Development of the AI HR Knowledge Base Assistant
2
1
3
4
5
The Brief
User Research
Outputs
Quantitative assessment
Wrap up
Overview
Co-creation
Usability testing
Customer journeys
Service blueprint
Impact assessment
Surveys
Impact
#1 The Brief
Background
Booking.com has over 10,000 employees working in 140 offices in 70 countries. They utilise a ServiceNow database to ensure that employees access to relevant HR information and can create HR tickets. They receive thousands of tickets per month that could be answered if employees read and understood the information on the database. This requires manual work by HR Ops specialists to resolve. To reduce the number of these tickets, they decided to create the HR People Assistant. I was employed as a contractor to ensure that the tool would reduce the number of tickets while offering a great employee experience. The team initially wanted to launch the tool within the first month of me starting and hoped I would clear up a few usability hiccups and inform change management. Based on the research, we cleared up enormous usability issues, delayed launch until it was ready and shifted the product strategy to better align with the strengths of AI.
Challenges
-
Unclear acceptance criteria meaning AI contractor was trying to create a product that did not align to employee expectations
-
Pressure on launching the tool quickly to understand the impact on resourcing of People Ops team for 2026
#2 User Research
The initial research brief was to figure out the expected experience baseline and assess if the tool measured up it. This was done by employing co-creation and usability testing techniques.
Co-creation
Among many uses, the AI tool was designed to assist in scheduling vacations, enquiring about leave balances and querying policies. Employees were instructed to 'create' this flow - acting as both the AI and the employee.
Usability testing
Employees tested the above scenarios with the tool and then compared their expectations versus the reality of the solution. Over 30 usability issues were discovered, including 3 critical issues.
Impact
Launching the tool at this phase could cause an immense rejection from employees - ultimately causing it to fail. It was recommended to delay the launch until these were resolved. Leadership agreed and the launched was put on hold.
#3 Outputs
Leadership wanted to have a better overview of who we were building for, and what problems we were solving. Based on interviews, I created 17 user journeys of how different issues are solved today, and service blueprints of how they would be solved moving forward.
#4 Quantitative assessment
To prioritise problems by business impact, I conducted an assessment of 1,600+ HR tickets from 2025 and outlined which journeys they were associated with (see above journey for the ticketing numbers).
Post launch, I coordinated and sent a survey to a representative sample of employees to assess how employee sentiment towards the tool. 76% of employees found the tool trustworthy, and 80% of employees would encourage others to use the tool. Ultimate success of the tool is that it has decreased the amount of tickets per month by 500 - further streamlining the HR Ops department.
#5 Wrap up
Expectations for the tool were high from the start. Anchoring the product development process with real users delayed the launched, but ensured acceptance among Booking employees. The tool is widely adopted and praised for reducing time to query their HR needs and streamlining ticketing flows.