Agents – Unlike most use cases for AI over the past several years, Agents do the work WITHOUT the human. They run autonomously, based on conditions and then provide a multiple step result. They take some input (a prompt, an event, a trigger) and then go through a series of steps, all tied together with AI. It’s like zapier or other automations, but it is ‘intelligent’ in that it can think and make decisions, develop writing, etc. One straight forward example is to have it read daily news from 10 websites and send you bite sized summaries to your inbox in the format you’d like. Or – for us public health folk – review data as it is being collected/reported and have it create weekly narrative and visual summaries for us.
Example – I built one to look at a survey questionnaire and think through appropriate analysis, providing guidance and actual excel formulas to use. The idea was that a non-data/evaluator could input their survey and receive guidance on how to analyze it, without uploading data, just the questions. I also built one that will template and begin to draft a proposal if you give it the RFP and some background information. Neither are that cool (aka – they don’t work great), but it’s a start. I used MindStudio because I don’t know how to code and it was recommended to me. I have thoughts, but here is an example of one the workflows, each step contains detailed instructions supported by AI.
When we talk about AI doing routine/repetitive tasks, we are now finally seeing that come to fruition. Agents can make this happen. Ideas for use in public health include sending automated summaries of data from last week’s surveys to 5 key staff, organizing and sending summaries of public input, monitoring incident reporting, routinely checking data quality, or helping route website visitors to the right content or person via chatbot. This is a short list, I am confident you have more ideas!
