Dominic Rochon, the deputy minister and chief information officer of the Government of Canada has urged public servants to use technology to address the challenges facing the public service as he opened this year’s AccelerateGOV conference.

How Governments Can Make Smarter Decisions

Most safety testing efforts for large language models (LLMs) today focus on evaluating foundation models. However, there is a growing need to evaluate safety at the application level, as components such as system prompts, retrieval pipelines, and guardrails introduce additional factors that significantly influence the overall safety of LLM applications. In this paper, we introduce a practical framework for evaluating application-level safety in LLM systems, validated through real-world deployment across multiple use cases within our organization. The framework consists of two parts: (1) principles for developing customized safety risk taxonomies, and (2) practices for evaluating safety risks in LLM applications. We illustrate how the proposed framework was applied in our internal pilot, providing a reference point for organizations seeking to scale their safety testing efforts. This work aims to bridge the gap between theoretical concepts in AI safety and the operational realities of safeguarding LLM applications in practice, offering actionable guidance for safe and scalable deployment.

Happy Tuesday, Transformation Friends.

Happy Tuesday, Transformation Friends.

Happy Tuesday, Transformation Friends.

Happy Tuesday, Transformation Friends.

Unveiling the Secret to Service Excellence

Happy Tuesday, Transformation Friends.

All leaders—from new managers to seasoned executives—must delegate tasks to free up time and attention for the big-picture work their more-senior roles demand. But too often leaders find themselves caught in the weeds of execution. This article helps leaders determine which work to keep and which is better left to a team member. Then it identifies four challenges that stop even those who know how to delegate from doing it successfully: (1) an addiction to the dopamine hit of easy productivity, (2) a disinclination to reject requests for help, (3) a desire to meet their own bosses’ or clients’ unmanaged expectations, and (4) a misunderstanding of what “work” should mean for a manager. Drawing on research and two decades of experience advising senior leaders, the author offers concrete recommendations for overcoming each challenge. Even for leaders with the best of intentions, overcoming these challenges will take time. However, if you stick with it and learn to delegate effectively, you’ll get much more out of your team members—and yourself, as a developer of talent and as a big-picture, strategic thinker.

Happy Tuesday, Transformation Friends.

Project managers and planners are highly prone to believing that their projects are one of a kind—partly because those that seem new and distinctive are more likely to win support. But research on 1,300-plus projects reveals that few, if any, actually are unique. The problem is, the perception of uniqueness causes managers to think there’s nothing to learn from other projects, which leads them to underestimate risk, make poor decisions, and blow through budgets and schedules. The cure is to always assume that someone, somewhere has undertaken a project like yours. If you can’t find an analogue, break your project into components, which may prove comparable with other projects. Then use forecasting and other risk assessment tools to avoid biases that undermine good choices.