Between January 2025 and January 2026, INS built fourteen production applications: enterprise NetSuite integrations, industrial protocol platforms, network monitoring tools, and sales engineering utilities. None of them would exist without AI. Not because AI made them faster. Because AI made them possible.
This isn't a story about augmenting existing development capacity. It's a story about creating capacity that didn't exist. A company without a traditional software development team built 119,000 lines of production TypeScript, industrial-grade Python applications, and a suite of web tools that our customers and sales engineers use daily.
The portfolio spans five categories: enterprise workflow automation, industrial protocol platforms, network monitoring applications, technical sales tools, and industrial automation platforms. Together they represent what becomes possible when imagination and domain expertise meet AI capability in the right way.
Unleashing our imaginations coupled with our domain knowledge really opens up what's possible. These projects aren't just productivity tools. They're answers to questions we hadn't thought to ask, solutions to problems we'd accepted as unavoidable, and capabilities we couldn't have built any other way.
The Projects
Each project solved a real business problem. None were experiments or proofs of concept. All are in production use, serving customers and internal operations.
NS-Office-Add-In (Time Entry)
Dual-platform time entry application running as both a Microsoft Outlook Office Add-in and Progressive Web App. OAuth 2.0 authentication with NetSuite, AES-GCM encrypted token storage, cascading customer/project/task dropdowns, real-time budget tracking, Copy Week/Copy Day features, and budget overrun notifications to project managers.
ns-expense (Expense Report)
Expense report application with receipt upload, image compression with EXIF orientation correction, HEIC detection, billable/non-billable workflows, and dual budget tracking for Engineering Expenses and Billable Consumables. Built in 15 days by applying domain knowledge from the Time Entry application.
The Time Entry application took two months. The Expense Report application, built on the same architecture and domain knowledge, took fifteen days. Same quality standards. Same test coverage. Same production readiness. Domain knowledge compounds when captured in reusable patterns.
VibeCIP
High-performance ODVA-compliant EtherNet/IP integration platform. Python/FastAPI backend implementing the industrial CIP protocol with React/TypeScript frontend. Supports 200+ concurrent scanner instances with connection pooling, async packet processing, and real-time I/O via UDP Class 0/1 and TCP Class 3 transports.
Why VibeCIP Matters
VibeCIP demonstrates something significant: AI can help build industrial protocol implementations that comply with published standards like ODVA CIP. The protocol layer handles encapsulation, session management, and I/O connections exactly as the specification requires. This isn't a toy implementation. It's production infrastructure for Allen-Bradley PLC integration.
CellMetrics
Professional cellular network monitor for field technicians performing coverage audits on private LTE/5G networks. PyQt6 desktop application with real-time signal metrics (RSSI, RSRP, RSRQ, SINR), GPS tracking, iperf3 bandwidth testing, and SQLite data collection. Includes router integration for Cradlepoint and Peplink devices.
NetLat
Network latency utility for diagnosing connectivity issues and measuring round-trip times across network paths. Python-based tool used internally for troubleshooting customer network performance.
celona-ap-matrix
Interactive comparison matrix for Celona access point specifications. Helps sales engineers quickly compare AP models for customer deployments.
Cellular-Throughput-Calculator
Calculator for estimating cellular throughput based on signal conditions, modulation, and bandwidth allocation. Used during network design and customer consultations.
Antenna_Ground_Intercept
Tool for calculating antenna coverage patterns and ground intercept points. Essential for site surveys and deployment planning.
Spectrum-Viewer
RF spectrum visualization tool for understanding frequency allocations and identifying interference sources.
CBRS_Guide
Comprehensive reference guide for Citizens Broadband Radio Service (CBRS) regulations, spectrum access, and deployment considerations.
DPA_Zones
Dynamic Protection Area zone mapping tool for understanding CBRS spectrum availability near protected federal installations.
WiFi and CBRS Report Generators
Browser-based automated report generation tools for wireless network engineers delivering site surveys and acceptance tests. Self-contained HTML applications with AI-powered analysis via xAI Grok, supporting Ekahau data (WiFi) and multi-vendor test equipment formats including Keysight, R&S, Viavi, and XCAL (CBRS). Generates Site Survey, SAT, FAT, and PoC reports with interactive Leaflet.js maps. Reduces report generation from 4-12 hours to 15-30 minutes (85-95% time savings).
AIMS 2.0
Industrial automation platform built on Ignition SCADA for manufacturing environments. Features ISA-95 compliant equipment hierarchy via Unified Namespace, context-aware operations dashboards, Material Design 3 Perspective UI, and real-time monitoring. Compressed 14 months of traditional development into 2 months with AI.
DevIt
Web-based development environment and project manager for Ignition SCADA projects. React/TypeScript frontend with Node.js backend, featuring Docker-based dev environments, AI-powered assistance with specialized expert modes, drag-and-drop task management, and automated project documentation generation.
The Development Timeline
Twelve months of concentrated development, each project building on the last:
Each project made the next one faster. Not because the code was reused directly, but because the problem-solving patterns, the architectural decisions, and the understanding of how to work effectively with AI all compounded. Domain expertise plus AI fluency creates multiplicative results.
Why AI Made This Possible
Let's be clear about what happened here. This wasn't a traditional software development team using AI to code faster. This was domain experts with no professional programming background building production systems.
The Old Model: Access Through Developers
Before AI, building these applications would have required:
- Hiring developers who understand NetSuite's REST API, OAuth 2.0, and SuiteQL
- Finding specialists in ODVA CIP protocol implementation
- Contracting PyQt developers for desktop applications
- Explaining industrial networking domain requirements to people who don't live in this world
The translation layer between domain expertise and technical implementation was expensive, slow, and lossy. Ideas got simplified. Requirements got misunderstood. The people who knew what was needed couldn't build it, and the people who could build didn't know what was needed.
The New Model: Domain Experts Building
AI collapsed that translation layer. The person who understands why budget tracking matters for billable projects is the same person who builds the budget tracking system. The person who knows CBRS spectrum regulations is the same person who builds the reference guide. The person who troubleshoots EtherNet/IP integration issues is the same person who builds the CIP protocol platform.
What AI Actually Provided
- Technical translation: Converting domain requirements into working code without losing nuance
- Pattern application: Applying architectural patterns (Clean Architecture, repository pattern, dependency injection) consistently
- API navigation: Understanding complex APIs (NetSuite REST, OAuth 2.0, ODVA CIP) and implementing correct integrations
- Quality maintenance: Writing tests, handling edge cases, maintaining consistent code style
- Iteration speed: Moving from concept to working implementation in hours rather than weeks
The Skill Stack That Made It Work
Building with AI requires a different skill set than traditional development, but it's not the skill set most people expect.
What Mattered Most
- Domain expertise: Knowing what the applications needed to do and why. Understanding NetSuite workflows, industrial protocols, cellular network behavior
- Problem decomposition: Breaking complex goals into components that AI could implement and validate
- Quality recognition: Knowing what good looks like. Catching when AI output was subtly wrong or missing edge cases
- Architectural thinking: Making decisions about structure that would hold up as applications grew
- Persistence: Working through the inevitable failures and iterations to reach working solutions
What Mattered Less
- Syntax memorization: AI handles the details of TypeScript, Python, CSS
- API documentation reading: AI can navigate complex API documentation and implement correct calls
- Boilerplate generation: Project scaffolding, build configuration, test setup
- Code formatting: Consistency maintained automatically
The single most important skill was knowing what to ask for. Not "write me an expense report app," but "implement a budget tracking system that queries NetSuite Sales Orders for Engineering Expenses line items and calculates remaining unbilled amounts using SuiteQL." Domain expertise translated into precise requirements is the irreducible human contribution.
The Super Exponential Reality
We're on a super exponential timeline for AI capabilities. That's not hyperbole. AI capability is doubling roughly every four months. What was possible in September 2025 was less than what was possible in January 2026. What will be possible by mid-2026 will exceed both.
These fourteen projects represent a snapshot of what's achievable at this moment. The Time Entry application pushed against the edges of what AI could reliably handle in late 2025. The Expense Report application, built just weeks later, was measurably easier because both AI capability and our skill at using it had improved.
The learning compounds. Each project built skills that transferred to the next. Each interaction with AI taught patterns that applied more broadly. The gap between "domain expert with AI" and "domain expert without AI" is already substantial. By the end of 2026, it will be transformational.
People who figure out how to build with AI now, in early 2026, will have a compounding advantage over those who wait. Not because the tools won't be available later, but because the skill of working effectively with AI takes time to develop. The learning curve is the moat.
What This Means for INS
These fourteen projects represent a fundamental shift in how INS operates. The friction costs we used to accept as inevitable (the "NetSuite tax" of manual time entry, the spreadsheet-based expense tracking, the lack of real-time project budget visibility) are no longer fixed constraints. We can solve them ourselves.
Our New Capability
- We can build what we need. When sales engineers needed better tools for cellular network planning, we built them. When project managers needed budget visibility, we built it.
- Our domain expertise becomes the asset. Decades of industrial networking knowledge, NetSuite workflow understanding, and cellular deployment experience can now translate directly into working applications.
- We move at the speed of ideas. The Expense Report application went from concept to production in fifteen days. When we identify a problem, we can solve it.
- We control our own roadmap. No waiting for vendors to prioritize our feature requests. No compromise solutions because custom development was too expensive.
What We're Building Toward
The people closest to the problems at INS, the ones who understand why budget tracking matters for billable projects, who know the nuances of CBRS spectrum regulations, who troubleshoot EtherNet/IP integration issues daily, can now participate in creating solutions directly. Development capacity is no longer our bottleneck.
Our competitive advantage isn't access to better AI. It's our deep domain expertise in industrial networking, cellular infrastructure, and enterprise workflows combined with our growing fluency in building with AI. We know what to build because we live in this world every day.
The most exciting applications are still ahead: custom tools that deliver answers to questions we haven't even thought to ask yet. When imagination is no longer constrained by development capacity, the only limit is how well we understand our customers' needs. And we understand them very well.
2026: The Year Everything Changes
Unleashing our imaginations coupled with our domain knowledge really opens up what's possible. We believe 2026 will be a fundamentally transformative year for INS as we tap into our collective creativity and domain expertise to solve problems we used to accept as a "NetSuite tax."
AI's impact will go far beyond ERP and CRM, especially as we build custom applications that deliver answers to questions we hadn't even thought to ask.