Redback WMS
Artificial Intelligence is all the rage these days. If you go out and read a few articles about AI and field service management, you wouldn’t be blamed for thinking that Skynet can be tasked to run your whole field service operation.
That’s what typically happens with new technology: thought leaders are already light years ahead of everyone in the industry and forget that the bulk of the companies are still living in the present, with today’s technologies.
It’s undeniable that AI is an important part of many new technologies and will be a major factor in field service management in the coming years. But for those organisations that are still working on today’s field management systems, how can you start taking your first steps into the AI world?
The answer to that, in my opinion, is with baby steps. Start with small, realistic objectives. Make these work for you, and then grow from there.
Maybe not chatbots
Chatbots are touted for being the solution to streamline your interactions with your customers and your field force. They’ll be able to handle multiple parallel conversations with customers, automatically reporting issues or providing statuses on jobs.
They’ll also be able to quickly understand what your field technicians are asking for and fill in the knowledge gaps by scraping through the entire history of work and knowledge documents.
But stop for a minute and think about how you personally feel when you are trying to get help from a website, and you realise you are dealing with a chatbot.
The trick with chatbots is to ensure the handover to a human operator is correctly timed to happen before the user feels frustration but not too early that it would overwhelm the human operator behind it. Too late, and the risk is a reduction in customer satisfaction.
Chatbots are an awesome opportunity, but they are tricky to get right and typically depend on a large volume of historical chats to learn from. So maybe not the best first step to take.
Work Planning
Planning the work and the expected resource usages on a job is probably one of the most important steps to help increase the first-time fix rate, increase the utilisation, reduce travel time and thus, increase customer satisfaction. Not planning, or producing a bad plan means your schedulers are just fumbling in the dark as to who should be dispatched.
One of the things AI (or machine learning in this case) is good at is finding patterns in large amounts of data. If you think of the data you capture closing out work (labour and material actuals, time spent, outcomes, etc.) as that body of data, an AI engine can make easy work of extracting patterns of job plans for different conditions. In effect, it can learn from what you’ve done in the past to help you better plan your future work, thus making good on your scheduling related metrics (FTFR, Utilisation, Travel, etc.)
And the easy part for an organisation that is just starting out using AI is that implementing this can be done easily. The AI engine doesn’t even need to be built into the field service management systems. It can be a program on the side, that runs nightly and simply outputs recommendations of how workplans can be refined.
Work Review
The above use of AI implies that you have good quality work data to review and learn from. Coincidently, this is also another easy first step for the use AI: helping review the quality of the close-out data.
Some of the first tutorials you encounter when you start looking into AI is image recognition or being able to label what is represented in a picture. So, what about using image recognition to evaluate the quality of the photos captured in the field. This, of course, helps you reduce the number of people reviewing field pictures, but more importantly helps you focus on the lower quality pictures and in turn helping those technicians responsible for lower quality images increase their performance.
But it doesn’t have to be for pictures only. AI engine can review all types of outcome data and rate its quality. For example, start and finish events being punched in the mobile app 10 kilometres away from the work site could attract a lower quality rating and get flagged for further review. Or, as another example, unnecessary material consistently being used by a specific agent would be recognised and flagged for review.
Here again, the implementation can be as simple as an offline nightly run on the day’s data resulting in a quality rating to each job.
Higher quality data
Another undeniable advantage of starting with these baby steps is that it will naturally increase the quality of your body of data, and high-quality data means more opportunity to learn from for AI systems.
You’ll then have something to feed into your chatbots, when you get to that point, to ensure they can actually be helpful for customers and field agents.