AgenticQuay.aiAccelerate Superintelligence
Book a call
AgenticQuay.aiAccelerate Superintelligence
Book a strategy call
  1. Home
  2. Case Studies
Case studies

What an AgenticQuay engagement looks like in practice

Eleven representative engagements across our consulting, integration, development, managed AI and training services — the challenge, what we built, and the outcomes clients measure.

Representative engagements. They show how we typically deliver; client names are withheld and the figures are illustrative.

Regional health networkHealthcare
3×
faster prior-authorization packet assembly
ClaudeMCPFHIRHuman-in-the-loop
Integration · Development

Claude agents that assemble prior-authorization packets

Challenge

Nurses spent hours copying chart notes into payer-specific forms, and incomplete documentation drove avoidable denials.

What we built

MCP servers over the EHR (FHIR) and payer policy library, plus a Claude agent that drafts packets and medical-necessity letters for clinician sign-off.

Results

  • Packets ready in minutes, not hours
  • Fewer documentation-related denials
  • Every statement linked to its chart source
National law firm · 400 seatsLegal
86%
weekly active use 60 days after launch
Claude EnterpriseSkillsTraining
Claude consulting · Training

A governed Claude Enterprise rollout across practice groups

Challenge

The firm had licenses, but usage sat with a handful of partners and IT worried about client confidentiality.

What we built

SSO and SCIM, practice-group Skills for clause review and deposition summaries, an approved document-system connector, a usage policy and role-based cohorts.

Results

  • Most lawyers using Claude every week
  • Standard Skills in every practice group
  • Policy signed off by the GC and IT
About our Claude practice
Mid-market P&C insurerInsurance
60%
of submissions triaged without manual re-keying
ClaudeTool useDocument AIIntegration
Development · Integration

A Claude submission-intake agent for commercial underwriting

Challenge

Underwriters re-keyed ACORD forms and loss runs from email, so quotes went out days later than brokers wanted.

What we built

A Claude agent that extracts ACORD and loss-run data into a validated schema with tool use, enriches it with third-party data, scores appetite and feeds a review queue inside the policy-admin system.

Results

  • Faster quote turnaround for brokers
  • Underwriters focused on risk, not data entry
  • Full audit trail for every decision
About our Claude practice
Major US food importer · 40+ countries of originSupply chain
42%
fewer late or held inbound containers in two quarters
ClaudeClaude Agent SDKMCPERP + TMS
Integration · Development · Managed AI

Claude agents that run inbound supply chain for a food importer

Challenge

Planners tracked thousands of containers across email, carrier portals and spreadsheets. FDA holds, missing documents and port delays surfaced too late, so stock-outs and demurrage kept climbing.

What we built

MCP servers over the ERP, transportation system, carrier and customs-broker feeds, plus Claude agents built with the Claude Agent SDK that reconcile shipment documents, flag FSVP and Prior Notice gaps, forecast arrival risk and draft re-routing and reorder proposals for planner approval.

Results

  • Document gaps caught before vessels sail
  • Demurrage and detention charges cut sharply
  • Planners approve exceptions instead of chasing them
About our Claude practice
Global testing company · 120+ countriesEducation & testing
5×
more remote sessions reviewed per proctor
ClaudeVisionHuman-in-the-loopManaged AI
Development · Managed AI

An AI proctoring solution built on Claude for a global testing company

Challenge

Remote exam volume outgrew live proctoring. Reviewers watched hours of footage to find a few incidents, flags were inconsistent across regions and candidates waited weeks for results.

What we built

A Claude-based review pipeline that reads session events, screen and webcam frames and ID checks, then writes an evidence-linked incident summary with a risk score. Proctors make every final call in a review console. Built with bias testing, data-retention controls and regional data residency.

Results

  • Proctors focus on the sessions that need them
  • Consistent incident decisions across regions
  • Results released days sooner, with a full audit trail
About managed AI services
BookRails.ai · HVAC company near New York CityHome services
$1.2M
in booked jobs attributed to BookRails in the first year
OpenAI Realtime APIVoice agentMCPAgents SDK
Development · Integration · Managed AI

BookRails.ai: an AI revenue employee for an HVAC company

Challenge

During heat waves and cold snaps a third of calls went to voicemail. Web leads waited hours for a callback, open estimates were never chased and thousands of past customers never heard from the company again.

What we built

BookRails.ai, an AI revenue employee on the OpenAI Realtime API and Agents SDK. It answers every call and web lead around the clock in English and Spanish, texts back missed calls within a minute, qualifies homeowners on system age, symptoms and service area, and books jobs into the field-service calendar through MCP servers over the FSM, CRM and SMS stack. It also follows up open estimates, runs tune-up and replacement reactivation campaigns, and attributes revenue to each workflow in a weekly report.

Results

  • Every call and lead answered, 24/7
  • 31% of open estimates recovered into sold jobs
  • Owner sees the revenue each workflow produced
About our OpenAI practice
Immigration law practice · 3 offices, 25 attorneysLegal
60%
less attorney time per petition draft
OpenAIVoice intakeMCPHuman review
Development · Integration

An AI copilot for immigration attorneys

Challenge

Paralegals rebuilt the same H-1B, O-1, EB-2 NIW and family-based packets by hand. Intake calls in five languages ran long, evidence lived across email and shared drives, and RFEs arrived because exhibits did not match the forms.

What we built

A copilot on OpenAI models with MCP servers over the case-management system, document store and a curated library of USCIS policy-manual sections and AAO decisions. A multilingual voice agent runs client intake and status calls and writes a structured case summary. The copilot pre-fills forms, builds exhibit lists that cross-check every claim against the evidence, drafts support letters and RFE responses with cited sources, and tracks filing deadlines. Attorneys review and sign every filing.

Results

  • Petition drafts in hours instead of days
  • Fewer RFEs from mismatched evidence
  • Clients get status updates in their own language
About our OpenAI practice
Marketing SaaS for small businesses · 4,000+ SMB customersMarketing tech
3.4×
more inbound leads per customer after six months
OpenAIVoice agentMCPAEO
Development · Managed AI

An AI marketing platform that runs marketing for small businesses

Challenge

Small-business owners had no time for marketing. They paid for tools they never configured, rarely showed up in ChatGPT or Google AI Overviews answers, and let leads go cold over the weekend.

What we built

A multi-tenant platform on OpenAI models with MCP connectors to Google Business Profile, the website CMS, social accounts, email, ads and CRM. It runs AEO audits that track how often the business is cited by AI assistants, then writes FAQ, schema and local landing pages to close the gaps. A brand-voice engine drafts blogs, social posts, emails and review replies for one-tap approval. A voice and chat agent answers inquiries and books appointments, and automated nurture journeys follow every lead with a weekly ROI summary.

Results

  • Customers cited in AI-assistant answers for local searches
  • A month of content generated in one sitting
  • Every lead answered and followed up automatically
About our OpenAI practice
B2B SaaS company · 120 engineersSoftware
35%
shorter pull-request cycle time
Claude CodeCodexBootcamp
Consulting · Training

AI coding agents rolled out across engineering

Challenge

Engineers used AI coding tools ad hoc. Quality varied, conventions were ignored and nobody could show the impact.

What we built

A two-week bake-off of Claude Code and Codex on real tickets, shared CLAUDE.md and AGENTS.md conventions, CI hooks, a review checklist and a four-week bootcamp.

Results

  • Measurably shorter cycle time
  • One set of conventions across 40 repos
  • Review load tracked every month
About our Claude Code bootcamp
Multi-brand retailerRetail
24/7
multilingual order support through peak season
Realtime voiceChat agentManaged AI
Development · Managed AI

Voice and chat agents that absorb the holiday spike

Challenge

Contact volume tripled every holiday season, hold times spiked and temporary hiring never caught up.

What we built

A realtime voice agent and a chat agent with tool calls into order management and returns, warm handoff to people, and QA scoring on every conversation.

Results

  • Hold times stayed flat during peak
  • Routine orders resolved without an agent
  • Quality and cost monitored daily as a managed service
About managed AI services
PE-backed services groupProfessional services
6
portfolio companies trained in one quarter
WorkshopsChampionsChatGPT + Claude
AI Enablement & Training

An AI enablement program across a portfolio

Challenge

Operating partners wanted consistent AI adoption across companies that used different tools and had very different maturity.

What we built

Executive briefings, role-based workshops on ChatGPT and Claude, a cross-portfolio champions program, a shared use-case library and a quarterly adoption dashboard.

Results

  • Funded pilots at every company
  • A champions network sharing wins across the group
  • Adoption reported to the board each quarter
About AI training & workshops
Accelerators

What carries over to your project

Clients don't start from zero. These battle-tested components come from our past builds and shorten time to production.

Voice agent stack

Low-latency, multilingual telephony agents with interruption handling, call transfer and transcripts.

Cited retrieval engine

Retrieval with source-level citations, access controls and evaluation datasets out of the box.

Durable orchestrator

Stateful multi-agent workflows with retries, human approval steps and full audit history.

Connector library

FHIR, CRM, ERP, ticketing and messaging connectors plus MCP server templates.

“We'd seen plenty of AI demos. AgenticQuay was the first team that showed us systems running in production — and then built one for us on the same foundations.”
COChief Operating OfficerHealthcare services company
Ready when you are

Let's put AI to work in your business.

Book a 30-minute strategy call. You'll talk to an engineer who ships production AI — and leave with a clear next step, whether or not we work together.

Book a strategy call sales@agenticquay.com