P402
P402.io·Meter
DocsPricingDashboard
> _ AI COST OPTIMIZATION READINESS

AI cost optimization
readiness.

P402 Optimize prepares AI spend for measured savings across models, cache, retries, and context. Recommendations are gated until baseline and outcome data prove the recommendation.

Status: readiness checks live. Recommendations gated.
See optimization readinessRead docs

Built on the P402 Optimize surface. Parent product page: /optimize.

The problem

AI cost claims without proof do not survive review.

Recommendations like switch this model, raise that cache, cut this retry, or trim that context only matter if a baseline and an outcome back them. The readiness page makes the data the recommendation needs explicit, and gates the recommendation until that data is in the ledger.

What this surface shows
Model readiness
Per workflow, do we have enough outcomes on two models to compare cost at equal quality.
Cache readiness
Per feature, is there a repeated-input signature with measurable hit potential against current cache.
Retry readiness
Per feature, are retries tagged and is their no-new-outcome share large enough to act on.
Context readiness
Per feature, is there context that does not change the outcome of the call across enough events.
Baseline coverage
Share of events with cost, tokens, outcome, and owner attached, by feature and workflow.
Outcome coverage
Share of events with outcome accepted, rejected, revised, escalated, or failed recorded.
Owner coverage
Share of events with customer, feature, workflow, and department attached.
Gate reasons
Why each readiness check has not yet released a recommendation, with the missing data named.
Privacy posture
Privacy mode per workflow. Readiness checks honor the same posture as Meter.
How it works
  1. 01
    Create a P402 key.
    Sign in, generate a key. The key is returned once. P402 stores only a hash.
  2. 02
    Send metered events for the AI calls you want to optimize later.
    Use the OpenAI-compatible endpoint or post a meter-only event from your backend. The readiness checks run on the events the ledger holds.
  3. 03
    Open the optimization readiness page.
    The page lists each readiness check with the data still missing before a recommendation can ship. No claim is made before the underlying data is in.
  4. 04
    Watch checks turn ready as data accumulates.
    A check stays gated until baseline and outcome data are measured. When a check goes ready, the recommendation it backs becomes available.
Privacy

Meter economics, not content.

Readiness checks read economic metadata. Prompt and response storage are off by default. Workflows that need stronger privacy can run under private gateway mode, where the inference stays in your environment and P402 records the economic event over a signed channel.

  • metadata_onlyOwner, cost, tokens, budget, policy, outcome, and evidence status. No prompts. No responses. Default.
  • fingerprint_onlyAdds a hash fingerprint of prompt and response for dedup and replay protection. Content stays out.
  • redacted_tracePrompt and response retained after a redaction pass for fields the tenant policy allows.
  • private_gatewayYour environment hosts the inference, P402 records the economic event over a signed channel.
  • full_traceOpt-in. Prompts and responses retained verbatim with the event. Requires explicit tenant policy.
Read the trust posture
Proof

Same metering layer, four shipped workflows.

Each vertical demo is a working surface on the same metering layer. Use them to see baseline coverage, outcome coverage, and owner coverage on concrete workflows readiness checks would run against.

Healthcare
Medicaid prior authorization with HIPAA-aligned demo mode and human review gate.
Legal
M&A contract due diligence across an 8-document synthetic data room.
Real estate
Tenant application screening with fraud scoring and HUD fair-housing audit trail.
Enterprise
Department, employee, model, and workflow attribution across one organization.
For developers

See what data a recommendation still needs.

Each readiness check states the data it needs. Add the missing owner tags, outcome labels, or context flags on the metered events you already emit. A recommendation ships only after the check it depends on goes ready.

Read docs
For enterprise

Gated claims survive finance review.

Recommendations are gated. No claim ships before measured baseline and outcome data back it. Finance and audit can read the same ledger the readiness checks run on.

Read the trust posture
FAQ
Are recommendations live today?+

No. Recommendations are gated. Readiness checks are live and run continuously over the recorded ledger. A recommendation ships only after baseline and outcome data measured on the ledger back the claim.

What does the readiness page show?+

The set of readiness checks across models, cache, retries, and context, with the data the check still needs before a recommendation can ship. The page is a prep surface, not a runtime optimizer.

Does this change the runtime path?+

No. Preparing the data plane does not change how your code calls the model. Readiness checks run over events the ledger already records.

What is the relationship between readiness and the AI Spend Audit?+

Both run on the same metering ledger. The AI Spend Audit produces a one-time delivered report. The optimization readiness page is the ongoing surface that shows which checks have enough data to produce a recommendation.

What is the first integration step?+

Create a P402 key, send metered events, then open the optimization readiness page. The readiness checks run on the events the ledger already holds.

Does the readiness surface require sharing prompts?+

No. Default privacy mode is metadata-only. Readiness checks read owner, model, tokens, cost, outcome, and evidence status. Prompt content stays out unless tenant policy opts in.

Get started

Prepare the data plane first. The claim comes after.

Readiness checks are live. Recommendations are gated. Start with the Optimize surface, or read the docs to see what data each readiness check needs.

See optimizeRead docs