Daily News Intelligence Agent

The Challenge

A communications director at a major public pension system needed to stay on top of every media mention of his executive leadership, every piece of news about the organization, and broader industry trends affecting public retirement systems.

He was paying $5,000 a year for a media monitoring service. And it still wasn’t working.

The clipping service returned a flood of loosely relevant results. He was spending 30 to 45 minutes every morning—sometimes a full hour—manually sorting through them, deciding what actually mattered. Critical coverage slipped through because the search terms were too blunt. He needed to track everything from executive mentions to ESG investing regulations to pension fraud cases, and the existing tool couldn’t distinguish signal from noise.

He needed comprehensive daily intelligence across three priority tiers: executive mentions first, organization-specific news second, and industry trends third. What he was getting was a firehose.

The Approach

Rather than recommending another monitoring subscription, I built a custom AI news agent on the Relay platform.

The setup was straightforward. We configured 11 targeted search queries covering his priority areas: executive mentions, organization coverage, ESG investing trends, pension legislation, fraud cases, and developments at other retirement systems. The agent runs daily at 10am, pulls the previous 24 hours of results, then passes everything through an AI filtering and compilation step.

The key design decision was the dual reporting system. He gets a daily news alert for immediate coverage, plus a weekly trend analysis for strategic insights. The daily report includes an executive summary, importance ratings for each article, organizational relevance notes, and paywall indicators so he knows what he can actually read.

We refined the system live. Early testing revealed that generic search terms were missing relevant articles. Using an AI copilot to diagnose the gaps, we added targeted terms and confirmed the workflow was now capturing coverage from industry publications it had previously missed entirely.

The Results

The shift was immediate and measurable:

He saves over 150 hours annually in manual sorting time. He eliminated a $5,000 annual contract. And he’s getting better coverage than before, because the AI can evaluate relevance in ways that keyword matching simply can’t.

But the strategic value might matter more than the time savings. The weekly trend analysis gives him predictive insight into regulatory shifts and investment opportunities, turning a reactive monitoring task into a proactive intelligence function.

Why This Worked

Three things made this implementation land:

• Solving the right problem. He didn’t need more search results. He needed less noise and better judgment about what mattered. The AI filtering layer transformed the workflow from “sort through everything” to “here’s what you need to know.”
• Right-sized technology. No enterprise platform. No IT involvement. No six-month implementation. We built this on an accessible agent platform with a lightweight AI model. He was operational within days.
• Iterative refinement. We didn’t try to get it perfect on day one. We tested, found gaps, diagnosed them with AI, and adjusted. The system keeps getting better as we tune the search terms based on what he actually needs.

The Broader Lesson

This project illustrates something I keep seeing: the most impactful AI implementations aren’t the flashiest. They’re the ones that take a specific, recurring pain point and eliminate it completely.

A $5,000 annual clipping service was a band-aid over a workflow problem. The real issue wasn’t access to news. It was the daily cognitive load of figuring out what mattered. An AI agent that costs a fraction of the price now handles that judgment call, and does it better.

For any professional who spends time every day manually monitoring, filtering, or synthesizing information, this pattern applies. The question isn’t whether AI can help. It’s whether you’ve defined the problem clearly enough to let it.