Customer Story

Mazars - Increased time saving efficiency by 50 % through intelligent document classification

  • Ability to independently recognise different formats of the same document

  • Harmonised classification of documents

  • Enormous time saving and reduced risk of errors

Mazars und UKG HR Service Delivery - 50 % Zeitersparnis durch intelligente Dokumentenklassifizierung

Mazars is a leading international accounting firm specialising in tax and consulting.

Challenges

To facilitate the exchange of information between its 37 HR offices across France, Mazars has been working on the digitisation of employee files and the introduction of electronic signatures since 2014.

Information exchange and streamlined processes When it comes to HR organisation, Mazars is undoubtedly one of the most innovative companies. This is because the independent international audit, advisory and business services group has put HR innovation at the core of its organisation: working with innovative start-ups, testing new digital tools, introducing new HR practices, etc. “When we have the opportunity to experiment with new tools that meet our challenges in terms of digitising documents, we jump at it,” explains Sandrine Vassal, Head of Personal Services.

When Mazars’ HR teams were given the opportunity to beta test a new machine learning-based feature developed by UKG, they seized the opportunity. And for good reason: with almost 3,700 employees in France and 280,000 documents to manage, automatic document classification that is as fast as it is efficient is an important goal for this large accountancy group that wants to optimise its HR processes.

Solution

I In July 2020, a new module, Smart Document Classification, was added to the UKG Document Manager solution, which is used daily by HR teams to manage digitised employee files. The aim was to eliminate time-consuming and purely administrative activities when filing a document. “Previously, our HR teams had to click on a dropdown menu and select whether it was a bank statement, an ID document, proof of a family event (declaration of pregnancy, birth, death, etc.), a social security certificate, and so on. In fact, any document that relates to an employee’s professional life,” says Vassal. In total, Mazars teams juggle no less than 80 types of documents. However, Mazars had a good document distribution.

Smart Document Classification – a whole new experience 
Four important steps were taken in the summer of 2020, 

  • Data examination
  • Training of a machine learning model
  • The production launch
  • The beta phase 

And a fifth stage could be added: real-time adaptation, a kind of up-date.

We have halved the filing time, which of course has increased the number of documents filed and reduced errors! Recruiters no longer have to search the drop-down menu for the right title. The level of precision is such that they only need to validate the type of document proposed. It is obvious that we have gained in working comfort.”

Sandrine Vassal

Head of Personal Services

Results

“As HR teams submitted documents, the tool refined its recognition of each document type,” says Sandrine Vassal. She continues: “In the end, the tool has classified about 12,000 documents in the last six months. This is an impressive number. The number of documents to be processed can increase quickly. Which was the case in September, when we welcomed about 350 to 400 new employees. In managing all these new hires, document pre-diction has saved us a tremendous amount of time. We have halved the filing time, which of course has increased the number of documents filed and reduced errors! Recruiters no longer have to search the drop-down menu for the right title. The level of precision is such that they only need to validate the type of document proposed. It is obvious that we have gained in working comfort.

The Smart Document Classification tool even recognises documents that may be in different formats. It’s all a matter of practice! For example, pregnancy declarations are different depending on whether they are issued by a doctor, social security or another organisation”, explains Vassal. Even though the formats differ, the information entered is similar from one document to another. The model is “trained” so that it learns that the information is not always presented in the same order and that the type of document is nevertheless identical.

Conclusion and outlook 

“In the long term, we will have a uniform classification of documents, which will also make it easier to see and search for documents. The prospects at group level are very interesting,” concludes Sandrine Vassal.

12,000
Up to date 12,000 classified documents
50%
time savings in administrative tasks
Your better tomorrow starts today

Learn how our HR, payroll and workforce management solutions help you achieve a better work experience for all your people.

Learn more about our award-winning culture