AI adoption case study: Transforming job allocation and logistics at Thames Laboratories
techUK’s AI adoption collection of case studies showcases examples of how organisations are Seizing the AI Opportunity, either through the adoption of AI models within their organisations, or by developing AI tools that can be leveraged by others.
By shining a light on these use cases, techUK hopes to demonstrate examples of best practice from across sectors and from organisations of all sizes.
1. Challenge:
The company’s legacy job scheduling system was mostly manual, a resource-consuming operation that routinely resulted in sub-optimal customer service, inefficiently utilised, dissatisfied workforce, and decision making that was slow to adapt to real-time developments; all at a high environmental cost. The cumbersome logistics of dispatching engineers to attend calls at client locations resulted in significant operational losses.
The core challenges associated to improving job scheduling in Thames Labs:
Meeting conflicting objectives, such as increasing productivity whilst keeping staff workloads and carbon footprints within manageable limits.
Managing subtle, dynamic inter-dependencies between job priorities, preconditions, and engineering staff skillset.
Adapting to unforeseen changes in staff availability (engineers stuck in traffic) and to last-minute job booking modifications (postponements, new requirements added to the initial ask) with cascading implications on other scheduled jobs.
2. Solution:
The team of experts from Aston University and Thames Labs developed an innovative, explainable AI solution that automated the company’s onsite engineering service provision. This resulted in more evenly distributed workloads, better customer care and a lower carbon footprint. The AI tool leverages the strengths of evolutionary intelligence to analyse the company’s historical job allocation records and efficiently detect patterns, both value-adding and detrimental to the business. This enables scheduling incoming jobs in a way that perpetuates the former whilst avoiding the latter, whilst dynamically adjusting to realtime data (last-minute job alterations, unforeseen changes in engineers’ circumstances, traffic developments).
The intelligent job scheduling algorithm was successful at:
Assigning jobs to (teams of) engineers appropriately skilled to complete them, such that the itinerary of an engineer on a given day is as short as possible (i.e., the jobs assigned to the engineer are in a tight geographical cluster and close to the engineer’s base location) and the time spent by an engineer attending jobs is evenly distributed across days (avoiding allocations where jobs are amassed towards the end of the week).
Creating job schedules that met operational constraints: even workload distribution across all engineers on the payroll, meeting income targets, attending jobs in logical order (avoiding cases where the kitchen cabinet fitting team arrives on-site before the plumbing and electrical wiring have been completed), etc.
Reducing time spent on the road by company vehicles, thus decreasing the organisation’s carbon footprint and supporting the net zero agenda.
3. Barriers:
Conflicting, ambiguous feedback from company engineering and admin staff trialling the algorithm whilst performing their daily tasks.
Incongruous legacy database entries: job and engineer metadata, such as geospatial location, priority and completion status, were often incomplete or misaligned.
Incompatibilities with frontend requirements: the AI algorithm outputs needed careful calibration to match the way job information was presented to company staff in the user interface.
Schedules with limited explainability: non-technical stakeholders found it difficult to grasp the rationale behind the AI-generated job allocations, causing increased resistance to adopting the automated system within their daily work routines.
4. Impact:
Thames Labs has successfully adopted the innovative AI-powered job scheduling system to replace their legacy logistics but not their staff whose bandwidth is now freed of mechanical tasks and available for creative, fulfilling work leading to a sense of accomplishment and professional satisfaction as well as superior service for clients. The company is also making the intelligent job allocation tool available to the community in the form of SaaS.
The AI solution was benchmarked against the company’s legacy manual scheduling system in three practical test scenarios, each entailing intricate job-to-engineer-skills matching combined with travel and service time optimisation, all while reducing the company’s carbon footprint and increasing the quality of its services (and, implicitly, its profits). The results show that the performance of the AI algorithm is on par with that of the manual variant on one dataset and significantly surpasses it on the other two. Across the first and third datasets, an engineer’s daily tour in the AI-generated schedule is less than half as long as in the manual variant, daily CO2 emissions are cut by almost 40%, the drop in fuel consumption exceeds 35% and operation costs are reduced by 37%.
An admin team leader within Thames Labs has said:
It has been good to consider ways to maximise the use of surveyors’ time. The new system has made us more aware of how we are booking the appointments, allocating them to a specific surveyor, and deciding how long they should take.
Charlotte Burton, Thames Labs Development Manager has said:
