Field Operations Platform for Multi-Project Staffing & Coordination

The brief

A field operations company runs back-to-back multi-week projects where hundreds of people are deployed across multiple locations at the same time. A small admin team was trying to coordinate all of it — staffing, scheduling, travel, ground transport, expenses, attendance, payments — across spreadsheets, email threads, WhatsApp groups, and a fragile no-code app. They also had to send daily reports out to the program sponsor.

Information lived in five places at once and arrived a day late. They asked us to build a single field operations platform that could replace the entire stack — and to have it live before the next project cycle started.

What we built

A single platform built on three pillars:

• A multi-project, multi-location backbone. One project record per engagement, with locations, departments, and people scoped per project. New cycles spin up by creating a new project; existing people reuse their accounts. Multiple projects run cleanly side-by-side without bleeding into each other.
• One app for the people on the ground. A self-service portal covering everything they touch: their schedule, daily assignments, travel info, attendance, expenses, documents, daily notes, and live earnings. They open one app instead of digging through ten emails.
• A control room for the admins. A single console for staffing, approvals, change requests, expense review, document compliance, real-time travel and attendance status, broadcast communication, and reporting. Role-scoped — a project admin only sees their projects; a super admin gets the portfolio view across every project running.

What this looks like in practice

• End-to-end travel chain. Each person is tracked from booked flight → live flight status → ground pickup → hotel arrival → app login link delivered. Admins watch every traveller's stage on one live board, instead of pinging people individually to find out who's where.
• Daily carpool and ride coordination. Every morning, riders open the app and see who they're carpooling with and who their driver is — with the driver's name and number one tap away. Drivers see their full pickup list with each rider's contact. Nobody is left wondering who's picking them up or who they're picking up.
• In-app sign-in and sign-out at the site. People sign in from the app the moment they reach their location each day and sign out at the end of their shift. Admins know in real time who has actually arrived, who hasn't, and who's still on shift — instead of phoning around to find out. Attendance, hours, and presence all flow from one source of truth.
• Self-service expenses and live earnings. People upload expenses with a photo, see them flow through approval, and watch their earnings update in real time. Admins review, override caps where needed, and every change is audit-logged.
• Daily reports out to the program sponsor. Attendance, expense, and operational reports are generated and sent automatically every day — instead of an admin spending hours pulling numbers into a spreadsheet. Issues surface the same day they happen, not a week later.
• Broadcast communication built in. Super admins send a templated email or notification to every person on a project — or across all projects — straight from the app, with delivery tracking per recipient. No external mail-merge tools.
• Approval and change workflows. Shift swaps, replacements, taper-outs, expense overrides, document approvals — proper queues, every action audit-trailed, no Slack threads or "approved by reply" email chains.

Where AI does the first pass

We used AI where it saved real admin time, without taking humans out of the loop.

• Expense receipt analysis. Every uploaded receipt is read by AI — vendor, amount, category, date, and whether it falls inside the project's per-diem caps. Admins see an AI-prepared summary and approve, override, or reject in one tap. Manual data entry effectively disappears.
• Resume and basic onboarding review. New joiners upload routine onboarding items like their resume through the app. AI does a first pass on those — extracting key fields, flagging anything missing or inconsistent — so admins move through onboarding much faster than reading every CV cold. Anything sensitive still goes straight to a human; AI is only on the routine stuff, by design.
• Daily field notes summarised by AI. Each consultant captures notes from their day — observations, issues, anything worth flagging. Across ~300 people writing daily over a four-week project, that's thousands of notes per cycle. AI rolls them into a daily summary, surfaces anything important immediately, and produces a full project overview on demand. What used to take admins hours of reading and synthesising now takes minutes.

The client's own AI co-pilot — direct from Claude

The change the client team felt the most: we connected the platform's Supabase database directly to the client admins' Claude account using the Model Context Protocol (MCP). The admin team can now open Claude and ask their own questions of the live data, in plain English — "How many people are short on hours this week across all projects?", "Generate me an expense report grouped by category for the last cycle", "Which travellers haven't signed in yet today?" — and get an answer in seconds.

No developer in the loop. No ad-hoc reports to request. No back-and-forth on Slack to find out what the data already knows.

Custom reports that used to take hours of cross-checking now take a single sentence. The admin team is genuinely faster, more independent, and more confident — they describe it as the single biggest productivity change the platform brought.

They ended up giving the assistant a name — Clay — and they only half-joke when they say Clay is the most productive member of the admin team.

What changed for the admin team

From juggling five tools and chasing people, to one system with proper roles, approvals, audit trails, real-time visibility, and a live AI co-pilot they can query directly. A small team now does what previously needed many full-time admins — and the program sponsor gets cleaner, faster reporting than they ever had before.

How we built it

Production stack from day one — this was never going to be a prototype.

• Next.js + Supabase with TypeScript end-to-end and an RLS-first multi-tenant data model
• FlightAPI.io for live flight status and arrival projections
• TravelPerk integration to ingest corporate travel bookings automatically and reconcile against the schedule
• Whapi.cloud for WhatsApp automation — welcome messages on landing, hotel and transport directions
• Anthropic Claude powering the in-app assistant, expense receipt analysis, routine document review, and the AI summarisation of daily field notes
• Model Context Protocol (MCP) server exposing the live Supabase data, securely, to the client admins' own Claude account for natural-language reporting and ad-hoc queries
• PWA-enabled so admins and on-site coordinators use it from any device

The result

• Six weeks from kick-off to live operation. The platform was running for the very next project cycle.
• One field operations platform replaces five+ disconnected tools — spreadsheets, email threads, manual WhatsApp groups, and several separate booking and reporting workflows.
• Multiple parallel projects now run cleanly side-by-side. New cycles spin up by creating a project record; people and processes carry over.
• AI handles the first pass on expenses, routine onboarding documents, and the thousands of daily field notes that admins used to read manually.
• The admin team queries the live platform directly from Claude via MCP — custom reports in seconds, not hours.
• Daily reporting to the program sponsor that used to take hours of manual work now happens automatically.
• In-app sign-in / sign-out gives admins live presence across every site, every day.
• Admins are back to running the program instead of running the spreadsheets.