Case study
The demo Reporting Assistant, built on public news.
This case study follows the demo Reporting Assistant, a working system I built on public news to show how a weekly coverage report can be put together by a system you can trust. It is a demonstration, not a client project.
The problem
At a small PR agency, someone searches the news for each client every week, copies the mentions into a document, summarises them and formats the report. It is careful work, it takes hours, and it is easy to miss something on a busy week.
What I built
A system that runs on its own schedule. For each client it collects articles and mentions from news sources, uses an AI model to pick out what matters and summarise it, and drafts the report in the agency’s own format. A person reads and approves the report before it goes to the client.
The demo works on public news only, so it needs no client data. For a client, the same system is set up in their own accounts and comes with written instructions, so the team can see what it does and change what it watches.
How it was tested
Before a system like this is trusted, I check its reports against reports made by hand. Does it find the same coverage? Are the summaries accurate? Does it say so when it is unsure, instead of guessing?
Each step has its own automated tests, and every run keeps a log, so a mistake can be traced back to where it started and fixed there.
The result
The aim is a weekly report that is drafted before the working day starts, so the time that went into searching, copying and formatting goes into reading and advice instead.
The next step
If your team spends hours on reports like these, start with a Workflow Review. I map how your reporting is done today and price two or three automations. The fee is credited if you go ahead.