“This is a really good tool. I wish I’d had it as a new rep. In fact, even though I’m an established Rep, I would hope that it would be made available to all of us because sometimes we need the practice.”
Union building relies heavily on one-to-one conversations in the workplace - whether persuading workers to join, encouraging members to take action, or resolving workplace issues. These conversations place high expectations on union reps, and whilst some find this comes naturally, many need additional support.
The challenge for unions is that whilst training courses are often excellent, there can be a significant gap between completing training and actually having those crucial conversations. After a training session, it might be weeks or months before a rep gets to put their learning into practice. This means that when the moment comes to have that important recruitment conversation, newer reps especially can feel daunting and under-prepared.
Traditional role-playing in training sessions helps, but reps often want more opportunities to practice and build confidence before they're out there representing the union in real workplace situations.
The solution
The UK Public and Commercial Services Union (PCS), working with the TUC Digital Lab, Campaign Lab, and development consultancy Poteris, developed RepCoach - an AI-powered conversation simulator that allows reps to practice recruitment conversations with realistic virtual colleagues.
The tool uses OpenAI's ChatGPT technology, wrapped in a simple web interface that works on both desktop and mobile devices. Reps can engage in text-based conversations with AI personas trained on PCS resources and realistic workplace scenarios. The tool focuses specifically on recruitment conversations, allowing users to practice their approach with different types of potential members.
Generative AI systems like the one behind ChatGPT are sometimes too risky for use in chatbots: they can make things up, give biased responses and introduce other ethical, safety or accuracy challenges. But for simulations used by limited audiences, they can be appropriate. "We decided that although this chatbot technically falls into the higher risk type, the risk was actually far lower because of the context. The interaction is billed as a simulated chat, not authoritative advice," explains the development team. The AI represents fictional coworkers rather than official union positions, and it's designed for internal use by union reps rather than public-facing advice.
The system provides feedback after each conversation, scoring the rep's performance and offering constructive suggestions for improvement. This gamified element helps maintain engagement whilst building skills.
Key benefits
Builds confidence: Reps can practice difficult conversations in a safe environment before facing real workplace situations
Available on-demand: Unlike traditional training, reps can access practice sessions whenever convenient, including evenings and weekends
Realistic scenarios: AI personas respond naturally, creating believable conversation experiences
Immediate feedback: Users receive instant analysis of their approach with specific suggestions for improvement
Flexible access: Works on smartphones, tablets, and computers, allowing practice during breaks or commuting
Bridges training gaps: Provides ongoing support between formal training sessions
Common questions
How difficult is it to implement?
"The development was iterative, allowing us to initially rapidly prototype a basic version of the tool," explains Peyman Owladi from Poteris. The basic framework can be adapted relatively easily, though each new scenario requires research and background data specific to that conversation type.
What skills are needed?
For users, no technical skills are required - it works like a normal text messaging interface. For implementation, unions need access to web development expertise and AI integration capabilities. This is what Poteris offered in this build.
What are the main challenges?
Data protection is crucial - the team ensured they only used AI services compliant with GDPR that guarantee conversations won't be used for training the AI model. There's also the challenge of creating realistic personas and scenarios that feel authentic to union reps.
How do you ensure quality and accuracy?
The AI was trained using PCS resources and materials, and underwent extensive testing with nine PCS reps. "We iterated these prompts until they gave outputs that were realistic and passable," notes the development team. Regular testing and refinement helps maintain quality.
What about data protection?
The team specifically chose AI services that guarantee that conversation data won't be used to train their models. The tool is designed for internal union use, reducing the risk of sensitive information being shared inappropriately.
Getting started
First steps:
Identify the specific conversation types most valuable for your reps (recruitment, retention, campaign mobilisation)
Gather existing training materials and resources to inform the AI training
Consider starting with a simple prototype focusing on one conversation type
Plan user testing with a small group of experienced reps
Common pitfalls to avoid:
Don't try to build everything at once - start with one scenario and expand
Ensure clear communication that this supplements rather than replaces traditional training
Be transparent about AI use and limitations
Don't skip user testing - rep feedback is crucial for refinement
Resources needed:
Web development capabilities or external development support
Access to AI services (budget for API usage)
Training materials and union resources for AI training
Time for user testing and refinement
Training requirements:
Brief introduction for reps on how to use the interface
Clear explanation of the tool's purpose as practice, not official advice
Integration with existing training programmes works best
Looking ahead
PCS found the testing results "extremely positive" with all participating reps seeing value in the tool. Future development could include:
Extended scenarios: Conversations around retention, campaign talking points, or representation meetings
Voice integration: Adding voice chat options, particularly useful for mobile users
Advanced feedback: More detailed analysis linking to specific union guidance and policies
Organiser dashboard: Interface for organisers to set scenarios and track engagement
Gamification: Badges and progressive difficulty levels to maintain engagement
The team has made the framework open source, allowing other unions to adapt the approach for their own needs.
AI transparency statement
This case study was written by Claude.ai based on a pre-defined case study template and prompt, and using two data sources as context:
The TUC Digital Lab article "Augmenting union education with AI – PCS Case study" (28 May 2025)
A conversation transcript between Nick Scott (Centre for Responsible Union AI), Peyman Owladi (Poteris), and Bobi Robson about the RepCoach project development.
The text was then edited by a human, and approved by the participants in the call and the lead of the TUC Digital Lab.
