Breadth · computed from the registry
One operator, 21 industries
The same instruments, applied across the industries where AI is being deployed today — 4 of them first-hand, the rest studied and labeled as such. Every count here is computed from the use-case registry, not asserted. Pick an industry to see everything in it.
21 industries4 first-hand domains69 worked scenarios
Financial services9 first-hand10 scenarios
- Enterprise bank — MCP core + A2AProtocol Selection Lab · 40 systems and many agent teams — the two-layer stack wins.
- Credit decisions — autonomy by fair-lending riskHuman-in-the-Loop Approval · Auto-approve the clear cases; the thin-file applicant stays with a human.
- A retail bank's AI bookAI Initiative Portfolio Dashboard · Cards, fraud, KYC, wealth — governed like capital; two get cut.
- Fraud model: build or buy?Build vs Buy vs Fine-Tune · Data gravity and real-time latency pin it to self-host.
- Agent-assist for 900 servicing repsAdoption & Change Readiness · Handle-time scorecards quietly punish the reps who use it.
- Bank AI Center of ExcellenceCapacity & Resourcing Planner · In a regulated shop the platform and model-risk skills gate everything — and both are short.
- Credit / loan eligibility decisioningAI Compliance Readiness Navigator · High-risk plus rights-affecting — fair-lending and model-risk stack on top of the Act.
- Bank AI Center of Excellence upskillingTalent & Upskilling Planner · Governance and model-risk carry the highest targets — and can't be partnered out.
- Fintech feature-team mobilizationOnboarding & KT Tracker · Prod-data and compliance access gate the ramp; the founding engineer holds every runbook.
- Board QBR on the AI programExecutive Communication Studio · A board read-out that ends in the two decisions only the board can make.
Healthcare6 scenarios
- Wrap an EHR (FHIR R4) for a care agentMCP Server Playground · Read clinical data freely; propose orders — never auto-write.
- Clinical note → claim codesTool-Use & Structured Output · A wrong-side modifier is caught by the gate before the claim writes.
- Clinical orders — acuity as the autonomy dialHuman-in-the-Loop Approval · Low-acuity orders auto-file; high-acuity always waits for a clinician.
- A health system's AI bookAI Initiative Portfolio Dashboard · Clinical and admin AI, allocated under one lens.
- Ambient scribe for 2,000 cliniciansAdoption & Change Readiness · The model works; the clinicians don't trust it yet.
- Clinical decision support / triageAI Compliance Readiness Navigator · It may be a regulated medical device before it's an AI-Act system.
Manufacturing5 scenarios
- MES / historian, read-only across OT/ITMCP Server Playground · Everything is read-only — the agent can observe the plant, never actuate it.
- A manufacturer's AI bookAI Initiative Portfolio Dashboard · Predictive maintenance to procurement — where the money actually is.
- Defect-detection rollout on the lineBusiness Case / ROI Builder · The scrap-rate assumption is the whole business case.
- Plant-floor AI capability buildTalent & Upskilling Planner · The gap isn't ML — it's the OT/IT and edge skills that make ML run on the floor.
- Plant-leadership update tying AI to OEEExecutive Communication Studio · Everything framed in OEE — the metric plant leadership actually owns.
Insurance4 scenarios
- FNOL triage — the validation gate as fraud controlAgent Loop & Failure Inspector · A hallucinated policy number gets rejected at the gate — which is also a fraud control.
- Document-heavy claims processingInference Cost Forecaster · Bursty claims volume means idle GPUs — utilization is the whole story.
- Claims-automation program — the drift hidesDelivery Health & RAID Radar · Extraction reads green on rollout while model drift pulls it amber.
- Claims triage & extraction estimateEstimation & Scoping Studio · The blow-up isn't the model — it's the messy claims documents nobody scoped.
Pharma & life sciences4 scenarios
- Competitive-intel brief on a Phase II readoutMulti-Agent Orchestration Board · Four agents turn a rival's trial data into a defensible brief.
- Regulated-research LLM vendorVendor Evaluation & Risk Monitor · Security and validation dominate — and concentration risk lurks.
- R&D discovery-informatics programStakeholder & Sponsor Cockpit · Regulatory/QA is the gate; bench-scientist trust is the adoption you can't mandate.
- Evidence-synthesis assistant estimateEstimation & Scoping Studio · SME-graded eval and GxP documentation are the hidden two-thirds of the estimate.
Retail & e-commerce4 scenarios
- Order management + returns as toolsMCP Server Playground · Mutations are idempotent and rate-limited so a retrying agent can't double-refund.
- Refresh copy for 4M SKUsPrompt Cost & Token Simulator · Caching the shared brand prompt collapses the seasonal bill.
- Personalization program — adoption reversesDelivery Health & RAID Radar · Search ships green while merchandisers quietly revert to manual.
- Peak-season personalization bidRFP/RFI Response War Room · Strong fit and a hard date — win on measurable lift, not lowest price.
Capital markets3 first-hand3 scenarios
- Equity-research swarm on an investment thesisMulti-Agent Orchestration Board · Analyst, risk, and a compliance critic pressure-test a thesis.
- Client goals across quarterly reviewsContext & Memory Engineering · The memory policy is the relationship — what survives between reviews.
- Wealth-platform advisor-assist programStakeholder & Sponsor Cockpit · Compliance can veto; the field advisor council decides whether it's ever used.
Consulting & prof. services3 first-hand3 scenarios
- Global-SI delivery pod across engagementsCapacity & Resourcing Planner · 48 people staffed across five client engagements — short in exactly one skill that gates all of them.
- Global-SI delivery pod mobilizationOnboarding & KT Tracker · Offshore client-system clearance is the longest pole — and it's not on the training plan.
- Partner update on the firm's AI capabilityExecutive Communication Studio · A partner read-out that leads with the investment decisions, not the build status.
Legal3 scenarios
- Contract-review swarm on an MSAMulti-Agent Orchestration Board · A four-agent swarm redlines a vendor MSA — the critic earns its keep.
- 50k-document e-discovery under budgetContext & Memory Engineering · At corpus scale, compress or hand off — full-dump is off the table.
- Contract analysis: buy, build, or fine-tune?Build vs Buy vs Fine-Tune · The firm's precedents are the moat — does that flip it to fine-tune?
Public sector3 scenarios
- Vendor under data-sovereignty rulesVendor Evaluation & Risk Monitor · Sovereignty and procurement rules become weights — OSS rises.
- Citizen-services automation programStakeholder & Sponsor Cockpit · The sponsoring CIO cools while a records officer quietly gains gate power.
- Benefits-eligibility AI platform bidRFP/RFI Response War Room · A biddable win — but only if you close the ATO and past-performance gates first.
Technology / SaaS3 scenarios
- A 40-turn support thread across two sessionsContext & Memory Engineering · As the thread grows, summarize beats full-dump on cost and overflow.
- Startup — two tools, ship this quarterProtocol Selection Lab · One agent, two tools — a protocol is premature.
- A chat feature at tens of millions of DAUInference Cost Forecaster · At consumer scale, the self-host cliff arrives early.
Telecom3 first-hand3 scenarios
- NOC incident correlation — 503 under loadAgent Loop & Failure Inspector · A tool 503s during the incident; retry/backoff plus a cached fallback keeps it moving.
- Summarize 1.2M care calls a monthPrompt Cost & Token Simulator · The batch discount is the go/no-go on the whole program.
- Network-ops assistant across NOCsInference Cost Forecaster · Portfolio run-rate across NOCs — the cliff is real but mid-horizon.
Cybersecurity2 scenarios
- SOC alert triage — breaking the flapping-alert loopAgent Loop & Failure Inspector · The agent loops on a flapping alert; the loop-breaker caps it and escalates.
- Autonomous SOC triage — the disciplined no-bidRFP/RFI Response War Room · Attractive scope, un-survivable terms — the senior move is to decline and say why.
Education2 scenarios
Energy & utilities2 scenarios
HR & talent2 scenarios
- HR-tech AI feature teamCapacity & Resourcing Planner · The scarce skill isn't engineering — it's the fairness scientist who keeps the feature legal.
- Resume screening / candidate rankingAI Compliance Readiness Navigator · Employment is high-risk by category — bias-audit law applies even with a human in the loop.
Logistics & supply chain2 scenarios
Marketing & adtech2 scenarios
Media & entertainment2 scenarios
Real estate / proptech2 scenarios
Coverage grows as the use-case layer is authored across every lab; the counts above are read live from the registry. Dot color: filled = first-hand (I ran the pattern), grey = studied (informed by public industry patterns, with sources on each brief).