The city of Helsinki is conducting agile trials on using artificial intelligence (AI) and robotics to automate city services.

Helsinki aims to be the best city in the world in the use of digitalisation. To achieve this goal, it is conducting agile trials that explore how AI and robotic automation can be used in city services.

“I believe that artificial intelligence, robotics and data are at the core of future developments,” said Helsinki mayor Jan Vapaavuori. “They will also impact the services provided by the city of Helsinki to a substantial degree. The most important thing about these trials is to gain new ideas that can then be scaled to suit the production of various city services. The trials also represent a shift in thinking whereby the aim is to make the city organisation’s operating culture more conducive to experimentation.”

Helsinki is developing internal innovations in the city’s experimentation accelerator, which began its operations in autumn 2019. In 2021, the city completed its second campaign of agile trials, which tested ten ideas submitted by city employees for the use of AI and robotic process automation. City personnel submitted 60 trial suggestions for the two campaign rounds, 17 of which progressed to the trial phase.

The quantity indicates there is great deal of interest within the city organisation with regard to the possibilities of AI. The campaign involved a wide variety of trials ranging from the quality monitoring of student information to a chatbot that can handle matters related to business trips.

“Through agile trials, we can learn to understand how we can use artificial intelligence to build even better digital solutions for the residents and our own employees,” said Mikko Rusama, Helsinki’s chief digitalisation officer. “We hope that the lessons learnt from these trials help us to avoid expensive and long projects that fail to meet actual needs. We also want to renew our culture and policies and make the city a bit more agile.”

The proposals that progressed to the trial phase received €10,000 in funding for implementation in cooperation with corporate partners. The city aims for genuine agility in the trials, and the target for proceeding from an idea to execution is 90 days.

Due to coronavirus restrictions, the trial campaign’s arrangements were entirely virtual. Applications were open from May to June 2020, and the trials were carried out during autumn 2020.

“The experimentation accelerator is used to refine innovations and support experimental activities that leverage digitalisation within the city organisation,” said project manager Ville Meloni. “The trials have also confirmed that it is often not necessary to replace an entire work phase with a robot. In many cases, robot assistance is enough. The trials also provide us with valuable information about the limitations in the data collected by the city. In this context, we sometimes notice that it may not be feasible to automate work tasks in a way that emulates how a person would complete them. Instead, it may be best to redesign the entire process.”

The starting point for the digital development is to facilitate the daily lives of Helsinki residents through the digitalisation of services and to help them find services more efficiently than before. The same applies to city employees: digitalisation expedites work through the automation of routine tasks. In this regard, ideas such as AI-based recommendation engines, chatbots and the machine-assisted identification of electronic signatures may become important in years to come.

The city’s new trials focusing on AI and robotic automation sought answers to the following questions:

• Can AI help with the management of employment?
• Can AI benefit work shift planning?
• Can AI help gain a more realistic and accurate situational picture of traffic emissions in Helsinki?
• Could AI help with the analysis of performance reviews to enable the development needs of employees to be supported better than before?
• Can robotic process automation be used to facilitate the quality monitoring of student information?
• Can a chatbot guide and advise business travellers?
• What is the benefit of a neural network-based AI identifying signatures in PDF documents?

It was often found that an AI is exactly as smart as its teacher. A human is needed to teach the machine, and any inaccuracies in the data or the teacher’s thinking are reflected in the AI’s actions. As such, products and services based on AI should be assessed from a variety of perspectives from the very start of the project. Among these viewpoints, the user’s perspective is the most important to ensure the end product is clear and easy to use. A cumbersome and illogical service will be quickly abandoned.

An important criterion for the trials was to document the lessons learned and share them with the entire city organisation. This will help later AI developers avoid the initial pitfalls related to aspects such as procurement and data security issues.

At the same time, it is important to ensure the operating logic of the AI-assisted services is transparent. AI-assisted services enable city employees to save significant amounts of time. Even now, there are plenty of potential applications for recommendation algorithms, in particular, in the context of library services and training courses provided by the city, for example. In terms of the trials, it is also important to identify which technologies could be more widely viable for the city organisation. This ensures that the city gains the most benefit from AI systems.

The planning of the trials was supported by Ai4Value, Aureolis, Deloitte, Digital Workforce Services, Helsinki Intelligence, Houston Analytics, Gofore, Lekab, Microsoft, Reaktor, SAS, Siili, Silo.ai and Sofor. The implementation partners for the trial project were Gofore, Lekab, Reaktor, SAS and Siili.