Services

Case study · Media reporting

Three weeks of reporting. Done in days.

How Loop built an AI agent system that took the mechanical work of monthly media reporting off a 200-person agency’s team, so their time went to analysis, recommendations and client strategy.

Full-service agency200 peopleMedia and analytics

A polybit figure of an account manager sitting on a printed media report on a desk, reading a tablet, beside a pink sticky note

The results

From the first live monthly cycle.

30–40 hrssaved per month, by the team’s estimate

3 weeks → daysto produce the monthly report draft

Zerodata errors found in review of the first live report

Fully designeddrafted straight into the agency’s branded template

Context

A 40-year agency with a three-week problem.

Our client is a full-service, 200-person agency that has served clients in higher education, healthcare, travel and other industries for more than 40 years. Like many agencies pulling together monthly media reports, much of their time went to tedious, repetitive tasks that left little room for the critical thinking and insight clients value.

We spend so much time trying to get the numbers straight and identify what’s changed. The clock is always ticking. It’s a race to the end of the month.

VP of Media

So the agency partnered with Loop to build an AI agent system that does the mechanical work of monthly reporting: aggregating the data, verifying every number, and drafting the report in the agency’s own template and voice. The team’s judgment stays in charge of every step.

The problem

Three weeks to report on a month that’s already over.

Monthly media performance reports are among the most stubborn time sinks in agency life. Data lives across a dozen ad platforms, analytics tools and vendor dashboards. Assembling it into a client-ready story takes so long that by the time the report ships, the activity it describes is old news.

This media team was spending three weeks of every month on reports that didn’t reflect their knowledge or expertise, only to start the whole process over again. The real cost was the strategic thinking those hours crowded out.

The goal of the engagement was direct: train AI to do the mechanical work of reporting so the team could use their expertise and judgment to help their clients succeed.

The approach

An AI agent system that runs the monthly report, end to end.

Loop started with discovery. Four collaborative workshops with the agency’s Media, Analytics and Account teams mapped the entire media workflow and pinpointed where the time went. From there, the pilot went deep on a single flagship higher-education account.

The system

Five stages. Human judgment at every one.

Stage 1

Data aggregation

A data-readiness check confirms that data from every campaign and channel is present before anyone drafts a word.

Stage 2

Data confirmation

Every metric is computed by executed code, never estimated by the AI, and cross-checked against account totals. Anything that can’t be verified is kept out of the report.

Stage 3

Insight drafting

The AI drafts insights in the agency’s voice, codified from their real narratives into principles it follows.

Stage 4

Branded report

Trained on the agency’s brand system, the AI delivers a fully editable PowerPoint in their own template.

Stage 5

Human review

The team reviews, edits and approves everything. Nothing ships without human oversight and judgment.

The results

One strategist. One live cycle. Zero errors.

The real test came quickly: a live monthly cycle, with real client data and a real deadline. The question was whether one strategist could take the report from raw data to client-ready, alone.

In that first live cycle, the agency’s VP of Media ran the entire report without the usual back-and-forth with the media team or channel leads. When something looked off, she fixed it conversationally with the AI and re-ran the analysis. When one vendor’s data wasn’t available yet, the system flagged it, completed everything else, and slotted the numbers in when they arrived.

The system also surfaced metrics the team hadn’t typically reported on, and several are being adopted as ongoing KPIs. Every piece of reviewer feedback is written back into the system as a permanent rule, so the output improves with each cycle.

The report was completely generated [by the AI], the nuts-and-bolts work that would have taken the team at least two weeks. It probably saved us 30 to 40 hours a month.

VP of Media

I cross-checked the data, and I didn’t find any errors. And in terms of the way it wrote, it captured our tone. It hit the right chord.

VP of Media

What’s next

From one account to the whole roster.

The agency is now scaling the system: onboarding more client accounts from the documented playbook, extending it to quarterly business reviews, and standing up an internal operating model so the agency owns the system’s evolution. In the words of one senior leader, the pilot left them “light years ahead of where we were before.”

I’m almost viewing this as three separate agents: pulling the data, analyzing it, and presenting it. It’s relatively generalizable. I think we can extend it to other clients fairly easily.

Analytics Lead

The takeaway

A report that arrives earlier, with time to act on it.

The downstream effect is the one clients feel: a report that lands earlier in the month, with more actionable insights and the time to do something with them.

Interested in what this could look like for your agency?

Tell us how your team works today, and we’ll show you where AI moves the needle first.

Talk to us