For the firm · Platform stack + Claude seat + governance

The platform engineering firms build on.

Snowball is the structured substrate — ingestion, deployment, visualization, governance — that turns frontier models like Claude into firm-specific tooling your engineers can build on. Clarion (water asset management) is our first vertical layer, running in production with engineering-firm partners today.

AI that takes engineering judgment to the field and amplifies its impact. Not AI that replaces the engineer.

The thesis

Four things we believe, and bet the company on.

We sell skills, not software.

The days of walled software are over. Engineering firms don't need another product to evaluate. They need a platform they can build their own tooling on top of.

We do not compete with frontier models.

We are the harness for Claude and its peers — the substrate that turns a frontier model into firm-specific tooling. Our seriousness is about where the model gets plugged in, not about the model itself.

Clarion is our first vertical layer, not our product.

An engineering firm built Clarion on top of Snowball to serve water-utility clients. Future verticals — capital planning, compliance, adjacent infrastructure sectors — follow the same pattern. The platform is the thing; the vertical is the proof.

Engineers are the leverage point.

Not software. Not consultants. Not regulators. The firm is the protagonist; the utility is the beneficiary; Snowball is the club director that recruits and arms the coaches.

The ROI math

Half-day payback, every month.

The median US knowledge worker earns $56.50/hr. Typical reported AI time savings are roughly 50 minutes per day — about $47 of value. A monthly enterprise-AI seat costs $20–60. The seat pays for itself in half a day, and you have nineteen and a half workdays left.

$56.50/hr
Median US knowledge-worker wage
50 min
Average daily AI time savings
$47
Daily value unlocked per worker
$20–60
Monthly enterprise-AI seat cost
99%
Inference cost drop (12 months)
25×
Token-demand growth since Dec 2024
91%
Software-dev cost drop (8 months)
AI agent task-duration growth (2025)

Figures drawn from ARK Invest's Big Ideas 2026 report (AI Infrastructure & AI Productivity sections) and a Spark Labs presentation by Jozef Soja, April 2026. Updated quarterly.

The platform stack

Five layers. One substrate.

What your firm gets when you sign on. Water firms get Clarion on top. Firms in other infrastructure verticals co-develop the first layer with us.

Ingestion

Your client's data infrastructure — AMPs, SCADA, GIS, hydraulic models — loaded into a structured layer Claude can reach into. Water is where we start; other verticals are the same mechanic.

Deployment

Engineers configure in natural language from Claude Desktop; the platform does the mechanical work. Hours, not months. No ticket queue, no implementation consultants.

Visualization

A field surface the 60-year-old operator will actually use and a research-grade surface for the licensed P.E. — two registers, same substrate. Clarion is the current example for water.

Governance wrapper

Enterprise Claude seat hosted on our AWS Marketplace contract. Post-quantum cybersecurity certification in progress. The compliance frame your IT would have built in six months, absorbed.

Skills that compound

Every firm that builds on Snowball contributes a vertical pattern. Platform learns. The next firm starts further along. You are not the first firm figuring it out alone.

“We're not competing with frontier models. We're building the harness to use them.”

The honest mechanism

What Snowball owns. What you own.

We don't deliver water infrastructure. We don't replace the engineer. Here is the five-step sequence — Snowball owns the first three, your firm and the utility own the rest. Load-bearing boundary, drawn on purpose.

01

Up-tool the firm

Platform access + Claude seat + governance + training.

Snowball
02

Activate the plan

The engineer parses an AMP into a running data system — days, not months.

Snowball + engineer
03

Track real data

NRW, work orders, inspections, condition records — the utility has operational truth.

Snowball + engineer
04

Unlock funding

Grant applications get the proof they need. SRF, state AM incentives, federal programs.

Firm + utility
05

Real infrastructure work

Pipe replacement, treatment upgrades — done by the firm and contractors.

Firm

Steps 4 and 5 happen because of Snowball but not by Snowball. We stay in our lane. You stay the engineer.

30 minutes, peer-to-peer, no deck.

Tell us where your firm is with AI today. We'll send a short memo on what adoption looks like for firms at your stage before we meet, so the conversation starts somewhere useful.