A Four-Agent Workflow Designed for Media Responses

ications reach out about operations, earnings, energy policy, and breaking news. Every inquiry needs a response. Most need one fast.

The process for handling them was almost entirely manual. Inquiries landed in a shared inbox. Someone read each one, figured out the topic, decided who should see it, and forwarded it along. Triage was subjective — two people could rank the same inquiry differently. Deadlines lived on spreadsheets. Approvals required chasing people through email chains and hallway conversations.

The result: an average response time of 28 days. In a news cycle that moves in minutes, 28 days isn’t a slow response. It’s silence. And silence speaks louder than any statement a communications team can write.

The team wasn’t slow because they didn’t care. They were slow because every step was a bottleneck. Manual triage. Manual drafting. Manual routing. Manual follow-up. No one had visibility into where an inquiry stood at any given moment. By the time a response was approved, the story had already been written — without them.

Four AI agents, one engine, zero guesswork

The company’s MCCD deployed an AI-powered media response engine built by CloudSufi on top of their existing communications infrastructure. The engine runs on Vertex AI within the company’s own GCP environment — all processing stays inside its cloud, with no external API calls. Four AI agents, each responsible for one stage of the workflow, work as a single system.

The intake agent classifies and prioritizes every incoming inquiry the moment it arrives. It reads the message, identifies the topic, assesses urgency, and tags it — all in seconds. No more subjective triage. No more guessing which inquiries matter most.

The draft agent generates a brand-aligned response draft using the company’s approved messaging library. It doesn’t send anything on its own. It gives the communications team a starting point that’s already on-brand, already factual, and ready for human review.

The routing agent identifies the right approvers based on topic, region, and sensitivity. It sends the draft directly to the people who need to sign off — no forwarding chains, no guessing. Internal sign-off dropped from an hour to 15 minutes.

The follow-up agent tracks deadlines and escalates stalled items. If an approval sits untouched, the system nudges. If a deadline is approaching, it flags. Nothing falls through the cracks.

CloudSufi built the engine and integrated it with the company’s existing tools — including Formstack for journalist inquiry intake and CloudSQL for audit trails. The system didn’t replace the communications team. It removed the manual work that was keeping them from doing what they’re actually good at — shaping the message, building relationships, and thinking ahead.

From 28 days to 12 — and the silence is gone

The numbers tell a clear story. First-touch acknowledgment went from minutes to seconds. Internal sign-off that used to take an hour now closes in a quarter of that time. And the full cycle — from the moment an inquiry arrives to an approved response going out — shrank from nearly a month to under two weeks.

But the real change wasn’t in the numbers. It was in what the team could do with their time. Before the engine, most of the communications staff’s hours went to logistics: reading, sorting, forwarding, drafting from scratch, chasing approvals, checking spreadsheets. That work is now handled by the system. The team spends its time on strategic messaging, media relationship building, and proactive outreach.

Visibility also changed. Before, no one knew where an inquiry stood unless they asked. Now, the entire team can see the status of every active inquiry — who’s reviewing it, what stage it’s in, and whether it’s on track. That transparency alone cut the number of internal “where is this?” emails by a wide margin.

What’s next: from reactive to predictive

The engine was built to grow. The next phase expands into proactive media monitoring — tracking coverage across outlets and social channels in real time, so the team knows what’s being said before a journalist calls.

Sentiment tracking will flag shifts in tone and topic across the media landscape. And predictive crisis detection will surface early warning signals — unusual spikes in coverage, emerging negative themes, or questions that signal a bigger story forming — so the team can get ahead of it instead of reacting after the fact.

The goal isn’t just faster responses. It’s a communications function that sees around corners.