Arielle Frank

• 92% — Reduction in Manual Lead Review Time
• 70% — Of Clients Hire the First Firm to Respond

Employment law firms lose clients not from a lack of leads but from slow intake and query response times, with 70% of prospects hiring the first firm to reply. Every inquiry required a human to read, triage, and route it, which took 15 to 60 minutes per lead. High-value cases were buried in the same queue as low-priority ones, and the team had no way to prioritize in real time.

I rebuilt the intake process from the data layer up. I first redesigned intake forms to produce structured, machine-readable submissions optimized for LLM processing without additional modifications. I also built an AI automation layer in n8n that automatically summarizes, classifies by case type and complexity, scores against high-value criteria, and routes each submission to the right person. The system was architected to improve over time, with classification accuracy increasing as volume scales.

I achieved a 92% reduction in manual review of client intake forms, cutting the time needed from 60 minutes to under five minutes per inquiry. High-value cases surfaced immediately, increasing the win rate of high-value clients.

• 400% — Improvement in Identity Resolution Accuracy
• **45+ ** — Sources Unified Under a Single Governed Segment CDP Implementation

A large regional health insurer had more than 45 active Segment sources across many platforms and tools that tracked behavior independently with no consistent identity logic. Customer profiles were fragmented, profile merge rates were low, cross-channel reporting was unreliable, and automations misfired. No data foundation existed for personalization, lifecycle marketing, or AI adoption.

I first carried out a full identity resolution overhaul: isolated upstream merge failures, standardized identifiers across all sources, and optimized cross-source resolution logic. I then executed production-grade data backfills using Python and the Segment API to remedy missing data. I also designed a full personalization architecture for a digital plan enrollment tool spanning six functionalities: application progress saving, autofill personalization on return visits, AI-powered lead scoring, dynamic web content, real-time behavioral nudges, and multichannel retargeting.

I achieved a 400% improvement in identity resolution accuracy, improving cross-channel reporting accuracy. I also unlocked lifecycle automation, AI adoption, and a clear implementation roadmap for a fully personalized enrollment experience rollout.

• 1 Call — Generates Full Library of GTM Assets Automatically
• 2 Pipelines — Parallel Processing

Sales calls contain the richest GTM data a company produces—real objections, real buyer language, real signals—and almost none of it gets used. Insights sit in recordings nobody analyzes. Representatives operate without shared baselines. Marketing writes messaging disconnected from what customers actually say. The data exists, but there’s no system to extract and route it.

I built a multi-agent AI pipeline in n8n that transforms a single sales call into deployable GTM assets, encompassing objection scripts, coaching scorecards, optimized talk tracks, personalized follow-up email, and competitor analysis. Two parallel pipelines run simultaneously: per-call processing daily, and aggregate pattern detection weekly across all call volume. QA guardrails and human-in-the-loop routing ensure nothing ships without passing validation.

One sales call automatically generates a full GTM asset library, work that previously took hours of manual synthesis per rep or never happened at all. The aggregate pipeline compounds over time, surfacing market-level patterns from call volume.