Designing responsible AI-supported public services: lessons from Think AI for Government

AI can improve public services, but adopting it is only a starting point. Teams need to understand the problem they are trying to solve and involve the people affected. They also need to measure whether AI improves the service.
Solirius Reply consultants Rocio Baldasano and Bella Chesworth attended Think AI for Government. They explored what the discussions meant for design and business analysis. Sessions covered public engagement, local government and service delivery, with one question connecting these themes: How can teams design AI-supported public services that keep people informed and in control?
Key Takeaways | Summary |
Start with the problem | Understand people’s needs and the wider service journey before choosing AI. |
Test understanding, not just confidence | Check whether people can use AI-supported guidance to complete tasks correctly. |
Make transparency useful | Explain AI’s role, its limitations and how to question outcomes or get human help. |
Design with users and staff | Involve them early enough to shape whether and how AI is used. |
Measure real improvement | Assess outcomes for users alongside service performance, costs and review responsibilities. |
Start with the problem and the whole journey
Several sessions stressed the need to understand the problem before choosing AI as a solution. Carl Kinson from DXC Technology discussed how teams can prepare to adopt AI. This includes identifying the data they need and understanding how AI could affect people and processes.
User research, design and business analysis bring complementary approaches to this work.
User research helps identify where people struggle.
Design helps teams explore improvements and better understand the problem.
Business analysis examines processes to identify delays, repeated tasks and dependencies.
Together, these disciplines help teams define clear outcomes and assess whether AI has a useful role in the service.
Jose Puyana and Philippa Evans’ Online Trade Tariff presentation offered a practical example. It explored the needs of the “nervous trader”. This is someone who needs help understanding complex trade information so they can proceed with confidence.
The presentation showed how AI-supported guidance fits into the wider user journey, covering discovery, classification, preparation, declaration and repeat use. For us, this raised two questions: Does the guidance help traders understand their next steps and complete their tasks? Do they also recognise when they need specialist support?
Design and business analysis offer complementary ways to explore this:
Design: test whether people understand the guidance and can act on it.
Business analysis: define successful task completion, identify when people need specialist support, and agree how to measure process improvements.
If we introduce AI-supported guidance at Solirius Reply, we would need to test both confidence and understanding. People may feel reassured by an answer without understanding its limitations. Testing should therefore check whether they can use the guidance correctly.
Make transparency useful
The AI and the public panel stressed the need for transparency. This means explaining how AI works, where it contributes to a process and how it affects outcomes.
Depending on the service, people may need to understand:
Where AI contributes and what its limitations are.
What information they should check before acting.
How to correct information, question an outcome or get help.
Teams need clear requirements for data use and human review. They must also agree who is accountable. Designers, business analysts and technical specialists need to collaborate to explain AI clearly. They must describe its role and limitations and match how the service works.
Research should test whether explanations about AI help people make informed decisions within the service.
Co-design with the people affected
The Online Trade Tariff presentation summed up another key theme: “Co-designed, not imposed.”
The presentation described working with traders and specialists, with real users involved in testing at each stage. Other sessions explored public deliberation and the importance of understanding local needs.
Teams should involve people early so they can shape whether and how the service uses AI. This means understanding their challenges and exploring ideas together. Teams should then use what they learn to adapt the service and AI’s role.
Staff involvement matters too. AI can change daily tasks, review responsibilities and support needs. Research and process analysis can help teams identify these changes. Teams can then agree who owns the revised process and what training staff need. This could cover:
Understanding AI’s limitations.
Checking its outputs.
Knowing when to refer a problem for further review.
Measure whether the service improves
The event prompted us to consider how design and business analysis can help evaluate AI-supported services. Evaluation could cover the following user and operational outcomes:

These are proposed measures, rather than results demonstrated at the event. Teams would need a baseline and ongoing evaluation to understand whether improvements are sustained and where new problems emerge.
Putting the lessons into practice
For teams at Solirius Reply, the starting point is to define the problem and involve users and staff. Teams should make AI’s role clear and agree how to measure success. The user and operational outcomes above can guide this.
Design and business analysis help teams assess where AI-supported guidance adds value and identify where people need human or specialist support.
Success means more than users feeling confident. They also need to understand the guidance’s limitations and use it correctly.
About the authors
Rocio Baldasano is a Digital Design Consultant at Solirius Reply. She specialises in UX and interaction design for government services.
Bella Chesworth is a Digital Business Analyst at Solirius Reply. She focuses on requirements gathering and process analysis for public sector clients.
To learn more about our design and business analysis services, get in touch with Solirius Reply.




Comments