Case Studies

Real workflows. Real results.

A snapshot of teams we've shipped production AI workflows for — what we built, how long it took, and the operational lift it delivered.

SupportFlow AI dashboard — Halcyon case study
Halcyon · SaaS · 2025

AI-powered support automation for a growing SaaS platform

We deployed a tier-1 support agent trained on product docs, FAQs, and 18 months of historical tickets. It drafts, resolves, or routes — with every escalation arriving in the rep's queue with full context attached.

70%
Faster Replies
40%
Less Ticket Load
2 wk
Build Time

The Challenge

Halcyon's support queue was doubling quarter-over-quarter, but headcount couldn't. Repetitive tier-1 questions were burning out senior reps, while genuinely complex tickets were stuck behind them in the queue.

What We Built

  • Retrieval pipeline over product docs, changelog, and 18 months of resolved tickets
  • Confidence-gated auto-resolve for low-risk categories
  • Context-attached escalations that drop reps directly into the relevant thread
  • Eval suite catching tone, accuracy, and policy drift before each release

"Our reps stopped answering the same five questions all day. They're finally working on the hard problems they were hired for."

Head of Support, Halcyon
Northwind claims triage dashboard
Northwind · Logistics · 2025

Claims triage agent for a national logistics operator

Northwind processes tens of thousands of damage and delay claims a year. We replaced a four-person triage queue with an agent that classifies, prioritises, and pre-fills resolutions in under 90 seconds end-to-end.

68%
Faster First Reply
3.1×
Throughput / FTE
21 d
Time to Deploy

The Challenge

Claims sat in a triage queue for an average of 4.7 days before a human even read them. By the time a resolution was offered, the customer was already escalating to a manager.

What We Built

  • Multi-class intake classifier with photo + document understanding
  • Policy-aware resolution drafter trained on past settlements
  • Priority router that surfaces high-risk claims to humans first
  • Audit trail and human-approval gate on every payout decision

"The triage team didn't disappear — they got promoted. They now spend their day on complex claims that actually need judgment."

VP Operations, Northwind
Foundry 9 lead enrichment dashboard
Foundry 9 · B2B Marketplace · 2025

Lead enrichment pipeline for a B2B procurement marketplace

A quiet background workflow that enriches every inbound lead with firmographics, technographics, and intent signals before it reaches a rep. CRM hygiene that doesn't depend on humans remembering to do it.

Enrichment Coverage
$0.12
Per Lead
24/7
Always On

The Challenge

Foundry 9's SDRs were spending the first 15 minutes of every call hunting for context that should already have been in the CRM. The enrichment tools they paid for covered only 20% of inbound leads.

What We Built

  • Multi-source enrichment fanout with deduplication and conflict resolution
  • Technographic signal extraction from public footprint
  • Intent scoring blended from product signals and third-party data
  • Idempotent CRM writes — never overwrites a human-edited field

"Our SDRs open a lead and the story's already there. They're calling prepared. Conversion is up and ramp time is down."

Director of RevOps, Foundry 9
Meridian legal AI copilot dashboard
Meridian · Legal · 2025

Private internal copilot for a 600-person legal team

A role-scoped assistant grounded in Meridian's matter archive, playbooks, and template library. SSO from day one; an evals suite that catches drift before it reaches the partner reviewing the draft.

42 min
Saved Per Matter
SSO
Audit Log
Live
Continuous Evals

The Challenge

The firm wouldn't touch a public AI tool — confidentiality, privilege, and audit requirements made it a non-starter. But associates were spending hours per matter on tasks the technology could clearly accelerate.

What We Built

  • Private-deployment copilot with role-scoped retrieval over the matter archive
  • Template + clause library grounded in firm-preferred language
  • SSO, full audit logging, and per-matter access controls from day one
  • Evals suite covering accuracy, citation integrity, and tone drift

"It's the first AI tool we've trusted with a real matter. The audit trail alone is why we got partner buy-in."

Knowledge Counsel, Meridian

Could your need for Automation be next?

Tell us what you'd like to automate. We'll come back within a business day with an honest read on scope, timeline, and price.

Talk to an expert Book an intro call