In the United Kingdom, Government Communications has developed Assist, the first generative AI tool approved for cross-government use. Its rollout highlights that successful AI implementation requires more than technical excellence—it demands human-centred strategies. Embedding AI into public sector workflows involves behavioural, organisational and cultural considerations. Through a structured framework—Adopt, Sustain, Optimise—the UK Government has demonstrated how public bodies can bridge the gap between innovation and daily use, ensuring safe, sustained and effective AI integration.
Adoption: Removing Barriers to Initial Use
Developing an AI tool does not guarantee its uptake. People need to see clear relevance to their roles, feel confident in using the tool and believe that others around them are engaging with it. Adoption begins with understanding users’ perceptions and identifying barriers such as low AI literacy, lack of trust or uncertainty about the tool’s usefulness. Defining the user journey and simplifying early interactions, such as onboarding and initial training, supports uptake. Assist’s rollout showed that targeted communications, simplified processes and early involvement of leaders helped close adoption gaps.
It is important to identify who is and is not engaging with the tool. Gathering data on digital confidence, AI familiarity and demographic background helps uncover disparities and informs targeted interventions. Interviews with non-adopters reveal critical insights—such as the perceived lack of relevance to daily tasks—which can be addressed by adjusting messaging and support materials. Evaluating adoption strategies and adapting based on feedback ensures that initial engagement is not left to chance but is systematically nurtured.
Sustainment: Embedding AI into Daily Workflows
Once a tool is adopted, it must be embedded into daily routines to deliver meaningful impact. Sustained use requires habit formation supported by structured prompts, perceived benefits and repeated application. Routine engagement with GenAI tools does not emerge naturally; it must be designed and encouraged. Successful sustainment efforts begin by defining what counts as regular use and developing tools to track usage across the organisation.
Feedback loops and user research identify friction points, such as lack of time, unclear prompts or difficulty navigating interfaces. Mapping the user journey and pinpointing specific barriers allows support strategies to be deployed more effectively. Webinars, reminders, accessible resources and ongoing communication help reinforce the habit of using the tool. Encouraging team leaders to advocate use and setting clear expectations around when and how the tool can be used fosters consistency and normalises engagement.
Must Read: Scaling AI in Healthcare: What’s Holding Providers Back?
By adopting a robust impact measurement framework, organisations can track how AI tools contribute to efficiency and output quality. In the case of Assist, combining self-reported time savings with quasi-experimental methods provided a well-rounded view of its impact. Clear definitions of success, such as weekly use or contributions to specific tasks, help maintain focus and guide future improvements.
Optimisation: Ensuring Quality and Managing Risks
Beyond regular use lies the challenge of ensuring AI tools are used effectively and safely. Without appropriate training, oversight and risk mitigation, routine use can give rise to hidden risks. These include poor quality control, inappropriate applications and reliance on outputs without sufficient review. Many risks arise not from dramatic system failures but from ordinary human behaviour—fatigue, time pressure or unclear accountability.
Understanding how tools are being used in practice is essential for optimisation. Monitoring input patterns, conducting qualitative research and identifying mismatches between design intent and user behaviour enable adjustments that improve safety and performance. Responsible optimisation involves assessing whether the “human in the loop” has the necessary time, authority and skills to oversee AI use. It also involves creating psychological safety for users to flag concerns without fear of reprisal.
Structured risk identification and mitigation frameworks help delivery teams surface and manage potential harms. In the development of Assist, over 90 hypothetical risks were identified, ranging from workflow issues to quality assurance challenges. Distributing responsibility across interdisciplinary team members created ownership and made risk management part of routine delivery. Training programmes were adapted to help users understand the limits of the technology, build prompt engineering skills and develop judgement about when to rely on AI outputs.
Leaders play a key role in supporting optimisation. They need a clear understanding of the tool’s capabilities, its appropriate applications and the broader organisational safeguards required. By investing in interdisciplinary teams, ensuring feedback is acted upon and embedding evaluation into strategic planning, leaders create the conditions for AI tools to deliver their full value while managing the risks of misuse or disengagement.
Scaling AI tools within organisations depends not only on the robustness of the technology, but also on the human systems that support adoption, routine use and responsible optimisation. By applying the Adopt, Sustain, Optimise framework, organisations can increase engagement, build confidence and ensure safe, high-impact deployment of generative AI tools. Focusing on the people using these tools—how they learn, engage and apply them—ensures that AI becomes an integrated part of organisational workflows, rather than an underused solution.
Source: UK Government
Image Credit: iStock
