosmos Serverless, and scale-to-zero Functions have no idle cost — you only pay when traffic flows.
- The only standing-cost levers are SignalR Standard (~$49.73/mo), provisioned Cosmos, and always-ready Functions.
Compare plans → · How seats & overage work → · Wizard field guide →