An AI tool like Claude, ChatGPT, Copilot or Gemini is one you can have a written conversation with.* Give it a task in plain language and it works through it, often producing outputs that would take you much longer to create yourself or ones which would previously have needed software developer input.
The word "agentic" refers to the ability of these tools to take sequences of actions on your behalf, using tools like file access or web search, and producing a finished result. Claude Cowork, for example, can read documents from your computer, write new files, build websites, and run code following the instructions you give it. If you have set up your personalisation files properly, it becomes even more powerful and useful.
* AI-powered speech-to-text applications are increasingly accurate, and speaking natively using third party apps such as WisprFlow is now my favourite way to interact with Claude.
We are routinely asked to do more with less. There is often a huge gap between what is needed and what a (usually) small team can deliver.
Agentic AI does not close that gap by replacing professional judgement. It closes it by reducing or eliminating the mundane structural portion of your workload. Once you have briefed an AI properly by giving it your professional context, your preferred frameworks and your writing style, it can:
Apply frameworks such as systems thinking, social determinants frameworks and more to a public health issue, producing a complex analysis of an issue or intervention in minutes.
Turn a surveillance spreadsheet into an interactive visual dashboard without specialist software or a data team.
Write plain-language health information for vulnerable communities in multiple languages simultaneously — adapted for each audience, not just translated.
Produce a structured policy briefing from bullet-point notes in the format your audience expects with the right tone, length, and framing.
Generate and test multiple message variants for a health campaign before any budget is committed.
Create a complete, accessible health information website for a specific community and topic in minutes.
A basic chatbot gives you a response you then have to do something with. An agentic tool like Claude Cowork can read a policy document from your hard drive, apply a structured analytical framework to it, produce a formatted briefing note, save it as a Word document, and give you a draft ready to review — all in a single instruction. It remembers your preferences across sessions, builds on previous work, and applies your professional style to everything it produces. The setup guide on this site will hopefully get you to that point.
It is not a substitute for professional judgement. AI can structure an analysis, but it cannot replace the contextual knowledge, ethical reasoning, and local expertise that public health practitioners bring to the table. The outputs it produces are drafts, not final versions.
It can be wrong with confidence. AI tools sometimes produce plausible-sounding statements that are factually incorrect — a phenomenon called hallucination. This is more likely with specific statistics, citations, and recent developments. Always verify claims against primary sources before using them in formal work.
It needs context to perform well. An AI that knows nothing about you will produce generic outputs. The more you invest in setting it up — by using the personalisation files in the repository, giving it your professional background, specifying your frameworks and so on — the more useful it becomes.
It is NOT suitable for confidential data on consumer plans. Do NOT paste personal patient data or confidential organisational documents into a consumer (paid) AI account. Enterprise options with proper data protection exist — ask your organisation's data protection officer before using AI tools with sensitive data.