Scenario 1 – Co-design and transparency
An agency set out to develop an AI-powered virtual assistant to help citizens navigate complex government services. From the outset, the project team prioritised inclusive co-design, engaging a diverse group of stakeholders including frontline staff, citizens from different cultural and linguistic backgrounds, and accessibility advocates.
Through workshops and user testing, the team identified key needs – such as support for diverse accents, simplified language, and clear pathways for users with low digital literacy. These insights directly informed the design of the virtual assistant, which was built using a fine-tuned language model grounded in approved policy content.
To ensure the virtual assistant could move beyond a successful PoC and scale across the agency, the team embedded several critical practices from the start. The PoC was strategically aligned with a clear business priority – improving citizen access to government services – and measurable outcomes were defined in partnership with business owners. The solution was architected for scale, using modular components and open standards to support future integration and avoid vendor lock-in. Privacy, security and ethical considerations were addressed early, with oversight from a governance board and compliance with the Australian Government’s AI policy and technical standards. A clear plan was established for data governance and integration as the project moved towards production. Executive sponsorship was secured before the PoC commenced, and regular updates kept leadership engaged, ensuring ongoing support for scaling. The team also planned for workforce capability uplift and change management, recognising that successful scaling would require new skills and ongoing support for staff. Feedback loops and performance monitoring were built in from the pilot phase, enabling the team to adapt the solution based on real-world use and user feedback.
The result was a virtual assistant that improved service accessibility and user satisfaction through transparent, traceable responses. The project was recognised as a model for ethical and inclusive AI design, and the co-design approach has since been embedded into the agency’s broader digital transformation strategy.
What worked well
- Early and inclusive co‑design with frontline staff, culturally diverse users and accessibility advocates ensured the solution reflected real‑world user needs.
- Clear strategic alignment with a business priority – improving access to government services – and jointly defined measurable outcomes.
- Solution engineered for scale using modular components, open standards and early attention to privacy, security and ethical safeguards.
- Strong executive sponsorship and continuous leadership engagement sustained momentum.
- Investment in capability uplift and structured change management supported readiness for scaling.
- Embedded feedback and monitoring loops enabled iterative improvement during pilot testing.
Lessons learned
- Without continued resourcing and capability uplift, scaling risks stalling – even with a strong PoC.
- Scaling requires sustained alignment across business, ICT, legal, policy and operational teams; fragmentation risks slow‑downs.
- Even strong co‑design efforts should be maintained through production stages to avoid drift away from user needs.