AI-Powered Water Test Analysis and Equipment Recommendations

A water treatment company ran into a bottleneck every operations manager knows: the gap between field data collection and actionable recommendations. Technicians could run water tests all day. But turning those readings into professional analyses that customers understood—and that led to equipment sales—required someone with deep product knowledge to interpret each result.

Now technicians enter readings into a simple form. Within seconds, AI analyzes the data against regional water quality benchmarks, identifies problems connected to consumer impacts, recommends the right equipment configuration, and generates talking points for the customer conversation. The recommendation appears in both the original record and a team Slack channel.

The technician doesn’t need to be a water chemistry expert. The AI encodes that expertise and delivers it at the point of need.

The Problem

Field service businesses that sell based on diagnostic data face a consistent challenge:

Technical interpretation creates bottlenecks — Test results require expertise to translate into recommendations. When that expertise lives in one or two people’s heads, scaling becomes impossible. Technicians either wait for guidance or guess.
Inconsistent recommendations damage credibility — Without standardized interpretation, the same water readings might generate different recommendations depending on who reviews them. Customers notice. Trust erodes.
Sales conversations lack structure — Even when technicians know what equipment to recommend, explaining why in terms customers care about (health, appliances, skin, taste) is a separate skill. Many field techs are great at testing but struggle with consultative selling.

The result: slower quote turnaround, missed upsell opportunities, and a ceiling on how many jobs can flow through limited expertise.

What This System Does

The automation monitors a simple database for new water test entries, then generates comprehensive analyses that bridge the gap between raw readings and customer conversations.

Regional Context Injection

The AI doesn’t just analyze numbers in isolation. It uses the customer’s zip code to pull relevant context about local aquifers, common contaminants, and regional water quality patterns from sources like EPA databases. This makes recommendations feel location-aware, not generic.

Pre vs. Post Filtration Comparison

Many customers already have filtration equipment. The system captures both pre-filtration (raw well water) and post-filtration readings. This enables:

Assessment of existing system effectiveness
Identification of components that need replacement or adjustment
Honest acknowledgment of what’s already working

Tiered Recommendations

Rather than one-size-fits-all suggestions, the AI generates:

Base recommendation: Minimum equipment to address all identified issues
Premium option: Full system configuration for highest water quality
Existing equipment context: What they have, what’s working, what needs attention

Sales-Ready Output

Beyond the technical recommendation, the system generates a paragraph specifically designed for customer service representatives to use during conversations. This bridges the gap between diagnosis and appointment setting.

The Data Model

The system works with a straightforward data structure that technicians can populate from the field:

Input Fields

Field	Type	Purpose
Address	Text	Customer location
Zip Code	Number	Regional water context lookup
Tester	Select	Technician who ran the test
Date Taken	Date	When test was performed
Pre Hardness (gpg)	Number	Hardness before any filtration
Pre TDS	Number	Total dissolved solids before filtration
Pre Iron (mg/L)	Number	Iron content before filtration
Pre pH Level	Number	Acidity before filtration
Pre Alkalinity	Number	Alkalinity before filtration
Post Hardness	Number	Hardness after existing filtration
Post TDS	Number	TDS after existing filtration
Post Iron	Number	Iron after existing filtration
Post pH	Number	pH after existing filtration
Post Alkalinity	Number	Alkalinity after existing filtration
Existing Components	Text	Current filtration equipment installed

Output Fields

Field	Type	Content
Recommendation	Long Text	Full AI-generated analysis and recommendation

The Prompt Architecture

The AI’s effectiveness comes from a carefully structured prompt that encodes domain expertise:

You are a water specialist who sells filtration systems.

Analyze results from a water test and suggest a filtration system.

COMPONENT OPTIONS:
- Under sink or whole house reverse osmosis (RO)
- Softener
- Iron breaker
- Carbon filter
- UV Light

TESTING PARAMETERS:
- Hardness (gpg)
- Total Dissolved Solids (TDS)
- Iron (mg/L)
- pH Level
- Alkalinity

PROBLEM IMPACT MAPPING:
- High hardness → appliance damage, scale buildup, dry skin
- High TDS → affects drinking, cooking, cleaning, bathing water
- High iron → clothes spotting, rotten egg smell
- Low/high pH → corrosion, staining, acidic water

REQUIRED OUTPUT:

1. Regional water context from EPA/EWG data for this zip code
   - Known contaminants and their health effects
   - Local aquifer characteristics

2. Problem identification connected to consumer impacts
   - What's wrong with their water
   - How it affects their daily life

3. Base system recommendation
   - All components needed to address identified issues
   - Why each component is necessary
   - Consider existing equipment (mention what they have)

4. Premium upgrade option
   - Full system for highest water quality
   - Additional benefits of premium configuration

5. CSR talking points
   - Single paragraph for appointment-setting conversation
   - Consumer-friendly language
   - Focus on benefits, not technical specs

Why This Structure Works

Component options are constrained — The AI can only recommend from your actual product lineup. No hallucinated equipment that doesn’t exist.

Problem-to-impact mapping is explicit — Rather than letting the AI make up consequences, the prompt defines exactly how each water issue affects customers. This ensures consistent, accurate messaging.

Regional context adds credibility — When the analysis mentions “aquifers in your area are known for elevated iron content,” customers feel like they’re getting expert local knowledge, not generic advice.

Output structure is specified — The AI knows exactly what sections to produce and in what order. No variation in format between analyses.

Real-World Example

Input Data

Zip Code: 34771
Address: 1234 Oak Lane, St. Cloud, FL

Pre-Filtration Readings:
- Hardness: 18 gpg
- TDS: 380 ppm
- Iron: 0.8 mg/L
- pH: 6.8
- Alkalinity: 120

Post-Filtration Readings:
- Hardness: 12 gpg
- TDS: 340 ppm
- Iron: 0.5 mg/L
- pH: 6.9
- Alkalinity: 110

Existing System: Softener (8 years old)

AI-Generated Analysis

Regional Water Context

The 34771 zip code draws from the Floridan Aquifer, one of the most productive aquifers in the world but known for elevated mineral content. EPA data indicates common concerns in this region include hardness levels 2-3x the national average, naturally occurring iron, and slightly acidic pH due to organic matter in the aquifer. EWG reports have flagged elevated levels of total trihalomethanes in treated water supplies in Osceola County.

Water Issues Identified

Your pre-filtration readings confirm typical Floridan Aquifer characteristics:

Hardness at 18 gpg is classified as “very hard” (>10.5 gpg). This causes scale buildup in pipes, water heaters, and appliances, reducing their efficiency and lifespan. You may notice spots on dishes and dry skin after showering.

Iron at 0.8 mg/L exceeds the EPA secondary standard of 0.3 mg/L. This causes orange/brown staining on fixtures, laundry discoloration, and metallic taste.

TDS at 380 ppm is moderate but contributes to overall water taste and quality concerns.

pH at 6.8 is slightly acidic, which can accelerate pipe corrosion over time.

Current System Assessment

Your existing softener is reducing hardness from 18 to 12 gpg—it’s working, but not optimally. After 8 years, the resin bed has likely degraded, reducing efficiency. The system is also not addressing your iron or pH issues.

Recommended System

To address all identified issues:

Iron Breaker (new) — Will reduce iron from 0.8 to under 0.1 mg/L, eliminating staining and metallic taste

Softener Replacement — New unit will bring hardness below 3 gpg, protecting appliances and improving skin/hair feel

Acid Neutralizer — Will raise pH to 7.0-7.5 range, protecting pipes from corrosion

Premium Configuration

For the highest quality water throughout your home:

Add Whole-House Reverse Osmosis — Reduces TDS below 50 ppm, removes 95%+ of all dissolved contaminants

Add UV Light — Provides bacterial/viral protection, especially valuable for well water systems

Add Carbon Filter — Removes chlorine, VOCs, and improves taste/odor at every tap

For Your Customer Conversation

“Based on your water test, we’re seeing some common issues for this area—your water is quite hard and has elevated iron, which is why you might be noticing spots on dishes or that your skin feels dry after showers. Your current softener is working, but after 8 years it’s not keeping up the way it should. What I’d recommend is upgrading the softener and adding an iron breaker to handle the staining issue—that’ll solve the main problems. If you want to go further, we can add a whole-house RO system that’ll give you bottled-water quality at every tap. Can we schedule a time to walk through the options at your home?”

Slack Notification

*# New Water Test #*
Address: 1234 Oak Lane, St. Cloud, FL

*Date Pulled:* 2024-12-24
*Tester:* Ryan Bailey

*## Test Results ##*
Pre Filtration
*Hardness:* 18
*Substances:* 380
*Iron:* 0.8
*PH Level:* 6.8

Post Filtration
*Hardness:* 12
*Substances:* 340
*Iron:* 0.5
*PH Level:* 6.9

Current System: Softener (8 years old)

*## Recommendations ##*
[Full AI analysis appears here]

The Technical Flow

Implementation Notes

Trigger frequency: Polling every minute ensures near-instant analysis after data entry. For lower-volume operations, 5-minute intervals reduce API calls without meaningful delay.

Duplicate prevention: The workflow checks if a recommendation already exists before processing. This prevents re-analysis when records are updated for other reasons.

Field normalization: A transformation step cleans up Airtable field names before passing to the LLM. This keeps the prompt readable and consistent.

Parallel outputs: The recommendation writes back to Airtable AND posts to Slack simultaneously. The record becomes the system of record; Slack provides team visibility.

What You Get

Metric	Before	After
Time from test to recommendation	15-30 min (waiting for expert)	Under 1 minute
Recommendation consistency	Variable by reviewer	100% consistent methodology
Sales conversion (test to quote)	~40%	~65%+
Technician utilization	Limited by bottleneck	Full capacity

Operational Benefits:

Technicians can run more tests per day without waiting for analysis
New hires immediately have “expert” recommendations available
Every customer gets the same quality of analysis

Sales Benefits:

Talking points help technicians close more appointments
Regional context builds credibility with customers
Tiered recommendations enable upselling naturally

Quality Benefits:

Consistent methodology across all analyses
Recommendations always consider existing equipment
Documentation automatically generated for every test

Going Further

This foundation enables more sophisticated applications:

Quote generation integration — Connect recommendations to your quoting system (ServiceTitan, Housecall Pro) to auto-populate line items based on AI recommendations
Historical trending — Track water quality over time at specific addresses or zip codes to identify seasonal patterns or infrastructure changes
Competitive positioning — When test results show existing equipment from competitors, generate talking points specific to that equipment’s known limitations
Customer follow-up sequences — Trigger email/SMS sequences based on test results and whether the customer converted, with content tailored to their specific water issues
Technician performance analytics — Track conversion rates by technician, identifying who’s most effective at turning recommendations into sales

The hard part isn’t the automation. It’s encoding the domain expertise—knowing which readings matter, how they affect customers, and what equipment configurations solve each problem. That’s where experience compounds.