PRODUCT CASE STUDY: Scaling to $70M ARR

The 30-Day Turnaround of an AI-Native Onboarding Engine

Author: Edmund Russell | Senior Project Manager

Executive Summary

In early 2024, Silverware POS launched an in-house payment solution that quickly became victim to its own manual complexity. By October 2024, the initial rollout (supported by a team of six) had failed to gain meaningful traction. I requested ownership of the project and, within 30 days, designed and deployed an AI-native orchestration stack that reached “full effect” by November 2024.

By December 2024, just 60 days from takeover, the engine successfully scaled onboarding volume from less than $10M to $70M in monthly ARR ($840M+ annual run-rate) while reducing required staffing by 67%. This study details the technical architecture and product leadership required to achieve this rapid enterprise scale.

The Challenge: Three Layers of Friction

I identified that the initial launch failure was not a resource problem, but an architectural one. We were fighting three distinct scale-killers:

Merchant Communication Deadlock: Merchants frequently failed to provide critical technical and PCI data. The manual follow-up loop was inconsistent, causing projects to stall in silent limbo for weeks.
Capacity Matching Blindness: Scheduling used static one-hour slots. A merchant with three units was over-served, while a merchant with 30 units caused massive installer overruns and team burnout.
Data and Tool Fragmentation: Critical shipping info was trapped in inconsistent warehouse emails, while project state was split between ClickUp, Salesforce, and technician side-chats.

Phase I: The Foundation (Intake and Extraction)

The first goal was to regain control of unstructured data. I built an Email OS that automatically categorized inbound communications to prioritize urgent tasks and filter noise.

Technical Spotlight: Zero-Touch Extraction

Warehouse vendors sent shipping data in varying, non-standard formats. I used n8n and OpenAI to build an extraction engine that identified tracking variables, updated ClickUp records, and generated standardized merchant notifications with clickable UPS and FedEx links. This eliminated 10+ hours of manual data entry per week.

Phase II: The Execution Engine (Procurement and Visibility)

I moved the project from tracking to executing by centralizing all onboarding logic in ClickUp via dynamic forms and conditional logic:

Automated Procurement: Based on merchant hardware selections, the engine automatically updated project statuses and ordered required inventory.
Automated Licensing: The system identified licensing gaps and initiated automated conversations with both the client and Sales to secure necessary software permissions.
C-Suite Visibility: Built dynamic dashboards that updated leadership in real-time, pushing notifications for important events directly to the executive team via Teams.

Phase III: Capacity and Resilience (Stateful Orchestration)

To achieve the $70M ARR scale, I had to solve the supply-versus-demand problem of installer time.

Technical Spotlight: The Hardware Logic Matrix

I designed an Excel-based matrix that assigned specific time values to hardware SKUs (for example, a five-minute plug-and-play unit versus a 30-minute complex network install). The engine queried the hardware mix for each project, calculated the required window, and used Microsoft Bookings webhooks to offer only those specific blocks to the merchant. This ensured 100% schedule accuracy.

The Persistence Engine and HITL

To solve client non-responsiveness, I built a state-aware nudge system. The engine monitored deadlines and sent automated reminders, but automatically removed merchants from the sequence the moment data was ingested. For complex edge cases, I integrated Microsoft Teams human-in-the-loop (HITL) cards, where a single click from a team member in Teams dictated the next branch of the automated workflow.

Phase IV: Technician Enablement (Generative Setup)

The final bottleneck was technician support load. I shifted the technical burden from the manager to the engine.

The Generative Script Factory: Built an LLM-driven engine where installers selected the environment (Server vs. Workstation). The system generated custom, site-specific PowerShell code that enforced private network states and verified all software dependencies post-install.
Pre-Enablement ROI: Upon shipment, the system triggered automated prep instructions to the merchant, saving approximately 20 minutes of setup time per device during the live installation call.

Business Impact: The 60-Day Transformation

Financial Velocity: Scaled from less than $10M to $70M monthly ARR in 60 days.
Operational Leverage: 67% reduction in headcount (six staff to two) while volume grew 7x.
Installation Velocity: Increased from six units per hour to 15 units per hour (150% increase).
Quality Assurance: 100% address accuracy (Google Maps API) and near-zero configuration errors.

Product Philosophy: Operational Empathy

A recurring theme of this project was working around institutional constraints. When I lacked Salesforce API permissions, I built a reporting bridge that replicated the Salesforce schema. This allowed manual operators to update records with zero search time. This operational empathy ensured that our internal teams were as enabled as our merchants.

Technical Stack

Orchestration: n8n, Power Automate
Resource Management: Microsoft Bookings (webhooks), Microsoft Teams (HITL), Calendly (prototype)
Cloud and APIs: Google Cloud Storage (GCS), Google Maps API
AI Models: OpenAI GPT-4, Llama 3 (local via Ollama), Mistral, Qwen
Data: ClickUp, Excel logic matrix, Salesforce schema mapping