Do you want to improve the quality of your service while using less resources? Recent research* on the value of machine learning for public services concludes that implementing virtual assistants allows you to kill those two birds with one stone. In this article you’ll find some practical tips that will foster the implementation of machine learning in your organization.

1.  Find out what machine learning can offer your organization

Machine learning is concerned with discovering patterns in large (unstructured) data sets, in which machines are way better than humans. In our research machine learning was applied in virtual assistants answering questions from citizens and internal departments. There are plenty other applications of machine learning. For instance, reading through large documents to support decisions if someone’s unemployment benefits should be continued. Or to optimize routes for large logistical operations like collecting garbage or distributing goods. Moreover, it can help in identifying critical groups and areas within a serving region, such as at-risk youth, abusive households, and high-crime sensitive areas. Our advice is to start with identifying public processes that are not running smoothly enough, result in repetitive tasks or simply take too much time of employees. Also investigate what sort of services citizens expect from your organization. And, what if you had a crystal ball and could predict any outcome, what would it be?

2.  Think win-win-win

In our study we found that machine learning applications resulted in labor savings while the quality of the service was improved by offering quicker responses and 24/7 availability, and it increased employee satisfaction. Labor savings in public services were mostly reported as better utilization of resources. Technology could take over tasks from employees, like straightforward tasks that do not yet include a lot of ‘intelligence’, such as answering frequently asked questions. Personnel could focus on more difficult and interesting tasks, and they do not have to handle the same questions all day long. Staff can focus on exceptions and better serving those, which makes their job far more interesting and value adding. As a virtual assistant would be available 24/7, citizens’ perception of the quality of the service will also increase and answers could be given to them quicker and in a more consistent way than before. That’s a win for citizens, employees and the organization.

3.   Tackle potential barriers in an early stage

When considering implementing machine learning, it is important to take time to exploit the data that you have at your disposal. We advise you to assess the quality of your data and combine different data sources, including open data. Results are to be achieved faster if technical IT skills are available internally. However, the technical IT skills of a vendor are likely to be more advanced. Be prepared to re-assess the status-quo processes, as new ways of managing departments and information sharing will be required across the entire organization. Finally, a success factor is that everyone involved is willing to take the risk of implementing new technology. Both business and IT stakeholders need to go through the whole process together and should be involved right from the start.

4.  Think big, start small

Consider machine learning as an incremental journey towards the higher goal of citizen satisfaction, which is defined up front. We advise an agile way of working in which you iterate and slowly expand the solution to other processes and departments and take the citizen along with you in the journey. This allows the organization to adjust to the new technology while simultaneously gathering feedback for improvement from the users. Additionally, as expectations of such a hyped technology as machine learning are often high (and unrealistic), starting small lets you better manage these expectations.

* Study performed as an internship by Bart-Floris Lensink. This article is a subtract of his final thesis for graduation in MSc Business Information Management at Rotterdam School of Management, Erasmus University class of 2017/2018.