Bestie Express

Role: Head of Operations & Product · 2022 — Present · Philippines

The Problem

One in three deliveries at Bestie Express was failing SLA. Not because the couriers were not working. Because there was no system.

The company had more than 50 couriers on the road handling a mix of parcels and documents, a growing base of merchant clients, and operations that had not kept pace with the volume. Dispatch was done manually. There was no real-time view of where couriers were or which deliveries were running behind. When something went wrong, the first signal was a merchant or customer calling to complain. There were no self-service tools for merchants booking shipments, no tracking notifications for end customers, and no way for managers to see the state of the fleet without making calls.

SLA compliance was sitting at 68%. In a market where companies like Ninja Van and FAST Logistics were already running on automated infrastructure and real-time tracking, that number put Bestie at a serious competitive disadvantage. The gap between the tech-enabled players and everyone else in Philippine last-mile delivery was widening fast, and Bestie was on the wrong side of it.

The couriers were not the problem. The absence of any operational system behind them was.

The Solution

What Bestie Express needed was not a patch on top of the existing process. It needed a full rebuild, both the operations model and the technology layer at the same time. Those two things had to be designed together. Building software on top of a broken process just automates the chaos.

I designed and led the build of four interconnected products that covered every stakeholder in the delivery chain.

The Merchant app gave business clients a self-service interface for booking shipments, tracking orders in real time, and getting status updates without calling the ops team. The User app gave end customers visibility into their deliveries and automated notifications at key stages. The Driver app gave couriers their assignment queue, route context, and a proof-of-delivery flow that replaced paper and WhatsApp photos. The Admin web app became the operations nerve center: a live fleet view, SLA health dashboard, exception queue, and the primary tool for managers and supervisors running the day.

On top of the apps, I designed and implemented a multi-component AI layer that runs across the platform.

Merchant support AI. The merchant app includes a conversational AI agent connected directly to the platform's order management system. It handles merchant inquiries 24/7: shipment status, delivery exceptions, pickup scheduling, SLA queries, and billing questions — all answered in real time by pulling live data from the platform. Queries that require human judgment are escalated to the ops team with full context already attached. In logistics, "Where is my order?" type inquiries make up a large portion of support volume. Having the AI resolve those around the clock means the ops team handles only the cases that actually need a person.

Predictive SLA risk scoring. Every active delivery in the system carries a continuously updated risk score. The model factors in courier movement patterns, current location relative to the expected route, historical on-time rates for that courier, and the time window remaining before SLA breach. Deliveries crossing a risk threshold surface automatically in the admin web app and trigger supervisor notifications. Managers act on the highest-risk cases before they become failures.

Anomaly detection. The system monitors for operational integrity issues that manual oversight would miss at scale: couriers who have stopped moving for longer than expected, significant route deviations from assigned paths, and repeated proof-of-delivery submissions from the same GPS coordinates — a pattern that can indicate falsified completions. These surface as flags in the admin dashboard for review.

Courier performance intelligence. Each courier in the fleet has an AI-generated performance profile updated on a rolling basis: on-time delivery rate, route familiarity by area, and capacity utilization. Dispatch uses this data to match the right courier to each assignment rather than defaulting to whoever is available. Over time this created a feedback loop where consistently high-performing couriers received assignments that matched their strengths.

The Execution

Taking on both the operations redesign and the product build at the same time meant I had to be disciplined about where to start.

The first thing I did was map the existing process end to end, not by reading documentation but by sitting with the ops team and tracing how work actually moved. That exercise surfaced where the manual handoffs were, where data was being dropped, and where the real failure points in the delivery chain lived. That mapping became the foundation for everything that came after.

Before writing any product specs, I defined the SLA framework: delivery tiers, thresholds for each tier, and the escalation steps when a threshold was breached. The entire monitoring and AI layer depended on having those definitions locked down first. Without clear thresholds, there is nothing for the system to measure against.

From there I moved into product design and then managed the outsourced development team through the build. I was the single point of accountability between the business and the engineering team, translating operational requirements into specs, making scope decisions, and keeping delivery on track through sprint cycles.

The rollout was sequenced deliberately. The driver app went first, because the couriers were the primary source of real-time data that every other part of the system depended on. Once the driver app was live and feeding accurate movement data, the admin web app followed. Merchant and user apps came after, once the core ops layer was stable.

The AI alert logic was validated against real delivery data before going live. I worked with the ops team to test the rule-based thresholds and tune the predictive model against historical patterns, making sure the system was surfacing genuine risks rather than creating noise.

The Result

SLA compliance went from 68% to 98%.

That 30-point jump is the headline, but what it represents is more specific than a percentage. It means the operations team went from finding out about delivery failures after the fact to catching them before they happen. It means merchants stopped calling the ops team for updates and started using a self-service app instead. It means couriers have a proper assignment and proof-of-delivery workflow instead of piecing things together through WhatsApp. It means managers have a live view of the fleet rather than a lagging picture assembled from manual check-ins.

Every part of the operation, from the moment a merchant books a shipment to the moment a courier captures a delivery confirmation, now runs through a purpose-built system. The spreadsheets and group chats are gone. Bestie Express went from running one of the most manual operations in its segment to having the kind of real-time visibility and AI-assisted monitoring that most companies its size in the Philippines market do not have.