Warm Thoughts Leadership Meeting · April 17, 2026

How AI Transformed
Our Plumbing Department

From plumbers who fix water problems to plumbers who care about your water.

0+

Calls analyzed daily

0

AI-powered systems

$0K+

Single customer relationship

Call IntelligenceWater QualityProcess BreaksVoice AgentTranscript AnalysisCustomer ContextRepeat CallersDepartment RoutingCoaching InsightsRevenue PipelineSLA TrackingRisk DetectionCall IntelligenceWater QualityProcess BreaksVoice AgentTranscript AnalysisCustomer ContextRepeat CallersDepartment RoutingCoaching InsightsRevenue PipelineSLA TrackingRisk Detection

Where It Started

Two Problems, No Clear Solution

01

Inconsistent Schedules

Plumbers getting up at 3 AM for emergency calls. No consistency for family life. We needed recurring, predictable work.

“The guys care about two things: a consistent schedule so they can spend time with their family, and the opportunity to make more money.”

02

No Service Plan Stickiness

Service contracts with no real value proposition. A popped pipe is a one-time fix — there's no continued relationship.

“Craig and Peter want recurring revenue streams. What I talk about is a recurring relationship.”

“What I thought was a very easy change management for a department has been very challenging for a few different reasons.”

— Alexander Snyder, AI Strategy

The AI Journey

AI Is a Draft. You Are the Final Product.

Our AI journey wasn’t a straight line. We pivoted, we failed, we learned — and every draft made the final product stronger.

01

The Research Phase

Insurance Brokers — An Unlikely Start

We asked AI: how do other companies create recurring plumbing revenue? It pointed us to insurance partnerships — automatic shut-off valves, water monitoring for policy discounts. Interesting, but it didn't solve our stickiness problem.

Learned: AI surfaces ideas you'd never Google. But not every draft is the final answer.

02

The Pivot

Back to the Drawing Board

Insurance didn't get buy-in from the team. It felt like 'just one more product.' So we went back to the AI — 'This didn't work. What else?' That's when water quality surfaced as a massive, underserved opportunity.

Learned: If you have to pivot, that doesn't mean the work was wasted. The insurance research is still on the shelf — and we're now getting inbound inquiries about automatic shut-off valves.

03

The Discovery

All Water Is Dirty

We didn't realize how big water quality is as a problem. The mass public doesn't either. We had 5 years of water tests sitting in filing cabinets that nobody else has. AI helped us see the value in our own data.

Learned: Where AI really shines is when you focus on data YOU have that nobody else does.

04

CheckYourTap

Building What Would Have Cost Millions

With AI, Tom and I built a complete water quality platform — data infrastructure, health risk assessments, personalized reports by household. This would have cost millions before.

Launched: CheckYourTap.com — an educational tool powered by Valiant that tracks contaminants across Connecticut.

05

The Cultural Shift

Plumbers Who Care About Water

The hardest part wasn't the technology. It was getting plumbers comfortable talking about health, about contaminants, about things they've never discussed with customers. We role-played, we trained, we pushed through resistance.

Result: March 2026 — highest water treatment installs and service revenue in years. The weakest links started stepping up.

The Product

Know What Your Family
Is Drinking

An educational platform that helps homeowners understand their water quality — personalized by ZIP code, water source, and household members.

166+

Contaminants Tracked

Full coverage of EPA violations, state testing data, and health guidelines for city and well water

5

Household Populations

Personalized risk assessments for adults, elderly, children, infants, and even pets

CT

Statewide Coverage

Every ZIP code in Connecticut — powered by 5 years of Valiant's proprietary water testing data

Tap to play demo

Try CheckYourTap LiveEnter your ZIP code and see your water report
“How many times have we found bacteria in someone’s well, and they’re like, ‘I don’t want to do anything about it — I’ve been drinking it.’ But when you bring up the dog, or the grandkids, or the pregnant wife— that’s a different conversation.”

— Alexander Snyder

This isn’t an upsell. It’s a conversation starter — backed by data, not a sales pitch.

The Results

The Numbers Tell the Story

Customer Story

$0K+

From a single customer relationship

A Farmington customer had a water quality concern another company couldn’t solve. Nick tested her water, diagnosed the issues, and proposed treatment. That relationship grew into a new boiler, electrical upgrade, and multiple plumbing renovations.

It started with her water — and what was wrong with it.

March 2026

Record Month

Highest water treatment installs and service revenue in years. The plumbers who “wanted nothing to do with water” delivered the best numbers the team has seen.

“Hey, our water treatment installs and service in March were the highest in the years, because we’ve really focused on this.” — Tom

Cultural Shift

Data-Driven Plumbers

Our plumbing manager now talks about data warehouses and databases. The weakest links have started stepping up. The team works together more than ever — the stronger guys teach the newer guys.

“Before, there was really not much more than we knew past the iron, acidity, hardness type of thing. That’s what we always focused on.” — Tom

The Recurring Relationship

Before

A popped pipe is a one-time fix. The plumber shows up, fixes it, leaves. Maybe they come back in two years for the next emergency.

After

Water treatment requires annual filter changes, regular monitoring, and ongoing education. The intimacy of water is initially a big roadblock — but once you're in there and have that conversation, they don't go anywhere else. That's very different than 'my toilet's not working.'

The Shift

We're not redefining our identity. We're plumbers. But we're plumbers who care about your water. And that's a fundamentally different value proposition.

Honest Reflection

What Worked. What Didn’t.

What Worked

AI as Research Tool, Not Replacement

We never asked AI to replace our plumbers. We used it to explore what others were doing, find data patterns, and surface opportunities we'd never have Googled.

Building on Proprietary Data

5 years of water tests in filing cabinets. Nobody else has that intimate detail about specific ZIP codes. That's where the value proposition lives.

Making It Personal

People don't want to know about their own water. But they care deeply about their dogs, their kids, their grandparents. We learned to lead with who it affects, not what's in it.

Separate Brand for Education

CheckYourTap, not Valiant. If we ran it through Valiant, it's 'another company trying to sell me something.' CheckYourTap is about education first.

Cautiously Open Leadership

Tom pushed back when needed but stayed open. That's the best kind of leader for AI adoption — not a yes-man, not a blocker. Devil's advocate with forward motion.

What Didn’t

Expecting Quick Cultural Change

We thought plumbers who deal with water daily would embrace water treatment quickly. Wrong. Talking about someone's health is a fundamentally different conversation. Months, not weeks.

Underestimating Water Intimacy

People treat water quality as deeply personal. 'Can I check your water?' gets the same reaction as asking something private. We had to learn to massage the approach.

One-Size-Fits-All Training

What works for one plumber doesn't work for the next. The 20% early adopters got it fast. The 20% resisters needed completely different framing. The middle needed repetition and role-playing.

Asking Permission vs. Stating Action

'May I do a water test?' got a lot of no's. 'As part of your service today, I'm doing a water test' changed everything. Confidence, not permission.

Thinking AI Speed = People Speed

AI lets you build infrastructure in weeks. But the people side — training, culture, confidence — takes months. We constantly had to remind ourselves that the tech is the easy part.

In Development

Rethinking the First Phone Call

Valiant is building an AI phone attendant to handle incoming calls. Not “press 1 for service” — a warm, conversational agent trained to ask the right questions and route to the right department.

Six Companies in One

Valiant is six departments in one: HVAC, plumbing, electrical, pest control, fuel delivery, and energy audits. “I need a repair” could mean any of four departments. The agent needs to ask clarifying questions to route correctly.

Trained on Real Calls

The agent was trained by analyzing hundreds of real call transcripts — learning the actual language customers use, where routing breaks down, and what clarifying questions resolve ambiguity fastest.

How the Agent Routes Calls

Tap any route to see how the agent handles it — what triggers the route, how the agent responds, and how it was trained.

Customer Calls Valiant
Explore the Full Agent

Try the live voice agent, see personality rules & routing details

Start Simple, Grow Smart

Don't build a fully intuitive agent on day one. Replace the basic IVR with a conversational version, then add intelligence as you learn from real calls.

Transcripts Are Gold

Before building any voice agent, analyze your call transcripts. We found customers calling back six times for the same issue. You'll find broken processes no amount of AI can fix downstream.

Fix Upstream, Not Downstream

Most customer frustration comes from broken pipelines and broken policies, not bad call handling. A voice agent forces you to address the real issues first.

The Full Picture

Every Layer Builds on the Last

This isn’t a list of disconnected tools. Each system feeds the next — data becomes intelligence, intelligence becomes action, action becomes a product your customers can feel.

01

Foundation

Ingest Everything. Structure It. Trust It.

Before AI can do anything useful, you need clean data. We built a four-layer warehouse that takes raw Dialpad call data and transforms it into something GPT-4o can reason about.

Live

60

SQL migrations

Data Warehouse

Medallion architecture: raw → bronze → silver → gold. Every call flows through 4 cleaning stages before any AI touches it.

Live

800+

calls/day

Dialpad Ingestion

Three ETL pipelines run continuously — syncing calls, fetching transcripts, and backfilling gaps. Every call gets captured.

02

Intelligence

Understand What’s Actually Happening

GPT-4o analyzes every transcript and extracts 70+ structured data points — not just sentiment, but root cause, resolution status, promises made, coaching opportunities, and whether Valiant failed to execute its own process.

Live

70+

data points/call

Call Analysis Engine

Every call gets GPT-4o analysis: category, sentiment, resolution, action items, and a strict litmus test — did Valiant break its own process?

Live

30

day lookback

Customer Context

Before analyzing any call, the system pulls 30 days of history for that phone number. A first-time billing call is very different from someone calling for the sixth time.

Live

-47%

false positives

Transfer Deduplication

Dialpad splits transfers into separate records. AI clusters them by phone number within 15-minute windows, eliminating 38-47% of false-positive process breaks.

03

Action

Intelligence That Arrives Before Coffee

Data in a warehouse doesn't change behavior. These reports put the right intelligence in front of the right person at the right time — no dashboards to check, no queries to run.

Live

7 AM

daily delivery

Daily CX Report

7 AM every morning: star performers, coaching opportunities, verbatim customer quotes. The CX manager opens her email and knows exactly where to focus.

Live

$65/hr

cost model

Process Break Report

Every call where Valiant failed its own process — missed callbacks, broken promises, repeat failures. Each break quantified at $65/hr loaded cost.

Live

Monday

delivery

Weekly Management Brief

Every Monday: what's breaking across departments, who needs coaching, revenue opportunities missed, and systemic patterns.

04

Customer-Facing

Products Your Customers Can Feel

The layers above are internal. These touch customers directly — turning proprietary data into experiences that build trust and create recurring relationships.

Live

166+

contaminants

CheckYourTap.com

Enter your ZIP code, see exactly what's in your water — personalized by household members, powered by 5 years of proprietary test data nobody else has.

Building

6

departments

AI Phone Attendant

Replacing 'press 1 for service' with a warm, intelligent receptionist. Routes across 6 departments, asks clarifying questions, handles edge cases.

05

Where We’re Going

The Roadmap

This is a living system. Each layer we build reveals what the next layer needs to be.

Roadmap

Department-Specific BI

AI-generated daily briefings for each department manager — not the CX view, but operational intelligence specific to HVAC, plumbing, electrical.

Roadmap

ServiceTitan Integration

Merging phone intelligence with field service data. When a customer calls about a missed appointment, the AI already knows the technician's schedule.

Roadmap

AI Agents for Energy Companies

The process we built is repeatable. Every company is different, but the framework — transcripts, upstream fixes, then agents — scales.

For Your Company

How to Start Your AI Journey

Cookie-cutter AI won’t work. But after implementing across multiple companies, here’s the framework that does.

The question every executive asks:

“What’s the payoff and when?”

We get it — you’re trying to get less work, not more. This framework is designed for that reality.

01

Start with Your Transcripts

Before you build anything, analyze what your customers actually call about. Not what you think — what they actually say. We found customers calling back six times for the same issue. You'll find patterns and broken processes you never knew existed.

“I find broken promises are a big one, and then also gaps in their process flows. A lot of the customers are already pissed off, but it's not because of the way the CSRs are handling it — it's broken pipelines, broken policies.”

02

Fix Broken Processes Upstream

A voice agent won't fix a bad callback policy. AI identifies where your internal processes are failing — missed deliveries, broken promises, repeat callers getting different answers every time. Fix these first.

“Voice agents are great because they're going to force you to address a lot of these issues that you didn't really know that you had. Every customer I've done this with, we've changed almost every SOP they've had.”

03

Replace IVR First, Then Build Intelligence

Don't build a fully intuitive agent on day one. Replace 'press 1 for service' with a conversational AI. Then add intelligence as you learn from real call patterns.

“If you want just an IVR replacement, we can do that pretty easily. But if you want an actual intuitive, generative agent, there's a process.”

04

Invest in Your People, Not Just Your Tools

AI builds infrastructure in weeks. Cultural change takes months. Your leaders need to be cautiously open — not blockers, not yes-men. The team needs training, role-playing, and time.

“Our plumbing manager went from never thinking about data to talking about data warehouses. That didn't happen overnight. It happened because he pushed back when needed and stayed open to learning.”

05

Find the Data Nobody Else Has

Every home service company has proprietary data they're not using — call transcripts, service history, water tests, equipment records. That's where AI creates real competitive advantage.

“We had 5 years of water tests in filing cabinets. AI helped us see the value in our own data. Now we have insights about specific ZIP codes that no aggregator or competitor can match.”

Warning: Don’t Train on Bad Data

“Could you take transcripts from all your customer service calls and feed them in as a baseline?” Yes — but when you do that as a learning model base, it’s gonna learn bad behavior. It’s not gonna focus on the upstream issues.

Warning: Expect Early Frustration

“If they go live with these things and they’re frustrating — and many of them are at the beginning — they have little tolerance for frustrating their customers. They balk, and then the world is ending.”

“There are a lot of people — if they could solve the voice agent issue, if they can get one that really works, there would be a huge breakthroughfor them.”

— Industry Leader