You are not wrong to feel anxious, and you are not alone.
In the federal public service, the pressure to get on board with AI is real and coming from the top. When something arrives in the letter of authority, it quickly finds its way into departmental planning, workplace tools and daily expectations. So, if it feels like AI has gone from a spontaneous curiosity to an institutional priority too quickly, that’s because it has.
But hesitation doesn’t mean you’re behind. Maybe it means you’re paying attention.
Public servants are stewards of sensitive information and public trust, so adequate caution here is prudent. And the tension between innovation (“move fast”) and risk management (“be careful”) is, in many ways, the main challenge of the current administration. In the traditional civil service culture of fearlessly giving advice before giving honest implementation, much of the solution to this tension falls to civil servants tasked with translating policy ambition into action.
So, is speed itself the real problem? Not exactly. How these tools are established, managed and used are more important. A poorly managed adoption poses risks regardless of speed.
But ignoring AI completely won’t solve much either. More often than not, it creates “shadow usage,” where employees use unauthorized tools without protection or oversight. The most dangerous use of AI isn’t inside the sanctioned systems – it’s what’s happening quietly in the browser tabs in the corner of someone’s desk.
Ideally, AI adoption should be seen less as a general rollout and more as a controlled experiment: testing low-risk use cases, learning where these tools truly add value, and building governance as implementation progresses. However, meaningful boxing requires time, administrative ability, strategic patience and sustained investment, all of which can feel impossible in an environment increasingly defined by doing more and doing less.
That fact does not eliminate the need for caution. If anything, it makes thoughtful implementation and usability even more important.
Your concern about private companies is well placed, too. Most of the AI tools are developed by a small number of large companies, many of them foreign private sector – the same companies that governments now, perhaps more regrettably than before, depend on. That raises legitimate questions about data freedom, vendor lock-in, procurement integrity and the long-term control of public sector capacity.
To its credit, the federal government is excited AI strategy for public servicetogether Treasury Board Secretariat Manualit does not demonstrate efforts to establish safeguards around the use of AI, privacy, procurement and accountability. The work remains inconsistent and may continue to lag behind ambition, but it does indicate that governance is considered important, even if its parts are already in play.
It is also worth distinguishing between different types of AI applications. There is an important difference between posting sensitive information to a publicly available chatbot and using business AI tools approved by your department within protected government systems.
The former presents obvious privacy and cybersecurity concerns, while the latter is usually governed by strict contractual, technical and policy protections. In many enterprise-level systems, those controls are specifically designed to prevent sensitive corporate data from being widely referenced to train public-facing models.
That doesn’t make licensed systems risk-free either, but it does make them very different from free-for-all. And that difference is important because it reflects a broader reality of modernisation: the public service needs to protect against risk and adapt to expectations of greater efficiency and improved service delivery.
In practice, responsible use of public sector AI tends to follow a few common sense principles: don’t enter sensitive information into unauthorized systems, hold humans accountable for results, verify results carefully, be transparent about where/when AI is used, and treat AI as a support tool, not a decision maker.
An important comparison: AI is less like a roommate and more like a confident student or micro-analyst – useful for writing, summarizing, planning and reasoning – but also prone to errors, factual errors, shallow thinking and a sycophantic tendency to tell you what looks good rather than what is true. Although it may save time on certain processes, judgment, responsibility and final decisions should remain with you.
So, where does this leave you? Stick to approved tools. Avoid entering sensitive information into unscreened systems. Start with low-risk administrative tasks. Treat the results as first drafts, not final answers. And keep your skepticism in check – AI hype is real, and not every use case adds value.
Public service does not need blind AI users. It requires dedicated professionals who are willing to engage actively, understand the constraints, and speak up when governance, privacy or public trust may be compromised. Right now, perceived concern can be one of the greatest strengths of public service.
– Jacob Danto-Clancy, Civil Service Secrets
