How ACV-CVS is using AI to process large numbers of emails from members more effectively

This Belgian trade union is experimenting with AI to categorise huge numbers of emails from members more effectively

Communication is a very important thing for us… when you're handling 600 emails from members every day, you need to make sure urgent issues don't get lost in the crowd.

Stephen Willocx, head of AI in the business and IT department at ACV-CSC, Belgium's largest trade union with 1.5 million members.

The challenge

For many unions, member emails are a vital but overwhelming communication channel. Staff face multiple challenges:

  • Sorting through hundreds of emails daily

  • Identifying which issues need urgent attention

  • Spotting all the questions in lengthy emails

  • Categorising emails correctly for follow-up

  • Ensuring nothing important gets missed

"Everything feels urgent when it lands in your inbox. But realistically, we know about 5% of emails need immediate action. The challenge is finding these quickly and reliably."

Stephen Willocx

The solution

ACV-CSC developed an AI system that processes incoming emails within minutes of arrival. "We started with three of our largest mailboxes," Stephen explains. "The system runs on our own infrastructure for security - it's not ChatGPT, but rather the engine hosted at our place."

The system performs several key tasks:

  • Categorises each email using predefined categories chosen by each service

  • Identifies truly urgent issues requiring quick response

  • Highlights specific questions within each email

  • Adds clear markers in Outlook for staff

"We built it step by step," says Stephen. "First, we got the categorisation working. When that went well, we added urgency detection. Then we added question identification. Each addition made the tool more valuable for our staff."

Key benefits

  • Emails processed within one minute of arrival

  • Urgent issues identified quickly (calibrated to flag ~5% as urgent)

  • All questions in emails clearly highlighted

  • Consistent categorisation across teams

  • Staff can see key information at a glance

  • More time for responding to members

Common questions

Is it secure?

"Security was crucial," Stephen emphasises. "That's why we host everything on our own infrastructure. No member data leaves our systems."

How accurate is it?

The system was carefully calibrated based on real union needs. "We worked with teams to understand what 'urgent' really means in their context," Stephen explains. "Then we tuned the system to match those requirements."

What skills do staff need?

The system integrates directly with Outlook, so staff don't need special training. They just need to understand what the different categories and markers mean.

Can other unions implement this?

While this specific implementation requires technical resources, the approach could be adapted by other unions, potentially starting with simpler tools or focused pilots.

Getting started

For unions considering similar systems, Stephen suggests:

  1. Start with clear goals for what you want to achieve

  2. Work with teams to understand their needs

  3. Begin with a small pilot

  4. Build features gradually

  5. Get regular feedback from users

  6. Be prepared to adjust and refine

Looking ahead

The success of the email processing system has encouraged ACV-CSC to look for other ways AI can help reduce administrative burden. "This isn't about replacing staff," Stephen emphasises. "It's about helping them focus on what matters most - supporting our members."

AI Transparency statement

This article was generated using AI based on the transcript of a talk by Steven Willocx at a Centre for Responsible Union AI webinar in February 2024.

Reply

or to participate.