The brief: modernise an industry that's still on paper
AgroHQ's founder isn't a developer. He's an operator who grew up around farms and knew every fragmented tool the industry defaults to — WhatsApp for staff, ledgers for livestock, spreadsheets for buyers, weather apps for planning, and everyone's memory holding the rest together. He'd sat with two agencies who quoted 6–9 months for a discovery phase alone. He came to us with one question: can you get me a real pilot in production in a month?
"I didn't want a beautiful Figma file. I wanted farms actually using it, actually paying. Growvate went to the farms with me. That was the difference." — Founder, AgroHQ
Week 1: Discovery — 12 farm visits, one whiteboard
Before we opened a single tool, we spent five days on farms. Twelve visits across four regions. Watched morning stand-ups, sat in on buyer calls, followed workers with their phones. We surfaced three insights that reshaped the product:
- The phone is the desktop. Every operator we watched worked from their phone first, laptop second — or never. Mobile-first wasn't a design choice, it was the actual product.
- Workers speak in WhatsApp, not apps. Field workers already logged everything into WhatsApp groups. Any product that made them switch would fail. So we didn't — we built a WhatsApp bridge that turns messages into structured logs.
- Buyers are the daily crisis. Every operator we shadowed spent 30–60 minutes a day fielding buyer calls. Price coordination — not livestock tracking — was the pain we could relieve first.
We ran a card-sort exercise with 6 farm managers to build the information architecture. It killed our first draft of the top-level nav (six sections) and gave us the one that shipped — Farm, People, Buyers, AI Advisor.
Week 2: Design system + AI architecture
Two parallel sprints. The design team built a farm-friendly UI system — high-contrast, glove-friendly touch targets (48px minimum), large numerals for glance-reading in the sun, semantic colour for livestock status. Packaged as a Figma library and mirrored as a Flutter component library so mobile shipped from the same source of truth as the web dashboard.
The AI team built the farm advisor on OpenAI's Assistants v2 API — with a farm-context memory layer that keeps each operator's context (herd size, crop calendar, region-specific pest patterns) persistent across sessions. It doesn't hallucinate the farm's numbers; it looks them up.
Week 3: Build — the field ops dashboard + mobile app
The dashboard is the operator's cockpit — livestock health, worker check-ins, weather signals, buyer inbox, AI advisor thread — all one screen, all keyboard-navigable. The mobile app is the field partner — clock-in, photo logs, voice-note advisor queries, and offline-first sync (we lost data twice in early pilots when signal dropped — sync-first became a hard rule).
Laravel powers the backend — battle-tested, deployable anywhere, and reads/writes cleanly to MySQL with Redis for the AI advisor's memory layer. Boring stack. Deliberately.
Week 4: Pilot — 12 farms live
We onboarded 12 farms in the first week of pilot, in-person for the first four (so we could watch onboarding fail in real time) and remote for the rest. By day 21 of the pilot, 84% of operators were logging in weekly. By day 30, we had our first paying farm.
The metric we obsessed over — tools consolidated per farm — landed at 6 → 1 across the pilot cohort. Operators who ran their farm on 6 disconnected tools now ran it on AgroHQ.
What we'd do differently
Two things. First, we'd have shipped offline-first sync from day one instead of retrofitting it week three — farms have signal dead zones we hadn't sat in yet. Second, we'd have built the WhatsApp bridge before the mobile app, not alongside it — it's the interface workers already trust, and the mobile app is a slower behaviour change than we assumed.
Both are now standard parts of our field-work playbook.