Selected work · Anonymized client examples

AI that operates invisibly and delivers measurably.

A high-level look at recent projects we've shipped. Every system below is in production today, used daily by the client's team, and tied to a business metric that mattered before we arrived. Client names are anonymized; outcomes and timelines are real.

Common thread #1

Every engagement starts with diagnosis.

Each of these projects began with a brief that didn't quite match the actual bottleneck. We re-scoped before we built.

Common thread #2

Production from day one.

No proof-of-concept sandboxes that never make it past a board demo. The architecture is built for real volume, real data, and real users.

Common thread #3

Measurable on a business KPI.

Throughput, error rate, response time, NPS, hours reclaimed. We agree on the metric before we build, and we report against it after we ship.

Legal Technology Document Intelligence

Client profile

200-person legal services firm, mid-market, US-based. Multi-jurisdiction commercial contract workflow.

From 4 hours to 11 minutes per contract review.

The firm was spending 4+ hours per contract on manual clause extraction, risk flagging, and summary generation. Senior associate capacity was the constraint on revenue — not demand. We designed a document intelligence pipeline that combines a custom extraction model, a structured output layer mapped to their existing matter management system, and a reviewer-facing UI that lets associates accept, amend, or override AI output with a single click. Evaluation was built in from the first sprint so quality stayed stable as document volume scaled.

Processing time
95% ↓
Throughput per reviewer
17×
Time to production
8 weeks
Healthcare Operations Workflow Agent

Client profile

Regional multi-specialty healthcare group with high inbound referral volume across four specialty lines.

One agent replaced a 3-step manual triage with 14% error rate.

Four intake coordinators were manually triaging referrals — an 8–72 hour process with a 14% specialty routing error rate. Mis-routed referrals were costing the group both reputation and throughput. Kerning built an intelligent intake agent that classifies, routes, and drafts response communications automatically — with deliberate human escalation logic for any case that falls outside high-confidence bounds. Full audit trail on every decision so clinical leadership can sample and spot-check.

Routing error rate
14% → <2%
Avg response time
8–72 h → 4 m
Volume growth
+40% at flat headcount
B2B SaaS Product AI Integration

Client profile

300-person B2B SaaS, eight years of accumulated customer data across multiple internal systems.

An AI search layer that became the product's highest-rated feature in 60 days.

Eight years of customer data — support tickets, product feedback, usage logs — was functionally inaccessible to the product and CS teams. Insight required either an analyst ticket or a brittle dashboard nobody maintained. We designed and shipped a natural language query layer on top of the existing warehouse, scoped explicitly so any team member could ask real questions without needing to know SQL or the schema.

Feature NPS
72
Analyst time reclaimed
22 h/wk
Shipped in
11 weeks
Financial Services Revenue Copilot

Client profile

34-advisor wealth management firm, regulated, with deep long-cycle client relationships.

Sales team moved from 60% admin time to 60% selling time.

Advisors were spending the majority of their day on client prep, note-taking, and post-meeting follow-up — not client development. The constraint on the firm's growth was advisor calendar capacity, not advisor count. Kerning built a meeting intelligence layer integrated with their calendar and CRM: auto-generated prep briefs, real-time note structuring, and post-meeting action drafts ready to send.

Hours back per advisor
11/wk
Client-facing time
40% → 60%
Full team rollout
6 weeks

Which of these is closest to your situation?

The diagnostic takes 90 seconds and maps your specific profile to the AI opportunity area most likely to deliver real leverage. Or talk to the team and skip straight to a scoping conversation.

Take the Diagnostic Talk to the team