Exploring the value of AI-Powered Insights in DirtMate

Propeller's platform allows teams to measure and manage their sites using 3D mapping and visual tools.

One of these tools is DirtMate, a real-time progress tracking system that gives teams live visibility of material movement, helping them stay on schedule and reduce rework.

Image of CSV file

The Problem

DirtMate collects large amounts of valuable operational data, but users often struggle to turn that raw data into meaningful, actionable insights. Many spend significant time manually digging through reports or datasets to identify opportunities for improving efficiency.

The Goal

Explore how AI could make DirtMate data more accessible and useful by:

  • Surfacing relevant, personalised insights proactively.

  • Helping users identify improvements faster.

  • Delivering insights through an intuitive, conversational experience

Image from workshop

My Role: Product Designer

Over 8 weeks, and working with product managers, engineers, and sales I:

  • Led discovery, including customer and stakeholder interviews

  • Planned and designed an experimental approach to validate AI-powered insights

  • Created wireframes and prototypes to test design concepts

  • Aligned the design with technical feasibility

  • Defined success criteria and risks for each experiment

Research

Customer Interviews

Customer interviews with site managers, supervisors and operators revealed:

  • Users want quick, easy ways to spot inefficiencies.

  • Many aren’t confident interpreting raw data.

  • They lack quick ways to investigate issues

  • Operator performance (e.g. idle time, productivity) is a key focus.

  • There is strong interest in personalised, contextual insights.

Site visit image…

Image of Miro board with stakeholder insights …

Stakeholder interviews and workshops

Internal stakeholder interviews with sales, product and customer support revealed:

  • DirtMate provides valuable data, but extracting insights is time-consuming for customers

  • There’s an opportunity to surface key observations proactively and improve the product experience

  • Caution around ensuring AI insights have the right context and accuracy

  • AI hallucinations

Design Process

AI Suggestion Bot Concept

We designed a lightweight, AI-powered bot within DirtMate to proactively analyse site data and provide clear, actionable insights. Users can engage with the bot to ask questions and explore their data further, all within a simple, conversational interface.

To validate the concept and de-risk development, we planned three lightweight experiments with 60+ active DirtMate users, delivered using LaunchDarkly and Claude AI.

Experiment 1: Initial Insight Card

  • Test simple, contextual insights within the DirtMate interface

  • Measure discoverability and engagement (hover rates, thumbs up/down)

Image to show iterations of the design

Image to show iterations of the design

Experiment 2: AI-Generated Insights Carousel

  • Deliver multiple AI-generated insights focused on efficiency

  • Collect qualitative feedback and measure interaction with the carousel

Experiment 3: Single Query Prompt

  • Introduce a conversational prompt to encourage deeper data exploration

  • Understand what questions users naturally want to ask

  • Measure engagement and types of follow-up questions

Image to show iterations of the design

Impact and Reflection

Outcomes

This project is currently in the build phase. Once deployed, we will:

  • Track adoption and engagement across experiments

  • Collect qualitative and quantitative feedback

  •  Identify which types of insights are most valuable to different user personas

  •  Refine the AI models and conversational experience based on real user behaviour

If the experiments are successful, it will:

  • Make DirtMate data more accessible and actionable

  • Boost user engagement and platform value

  • Help users improve efficiency on site

Image to show iterations of the design

Key skills Demonstrated

  • Discovery and user research

  • Collaborative, cross-functional design process

  • AI-informed product design

  • Prototyping and rapid experimentation

  • Stakeholder alignment and communication

Explore Other Projects

Revolutionising the Digital Claims Journey

The UK’s first fully integrated digital motor claims form redefined the claims process for Direct Line Group brands, delivering a seamless, empathetic, and efficient experience that elevated customer satisfaction and streamlined operations.

Driving Digital Adoption for Claims

Through strategic updates to the claims landing page, this project increased the number of customers initiating claims online -improving accessibility, boosting engagement, and driving digital adoption across DLG brands.