The work behind the advice
Production AI systems, AI governance, and global customer support operations
Two kinds of experience, and the value of a fractional AI leader is in having both: I've architected AI systems that run in production, and I've run the organizations that have to adopt them.
Building production AI: Scaled Comp
I co-founded Scaled Comp to solve two problems California employers and their counsel keep running into: hidden wage-and-hour violations that turn into PAGA lawsuits, and the lack of real settlement data when those lawsuits land. I architected the AI platform behind both products.
What it does
- Scaled Comp Analyzer turns daily time records into compliance intelligence and produces the documented "reasonable steps" record now required under AB 2288 / SB 92.
- Vector Index is a settlement intelligence platform built from 5,900+ tracked California matters, 37,800+ court documents, and 29,100+ LWDA filings — powering plaintiff firm profiles, mediator benchmarks, and comparable-settlement reports for defense counsel.
How it's built
- Production systems on the Anthropic Claude and OpenAI APIs: document classification, field extraction, cross-document reconciliation, retrieval, and multi-step agentic workflows
- Confidence scoring with human-review checkpoints for anything below threshold, any unresolved reconciliation conflict, and anything client-facing
- Evaluation harnesses and QA gates so output holds up under legal scrutiny
- Unit economics: model routing by task complexity, prompt caching, batch processing, cost-per-run modeling
The AI governance framework behind it
- Approved-tooling standards and an AI systems register
- Data-handling and confidentiality policy; subprocessor inventory
- Output validation and audit-trail standards
- Quarterly change-log cadence and shadow-AI remediation
The buy-versus-build rule I applied there is the one I bring to clients: build where the logic is proprietary and differentiating; buy commoditized infrastructure. Keep humans on the decisions where a confident wrong answer costs credibility.
Running the customer support organizations that adopt it
15+ years leading global support, customer experience, and operations organizations through growth, acquisition, and transformation.
Eptura
- Merged nine separately run product support teams from three acquisitions into one 24/7 global organization; 16,000+ customer accounts unified within a year
- Held 93% CSAT through the full consolidation
- Migrated all nine onto a single Salesforce Service Cloud instance; built a FedRAMP-compliant support model with security
- Launched a paid premium support tier and a critical-account retention process with Customer Success
Smartsheet
- Follow-the-sun support across EMEA, APAC, and the Americas for 10M+ users
- Five platform migrations — CRM, contact center, telephony, customer portal, community — plus acquisition integration
- Drove a 30% increase in Professional Services capacity
PatientPop
- Built support from scratch through hypergrowth ($20M to $100M+ ARR)
- Cut resolution time 82% (10.5 days to 1.9) in ten months; first-response rate from 58% to 99%
- Offshore team handling 50% of frontline volume within a year; built Tier 3 Technical & Shared Services
Earlier
Head of Global Customer Care at DJO Global (350-position offshore transition, $500M key-accounts portfolio). Senior Director, Client Experience Delivery at AMN Healthcare (enterprise Salesforce deployment). Global Director, Customer Success & Technical Support at Infrascale (built 24/7 support from scratch).
Tools I work in
Anthropic Claude API and Claude Code, OpenAI, Python, Pandas, SQL, RAG and agentic workflows, AI governance frameworks, Salesforce Service Cloud, Zendesk, Azure, AWS, Power BI, Tableau.
Talk through your situation
If you're weighing a fractional AI leader, an AI governance review, or a support-operations transformation, start with a conversation.
info@advising.la