Case studies

Selected engagements.

A look at three recent engagements across professional services, legal, and logistics. Client names are withheld for confidentiality.

Case 01

Intellectual property

Org
GCC-based IP firm, ~30 attorneys, retained corporate clients
Process
Opportunity discovery — patent monitoring through quarterly client reports

Friction points identified

Manual patent database scanning
Associates query 3 databases (USPTO, WIPO, GCC IP Office) per client, dedupe by hand, ~12 hrs per client per quarter
Competitor filing alerts lost in volume
300+ daily alert emails per associate; ~70% never opened; relevant filings missed downstream
Quarterly opportunity report drafting
Custom write-up per client, templates underused, ~6 hrs per report, inconsistent format across attorneys

AI-fit assessment

Patent database scanning + dedup
Structured extraction across well-formed sources; deduplication is deterministic
High
Competitor filing triage + classification
Classifier agent over inbound feeds; ranks by client relevance and IP class
High
Opportunity report drafting
LLM drafts first version from structured findings; senior attorney signs off before send
Medium

Build proposal

  • Phase 1: Patent monitoring agent + unified findings store — $16,000, 6 weeks
  • Phase 2: Opportunity report drafting layer with attorney review gate — $9,000, 4 weeks (60 days after Phase 1)

What the audit found

The firm came in asking us to automate opportunity discovery. The audit established they were already finding opportunities — the bottleneck was triage and prioritization across 300+ daily alerts that never got opened. Reframing the brief from "find more" to "lose fewer" changed both the build scope and the expected ROI.

Case 02

Higher education

Org
Mid-sized UK university, ~7,000 undergraduate students, essay-heavy curriculum
Process
Assignment lifecycle — submission through grade return to student

Friction points identified

Fragmented submission intake
Students submit via LMS, email, and shared drives; TAs spend ~4 hrs per course per week consolidating
Inconsistent rubric formats across faculty
Each professor uses their own format (Word, Excel, verbal); blocks any standardised grading layer
Essay grading turnaround
14–21 day return time vs 7-day target; TA bandwidth is the binding constraint

AI-fit assessment

AI essay grading (as originally requested)
Unreliable on inconsistent rubrics; high risk of faculty rejection on first deployment
Low
Submission intake unification
Single ingestion endpoint + auto-routing by course code + plagiarism flag
High
Rubric standardisation tool for faculty
Assistant that converts existing rubrics to structured schema; voluntary adoption
Medium

Build proposal

  • Phase 1: Submission ingestion + routing pipeline — $22,000, 8 weeks
  • Phase 2: Rubric standardisation tool + faculty onboarding — $14,000, 5 weeks
  • AI grading deferred to Phase 3, contingent on rubric coverage above 60% across active courses

What the audit found

The university came in asking for AI grading. The audit established that grading wasn't the root issue — fragmented intake and inconsistent rubrics were. AI grading deployed on the existing foundation would have produced unreliable results and damaged faculty trust on first contact. By sequencing the work, we kept AI grading on the roadmap without setting it up to fail.

Case 03

Logistics

Org
Regional freight forwarder, 60 employees, 200+ shipments/week
Process
Exception handling — carrier notification to resolution and client communication

Friction points identified

Exception identification
Notifications across 7 carrier accounts, 2 staff monitoring, 30% miss rate at peak
Triage and routing
Manual categorization via Slack DMs, no audit trail, 2.5 hr triage delay
Client communication
Manual templated messaging, inconsistent tone and timing

AI-fit assessment

Exception identification
Email monitoring agent + structured extraction (type, carrier, shipment ID)
High
Triage and routing
Classifier + routing rules — needs taxonomy agreed with ops first
High
Client communication
Templated generation + human review gate is straightforward; full automation has tone risk
Medium

Build proposal

  • Exception identification + triage pipeline — $18,000, 8 weeks
  • Client communication module as Phase 2, optional — $6,000, 3 weeks

What the audit found

The 30% miss rate on exceptions was unknown before the audit. The firm tracked exceptions that were escalated, not exceptions that were missed entirely. The audit identified the gap by mapping the full process — not just the escalation log.

Have a process worth auditing?

Tell us about it. We respond within one business day.

Start the conversation