Time recording with AI – from manual allocation to automated efficiency

Employees save time, management benefits from clean data: an AI-supported solution for time recording in the project management tool Monday.com

Customer

Customer benefits

Technology stack

"What used to often lead to queries or errors is now a clearly structured and partially automated process. The combination of AI and integration in Monday.com has significantly simplified our day-to-day work."

Management

Challenges of the customer

The customer uses Monday.com as a project management tool in which project times are also recorded. Employees create daily bookings there with a descriptive text and had to assign each booking to a suitable task. The customer, project and cost center are clearly defined via a task – this means that the correct link is crucial: both for subsequent invoicing and for internal evaluations in management.

However, manual assignment proved to be time-consuming and error-prone. Employees often had to consult with each other or left tasks unassigned, which led to extra work at the end of the month. The aim was therefore to automate this process and increase data quality.

Our approach

jovoco developed a multi-level approach that combines AI methods and system integration to achieve reliable and efficient automation.

Analysis of different methods

At the beginning, the inventory data was divided into test and training data, then different approaches were tested - from simple GPT prompts to finely trained models. The aim was to assign the descriptions of the bookings to the appropriate tasks as precisely as possible. The process and technologies of the target system were also defined.

Structure of the similarity search

A semantic search system was developed that searches both current bookings and historical data. New entries are compared with similar past tasks to determine the most likely classification. The index of historical data is updated daily - manual corrections are also automatically incorporated into future classifications.

Integration of a specialized classification model

A model optimized for this use case receives the results of the similarity search and makes the final assignment decision on this basis. This ensures consistent and context-aware classifications and reduces errors in the long term.

Integration into existing systems

The solution was provided as an API in Azure AI Foundry and linked to the booking data via Microsoft Fabric. The automatically determined tasks are fed back into the project management tool via the Monday API. This has increased the level of automation and quality - and at the same time made it possible to carry out a final check of the assignment.

Testing and introduction

Before the rollout, a test phase was carried out in which assignments were checked and optimizations were made. The solution was then fully integrated into everyday working life - with a high level of acceptance among employees.