FieldFlow

FieldFlow is field service software for small trades businesses of one to five people. It handles the whole job cycle (estimates, scheduling, invoicing, payments) and includes an AI assistant, for a flat $19 a month with the whole team included. I founded it, and I'm the only designer and the only engineer. Every screen, flow, and line of copy is mine, and so are the pricing and positioning the design has to support.

The big platforms compete on breadth. They keep adding modules and add-ons, charge per seat and per feature, and end up needing a sales call to buy and an onboarding program to learn. A small trades business pays for a lot it never opens and works around screens that were designed to look good in a demo.

The idea came out of my own home renovations, where I had informal conversations with more than a dozen tradespeople. Many of them volunteered that they aren't technical. FieldFlow hasn't had formal user testing yet, so for now the evidence is those conversations, teardowns of competitors, and my own use.

So FieldFlow goes the other way. It covers what a service business does every day (estimates, scheduling, invoicing, payments, chasing late payments) and puts all the design effort into making those quick and modern, whether the person is an office manager at a desk or a solo operator invoicing from the truck between jobs.

The pricing says the same thing: one flat rate, every feature, unlimited seats. At $19 a month it shouldn't take a meeting to decide.

Field service software usually grows by addition: every release asks what else the product could do. Most of FieldFlow's feature list is a record of what I cut. The test was whether a feature belongs to the loop a service business runs to get paid: quote the work, schedule it, do it, invoice it, collect, and follow up with whoever hasn't paid. If it was on that loop, it had to be excellent. If it wasn't, it didn't ship.

What shipped:

• Estimates. Built in seconds from a reusable price book, then converted to an invoice or a scheduled job in one click. Customers sign on their phone, and every signed document is snapshotted and hashed for tamper evidence.
• Invoicing. A live PDF preview that matches exactly what the customer receives, with selectable styles, custom numbering, and live totals. Credit notes, refunds, and voids all leave an audit trail, and QuickBooks stays in sync.
• Scheduling. Month, week, and day views with drag-to-create and drag-to-reschedule, a full recurrence builder for repeating work, and one-tap timers for time on the job.
• Payments. Customers pay the way they actually pay: card through Stripe, a QR code for Venmo, Zelle, PayPal, or Apple Cash, or a check with mail-to details. Delivery status is tracked per channel, so a bounced email never shows up as sent.
• Reminders. Separate cadences for unpaid invoices, upcoming appointments, and open estimates, each with editable message templates and a log of everything sent.
• Expenses & reporting. The one exception I made. Expenses with receipt photos and categories, a mileage log, and reports that total income, expenses, and net profit for any period, with CSV export. Most small trades businesses use a small slice of QuickBooks, so covering that slice well lets them cancel a subscription instead of adding one.
• AI. Flow, the built-in assistant, can use ninety tools. It drafts, leaves anything consequential for a person to send, and flags what needs attention, like a finished job with no invoice.

An assistant can save a lot of time, but plenty of people in the trades don't trust AI. So Flow is optional. One toggle removes it from the interface completely, and every write tool it has matches an action in the regular UI, so nothing depends on using it.

For people who do use it, it's often faster than the UI. "Invoice the Hendersons for the water heater job and send it" is one sentence. Without Flow it's a navigation, a form, and a send dialog. On the web Flow knows what's on screen, so "send it" means the invoice you're looking at. Its tools create invoices, start timers, find open slots, and record payments, and results come back as cards that link to the record. Flow drafts; a person clicks send on anything that matters.

A triggers engine watches for what a good office manager would catch, like a finished job with no invoice or an accepted estimate that never got scheduled, and puts the top one in front of you as a one-click prompt. Requests start on a fast, cheap model and only move to a larger one when Flow tries to write something. That's how an assistant like this can exist at $19 a month.

Before designing FieldFlow's onboarding, I went through Jobber's step by step. It's eight steps, including some low-priority ones and some that serve the vendor more than the user. It promises personalization, but everyone lands in the same place: a to-do list with no clear call to action and none of your own data in the app.

What I took from it: empty software looks like work. Onboarding should hand back something that's already theirs.

FieldFlow's onboarding is three steps: your business, your trade, and your first client. A live preview beside the form fills in as you go, so setup feels like building your account rather than filling out paperwork, and picking a trade stocks the price book with starter services. Asking less means knowing less about who just signed up. I took that trade: the live preview and one real client over a fuller profile.

It took three versions. The first had the same three steps and the live preview, but it put the weight on configuring the price book and landed on an empty dashboard with a checklist, the same landing I'd flagged in Jobber. The second reworked the steps around creating a client, so people arrive with that client already in the app, and a prominent card hands the rest of setup to Flow. It still had no progressive nudges. The third added them.

The dashboard fills in as the account does. A new account sees a large card asking for a first client, and the open work, schedule, and AI overview stay hidden until there's something to put in them. Once a job, estimate, or invoice exists, the card goes away and an earnings chart and the AI overview take its place. Data added anywhere in the app counts, including through Flow.

The wizard asks only what it needs, and Flow picks up the rest in chat, one question at a time. It starts from what onboarding already collected, so nothing gets asked twice. It saves each answer as it's given and lets optional details be skipped. Nudges bring people there. They're triggered by fixed rules rather than the model, most of them point toward the chat, and the action is built into the toast. Three of them arrive on visits, in order and at most one per visit: finish your account, draft your first estimate or invoice, create your first job. Every nudge costs attention, so each one fires only once.

Customers size up a trades business quickly, on their phone. The estimate opens and they decide whether the outfit looks professional, whether to say yes, and when to pay. So every estimate, invoice, and job in FieldFlow has a branded public page behind a tokenized link. The customer doesn't need an account or an app.

Saying yes takes about ten seconds: enter an email, let the 6-digit code auto-fill from SMS, then sign by drawing or typing. The signed document is saved as a PDF snapshot and SHA-256 hashed into the operator's audit trail, so there's tamper evidence if a signature is ever disputed. Invoices take Stripe, a QR code for Venmo, Zelle, PayPal, or Apple Cash, or a check. After payment, a "How did we do?" page sends happy customers to the business's review sites, which is how a finished job turns into the next one.

The constraint here was unusual: the mobile app, the web app, and the PDFs the app generates all had to look and behave like one product. One shared color package feeds all three. Space Grotesk is the typeface everywhere, and 26 web primitives and a full mobile kit are built on the same tokens.

The sharing goes past visuals. 225 web files import one package for totals math, validation, permissions, and formatting, so an invoice can't add up differently on two screens. The system also changes by role. Operators get the full app, while crew and subcontractors get a three-tab mobile app, controlled by a permission matrix that mirrors the database security rules.

I set the style in Figma with a handful of screens that fixed the type, color, and density, then built the product in Claude Code. The mobile apps are Expo, the web app is Next.js, and both run on Supabase, with Stripe for payments, Twilio for customer messages, and an Astro marketing site in front.

Flow was the hard part. Each of its write tools mirrors a real action, so every response has to be the right shape before anything goes out, and a model doesn't return the same shape twice. QA was running my own jobs through it, logging every response that came back wrong, and fixing the tool definition or the prompt until the log stayed empty.