Skip to content
Clooney Network Clooney Network Book a Platform Demo

Customers · Production evidence

412 production customers.
No logo theater.

Most of the case studies on this page are anonymized on purpose. Logo rights take quarters; engineering teams evaluating orchestration layers need signal today. What follows is six archetypes, one deep dive, and the receipts — inference volume, uptime, and the median 11-day path from contract to first production agent.

No procurement required — 30 minutes, your stack or ours.

Six archetypes running on Clooney today

The breadth of the install base, named by shape rather than by logo.

From a Series C vertical-AI agent company to a Fortune-200 bank's innovation unit, the patterns below describe real workloads shipping on the platform in the last two quarters. Names are composite; metrics are not.

  1. 01

    Series C vertical-AI agent startup

    60 engineers · multi-tenant SaaS · 14M agent invocations / month

    Shipped four customer-facing agents (claims, scheduling, document review, voice triage) on a single Clooney control plane. Replaced a homegrown Python router and three MLOps tools. Reached SOC 2 Type II readiness in 11 days from contract to first production agent.

    Agents in production
    14
    Models routed
    9
    Inference spend reduction
    52%
  2. 02

    Top-5 US bank, innovation unit

    Regulated · regional isolation in US-East and EU-Frankfurt · 28M invocations / month

    Internal "agent factory" serving compliance, KYC, and underwriter copilots across three lines of business. Dedicated regional isolation since Q3 2023, HIPAA-eligible deployment posture, deterministic trace capture for every model call feeding an examiner-bound decision.

    Audit-grade traces / day
    920k
    PII redaction at router
    100%
    LoBs served
    3
  3. 03

    Public SaaS data platform, Series F

    1,800 employees · customer-facing copilot · 86M invocations / month

    Embedded an AI analyst copilot across the analytics suite. Uses Clooney's cost-aware routing to keep per-tenant inference spend predictable; uses the trace-correlation engine to reproduce customer-reported regressions in under four minutes on average.

    Tenants routed
    4,400
    Mean regression triage
    3.8 min
    Spend predictability (MAPE)
    4.1%
  4. 04

    US health system, clinical operations

    HIPAA-eligible · AP-Singapore & US-West regions · 6M invocations / month

    Ambient-documentation and prior-authorization agents running across 14 hospitals. Clooney sits behind the EHR integration layer and enforces model allowlists, prompt redaction, and PHI-aware logging before any model provider sees the payload.

    Hospitals live
    14
    Allowlisted providers
    3
    Avg. clinician time saved
    42 min / day
  5. 05

    Global e-commerce platform

    Peak traffic · 312M invocations / month across 4 regions

    Search, merchandising copy, and post-purchase support agents routed through Clooney with provider-fallback chains that hold latency budgets during Black Friday peaks. The platform absorbed a 6.4× traffic spike on a single Friday in November 2025 without a control-plane incident.

    Peak day volume
    38M
    Fallback chain depth
    4
    Holiday traffic multiplier
    6.4×
  6. 06

    Public-sector research lab

    Air-gapped tenant · on-prem control plane · 1.4M invocations / month

    Runs the Clooney control plane on dedicated hardware inside an isolated environment. Used to coordinate multi-agent scientific literature reviews across grant-funded research teams; trace graphs feed directly into the lab's reproducibility archive.

    Air-gapped tenants
    1
    Research teams
    23
    Reproducibility index
    99.2%

Practitioner voice

"We had four repos of glue code pretending to be a router and three dashboards pretending to be observability. Clooney didn't add a feature — it retired a category. Eleven days after signing, we had our claims-triage agent behind a real SLA, with traces an examiner could actually read. The thing I keep telling other directors is that the control plane is the product; the model is now just a row in a routing table."

Director of Platform Engineering Series C vertical-AI agent company · archetype 01

By the numbers · as of January 2026

The receipts an engineering leader scans first.

412
Production customers
Including 9 Fortune 500 enterprises
4.2B
Model + agent invocations / month
Across the entire customer base
99.97%
Trailing-12-month orchestrator uptime
Measured at the platform control plane
11d
Median time to first production agent
2024 customer cohort, signed to shipped
$28M
67
4.1k
14

Deep dive · archetype 01

From contract to production agent in 11 days — a Series C vertical-AI team tells the actual timeline.

No retrospective polish. The four steps below are the working notes the platform team kept during their first production deployment in Q3 2024.

Step 01

Pre-Clooney stack

A Python routing service maintained by two engineers, three separate MLOps tools for evals and traces, and a custom usage-metering pipeline feeding their billing system. Every new model provider was a six-week integration project.

  • Two engineers dedicated full-time to "router maintenance"
  • No consistent trace schema across agents
  • SOC 2 audit blocked on observability gaps for 4 months
Step 02

Decision criteria

The platform team's shortlist came down to four things: deterministic routing with cost-awareness, trace correlation across multi-agent graphs, HIPAA-eligible posture (not a checkbox — they had a pilot with a payer), and an SDK surface small enough that a new engineer could ship an agent in their first week.

  • Cost-aware deterministic routing across ≥ 5 providers
  • Trace-correlation engine benchmarked on a public suite
  • HIPAA-eligible deployment posture with BAA support
  • SDKs in Python and TypeScript with a single integration shape
Step 03

Integration timeline

Day 1: signed. Day 2-3: SDK install and a single-agent route against two providers. Day 4-6: cost policies written, fallback chains configured, eval suite ported to clooney-eval. Day 7-9: trace correlation validated against the SWE-Agent Trace Suite. Day 10: HIPAA-eligible tenant provisioning. Day 11: claims-triage agent cut over to Clooney behind a 99.9% control-plane SLA.

  • Day 1: contract executed
  • Day 3: first routed agent call from staging
  • Day 9: trace correlation passing internal benchmarks
  • Day 11: first production agent behind enterprise SLA
Step 04

Outcome, six months in

The router-maintenance backlog dropped to zero. Inference spend on the claims-triage agent fell 58% in the first quarter after cost-aware routing went live, without a measurable change in task-completion quality (measured against clooney-eval's held-out regression suite). Fourteen agents now ship behind the same control plane.

  • 58% spend reduction on the first agent routed through Clooney
  • 14 agents in production on one control plane
  • 0 router-maintenance tickets in the trailing 90 days

How the industry sees Clooney

Analyst recognition, open-source traction, and a deployment posture that clears regulated buyers.

  • Analyst · Q4 2024

    Strong Performer · The Forrester Wave™

    Recognized as a Strong Performer in The Forrester Wave™: AI Orchestration Platforms, Q4 2024.

  • Analyst · Oct 2024

    Cool Vendor · Gartner AI Infrastructure Summit

    Selected for the "Cool Vendor in Agentic Infrastructure" designation at Gartner's AI Infrastructure Summit, October 2024.

  • Open source · GitHub

    4.1k stars across three SDKs

    clooney-router, clooney-trace, and clooney-eval maintained by the platform team under Apache-2.0.

  • Security posture

    SOC 2 Type II · HIPAA-eligible

    Dedicated regional isolation in US-East, US-West, EU-Frankfurt, and AP-Singapore since Q3 2023.

Clooney is an orchestration and observability layer. We do not own customer model outputs, customer data, or any customer IP. Pricing is custom-quoted post-demo.

Next step

If the patterns above look like your stack, the next 30 minutes is yours.

Bring a workload, a routing decision you're stuck on, or a regulator's question. We'll walk through how Clooney would handle it on your tenant, with your model list, in your region.

No procurement required · 30 minutes · your stack or ours.