FieldFlo
AI HVAC Operations Platform
An AI-assisted operations platform for HVAC service businesses — scheduling, dispatch, and field work in one place.
- Category
- SaaS
- Type
- Product
- Status
- In development
- Website
- fieldflo.vercel.app
01 · Overview
The problem and the solution
Business problem
HVAC service companies juggle scheduling, dispatch, technician notes, and customer follow-up across disconnected tools. Context gets lost between the office and the field, and coordination eats into time that should go to customers.
Solution
FieldFlo brings service operations into a single platform and applies AI to reduce manual coordination — helping office teams plan the day and giving technicians the context they need on site.
02 · Functional
What the product does
Key features
Scheduling & dispatch
Plan jobs and assign technicians from one shared board.
Technician job context
Equipment, history, and notes available in the field.
AI-assisted job summaries
Turn technician notes into clear summaries for customers and the office.
Customer service history
A single timeline of visits and work for each customer.
User workflows
- Office staff capture a service request and schedule the visit on a shared board.
- Dispatch assigns a technician based on availability and skills; the technician sees the job, site, and equipment history on their device.
- On site, the technician records notes, readings, and photos.
- AI turns those notes into a clear job summary for the customer and the office.
- The visit is added to the customer's service history for the next call.
03 · Technical
How it's built
Technology stack
- Next.js
- TypeScript
- Node.js
- PostgreSQL
- OpenAI
- Docker
- AWS
- CICD
- Generative AI
- NestJS
- RAG
- Redis
- React
- AWS S3
- devOps
System architecture
FieldFlo is built as a modular monolith: a Next.js web app for the office, a mobile-friendly technician view, and a TypeScript API layer over PostgreSQL. Scheduling, dispatch, customers, and equipment live in separate modules with clear boundaries, so each can evolve without a rewrite.
AI features run as background jobs. Technician notes are sent to an LLM together with the job and equipment context, and the generated summary is stored as a draft for review rather than sent automatically.
Engineering highlights
- Domain model built around jobs, visits, equipment, and customers, so history follows the equipment across visits.
- AI summaries generated asynchronously with structured prompts and stored as editable drafts.
- Role-based access separating office, dispatch, and technician capabilities.
- Responsive technician UI designed for quick use in the field.
Challenges and design decisions
- Messy field input: technician notes are short and inconsistent, so summaries are grounded in structured job data (equipment, site, work type) instead of free text alone.
- Human in the loop: AI output is always a draft the office can edit before it reaches a customer.
- Scope discipline: a modular monolith was chosen over microservices to keep deployment and data consistency simple at this stage.
04 · Infrastructure
Hosting, delivery, and operations
- The Next.js app and API run as containers on AWS, with PostgreSQL on a managed database service with automated backups.
- Separate development, staging, and production environments, configured through environment variables and a secrets manager.
- A CI/CD pipeline runs type checks, tests, and database migrations before every deploy.
- AI summaries run as queued background jobs with retries, so slow model calls never block the technician UI.
- Centralized logging, error tracking, and uptime monitoring; AI calls are logged with latency and cost.
Interested in FieldFlo?
Visit the product website, or get in touch for a walkthrough or a technical discussion.