Everplan: AI-assisted planning of field operations

Introduction

Geckosoft is a digital company and AI consultancy founded in 2016 in Pisa, with a team of over twenty developers, designers and researchers, and is ISO-certified for quality management and information security. It develops proprietary AI-driven platforms for the management of fleets, personnel, assets and activities spread across a geographical area. These include Everplan, a real-time operational planning and optimisation solution aimed at businesses and organisations that coordinate teams and vehicles across multiple sites or for multiple clients.

The Problem

Those who provide technical services across large areas still plan their work manually: a manager organises the working days by geographical proximity, and the field worker decides for themselves the order in which to carry out the tasks. This results in vehicles being hired on a daily basis but used only part of the time, unproductive journeys that nobody measures, and no response when an unforeseen event brings an activity to a standstill.

Everplan was developed to solve this problem, but its adoption faced a recurring obstacle: without a benchmark against a real-world case, the benefit remains nothing more than a sales pitch. Geckosoft lacked real-world validation, based on actual projects and using a method capable of isolating the contribution of automated planning.

The solution

Through MADE’s ‘Test Before Invest’ service, Geckosoft carried out a field trial with Phacelia SB – a company specialising in the instrumental surveying of road surfaces – on two real-world contracts awarded by road authorities: approximately 7,000 km of roads and 300 routes.

A dedicated instance of the platform has been set up, featuring a planning back office, a replanning engine and a mobile app for operators. The comparison was based on three baselines – days estimated using the manual method, days planned by the engine, and actual figures – under a protocol requiring strict adherence to the plan, without any discretionary adjustments.

The trickiest part was mapping a new domain onto the platform’s data model: routes with a fixed direction, connection costs dependent on the sequence, and indivisible resources hired on a daily basis. It did not require any customisation of the engine.

Tecnologies

1. Everplan platform: back-office system for recording work packages and activities;
2. Multi-objective optimisation and re-scheduling engine with constraints relating to timetable, capacity, resource availability and route;
3. Mobile application for the real-time transmission of instructions to field operators and the collection of progress reports. Integration of geospatial data.

Expected impacts

This validation opens up a market expansion for Geckosoft beyond its original logistics scope: routine network maintenance, infrastructure inspection and technical services across a wide geographical area share the same problem structure and can be addressed without the need for bespoke developments. The next step will be multi-resource optimisation, involving multiple teams and vehicles working on a shared portfolio, where the expected benefit exceeds the measured benefit.

• From an environmental perspective, reducing the number of days spent on business trips whilst maintaining the same level of service means fewer kilometres travelled and lower fuel consumption;

• From a social perspective, a schedule that fills all available hours includes overtime and downtime for staff on business trips.

Benefits for the business

Geckosoft now has a measured and documented use case: verifiable figures to take into the commercial phase, rather than theoretical estimates. The trial has confirmed the general applicability of the platform’s data model across a non-native domain, thereby reducing the cost of entry into adjacent sectors. This has resulted in concrete product recommendations – a route passability attribute and a data validation checklist for onboarding – as well as a pilot protocol that can be replicated with new clients, thereby shortening the time-to-value.