Customer testimonial: Feedback on the implementation of the Kisio chatbot

Kisio called on Dydu to deploy a chatbot for their employees. Sarah Geret, HR Development Manager, and Hugo Gomez, HR Manager at Kisio, talk about the implementation of this project and future prospects.

Can you describe Kisio and your role ?

Kisio, a French subsidiary of the Keolis group, is working towards a societal, environmental and energy transition, by promoting sustainable mobility at the service of citizens. On a day-to-day basis, we combine consulting and operational services, offering local authorities, transport operators and businesses an expert, multidisciplinary and tailor-made approach to mobility. Kisio employs 700 people in 16 branches and 15 bus stations (Amiens, Nancy, Villeurbanne, Paris, Aix-en-Provence, etc.).

In parallel, we find Hove, also a subsidiary of Keolis. Made up of around a hundred employees, mainly based in Paris, it represents the IT side of the company, and has established itself as a key player in data-mobility for 24 years.

Toby, the chatbot, is deployed on Teams and used by the 800 employees of both subsidiaries.

What is the role of Selfcare tools, particularly chatbots, at Kisio?

Prior to the arrival of the chatbot, we didn’t have any self-care tools and were unfamiliar with the concept. However, the customer relations department was considering the use of innovative tools for their own customers, and was looking for volunteers to test the chatbot solution. Our HR department is often keen to experiment with this type of tool. This was not a need, but a perfect opportunity to test an innovative product.

Since the launch of the Toby bot in May 2023, we’ve really come to see it as a fully-fledged member of our team.

Which issues did you want to address by implementing a chatbot?

We saw the project as an opportunity to be a forerunner in this field, with specific objectives:

• Modernize service
• Improve the employee experience
• Shorten waiting times
• Handle first-level HR requests
• Make daily life easier for HR and managers, and give them more time to support their teams.

What perimeters and use cases does your chatbot address?

Our chatbot covers all HR perimeters, answering day-to-day questions relating to: paid leave, CET (time savings account), teleworking, mutual insurance, health insurance, CSE benefits, professional interviews, HRIS system, our internal tools (Sphinx), luncheon vouchers, etc.

At this stage, the bot is more geared towards employee rather than manager support, but we make sure to cover as many questions as possible and redirect to the right contact when needed.

Examples of typical questions:

• “How can I apply for leave/time-off?”
• “I’d like to resign, who do I send my letter to?”
• “How much will I be paid while I’m off work?”
• “I’ve been recognized as a disabled worker, who should I send it to?”
• “I’d like to declare a pregnancy, who do I contact?”
• “I’m moving house, how many days can I take off?”

What difficulties did you encounter in creating your knowledge base?

We have a bot that manages requests from two companies, which makes feeding the knowledge base a little more complex. Some answers that are valid for Kisio are not necessarily valid for Hove. In order to personalize answers according to each employee’s company, we set up “knowledge trees”.

To build our knowledge base, we had to collaborate with several divisions (Payroll, Legal, HR, Personnel Administration, Training) to cover as many questions as possible. Right from the start, we understood how important it was to coordinate and master the tool’s operation. Dydu’s pre-written HR knowledge base was a great help, serving as a working basis right from the start.

What were the different stages in setting up the project?

1. First scoping meeting: we presented our expectations for the project and how Dydu could meet them.
2. Soliciting employees: to vote on the bot’s name and design, and to give their opinion on the bot’s merits and fears.
3. First BMS training: for knowledge creation and HR knowledge base customization.
4. Internal meetings: we organized internal meetings over a 2-month period to feed the knowledge base. These meetings were broken down into batches, so that we could move forward in a targeted and efficient way without having to deal with all the knowledge at once.
5. Second Dydu training session: learning how to replay conversations, analyze statistics and KPIs.
6. Launch webinar: accompanied by our Dydu project manager, this was an opportunity not only to present the chatbot, but also to allay the fears expressed by employees beforehand.

Have you taken any specific actions to boost your bot’s traffic?

Since the launch of the chatbot, we’ve carried out several actions:

• Organization of a launch webinar in collaboration with Dydu.
• Internal communication to solicit employees: “Do you validate Toby’s trial period?” (80% expressed their approval).
• Integration of the chatbot into our welcome sessions and internal newsletters.
• Presence of the bot on our intranet, our organization chart and on a poster in our cafeteria.

We integrate the chatbot into our media as soon as possible, and plan regular communications to remind people of its existence and provide statistics on its use: the number of requests, the most frequent topics, etc…

Do you think the chatbot is a success? Why?

The HR chatbot has been completely adopted by employees, who don’t hesitate to ask for it. To further encourage its adoption, we celebrated the validation of its trial period 4 months after its launch, and in June 2024, we celebrated its first birthday! That same year, we had the opportunity to take part in the 2nd edition of the Keolis People’s way awards, a ceremony that rewards HR innovation in Group companies. Our chatbot was awarded 4th place worldwide for having contributed to making Keolis a “great place to work”! We proudly displayed its diploma in our office.

Likewise, we never miss an opportunity to remind people of its existence and the importance of keeping the bot up to date by giving feedback at the end of interactions. We are fortunate to have a 20-30% participation rate in the satisfaction survey, with a majority of positive feedback. This shows that 80% of our responses meet employee needs, which is very encouraging for us.

We’re also noticing a real reflex emerging among employees: they think of requesting the chatbot before going to HR. As a result, we are increasingly called upon for more complex questions, or those that require personalization.

What KPIs do you track to monitor the bot’s activity?

We focus mainly on the number of conversations and interactions, and the most frequently used themes. What’s most important to us is to know how much the bot is used, the satisfaction of the people who use it, the quality of the responses and our ability to improve them accordingly.

Analyzing these statistics allows us to see the qualification rate.

When bot usage drops, we know it’s time to launch an internal communication campaign to revitalize its use.

What lessons or best practices have you learned?

Communication : 

We strongly recommend maximizing communication before, during and especially after the project launch, using all available channels (Teams, Newsletters, Intranet, E-mail, Managerial Memos). In general, we communicate on new knowledge to demonstrate that the bot is regularly updated and can answer a growing number of questions. As a result, we see an increase in traffic and users think about requesting the chatbot. It’s also essential to integrate it fully into the employee onboarding process, as this is often when they have the most questions about the company.

Feeding and updating knowledge:

Regularly feeding the bot, re-reading and analyzing conversations is crucial to effective management. Employees quickly become frustrated when their questions are not answered to their satisfaction.

Rather than simply saying “I don’t have any knowledge on this subject”, we recommend directing the bot’s response differently, for example: “At the moment, I don’t have any knowledge on this subject, we’re working on it internally. Come back in 12 weeks for an update.” This approach not only reduces employee frustration, but also shows that the tool is regularly monitored and updated.

Why would you recommend the Dydu solution?

First of all, we were very well supported and trained by our project manager, and everything went very smoothly. The Dydu platform is easy to understand, even for us HRMs who have no knowledge of computer coding; all it takes is a little curiosity. What’s more, we were pleasantly surprised by your HR vertical, which is more than complete with its 300 pre-written knowledge items. All we had to do was customize them to our specific needs.

What developments and/or updates are you planning to improve the chatbot and encourage its use?

Currently available only on Teams, we plan to extend the chatbot to Slack and our new intranet to onboard a larger number of employees. We also plan to set up two separate authentication spaces for employees and managers, with responses tailored to their needs and status.

We feel that this project required an initial investment and took time to set up. Today, our main objective is to keep it up to date and continuously evolving.