Case Study

    Hundreds of Journalist Requests a Day. Zero Inbox Monitoring.

    How a UK PR agency replaced manual triage with an autonomous platform that only calls the team in when a human is needed

    PR / Digital Marketing Agency · ~20 employees · 3-person PR team handling client coverage·AI transformation·45-day implementation sprint

    Hundreds of journalist requests a day. All eyeballs, all the time.

    100s/day

    Journalist source requests landing in shared inboxes

    11.25 hrs/wk

    Operator triage time (45 min × 3 ops × 5 days)

    Minutes

    Window before journalists move on if no one responds

    0%

    Systematic matching of requests to clients or warm leads

    Journalists fire off hundreds of source requests every day. Miss the window, miss the placement. So three operators sat scrolling shared inboxes twice a day, eyeballing each request, deciding whether anyone in the client book could quote. When we asked the team to describe the workflow so we could help, the answers came back differently each time, the tell that nobody had ever mapped the process. That's where bad AI transformations begin, and it's why we map it ourselves before we touch the tooling.

    The Lean AI Method

    1. 1. Discovered

      Sat alongside the PR team and reconstructed the request workflow in the open. 80% triage, 20% craft.

    2. 2. Designed

      A data pipeline behind the shared inboxes, triage logic prototyped in Excel, and an AI layer trained on the team's own pitch voice. The system takes the boring 80%, the team owns the craft.

    3. 3. Deployed

      A full PR platform: live ingestion, auto-categorisation, client and lead matching, drafted research, one-click push to CRM, analytics. Production from day one.

    4. 4. Scaled

      Runs against every journalist request, all day. Operators get pinged only when a human is genuinely needed. Inbox monitoring as a scheduled activity, gone.

    Why process comes first

    We mapped the workflow ourselves before touching the tooling. When a team describes a process differently every time, that's the tell that nobody has mapped it — and that's where bad AI transformations begin.

    What we built

    • Live inbox listener

      Watches shared inboxes continuously. No scheduled scrolling, no missed response windows.

    • Auto-categorisation

      Each request tagged by topic, urgency, and outlet before a human ever sees it.

    • Client + lead matching

      Cross-references every request against the live client book and active prospects.

    • Brief & draft generator

      Researches the journalist, researches the topic request, ready for review.

    • Push-to-client workflow

      One click to send the opportunity to the CRM for further action.

    • Data trail for sales

      Every match, every pitch, every win becomes proof in the next new-business deck.

    Before vs after

    BeforeAfter
    Inbox monitoring45 min × 3 × 2/day0
    Operator triage time11.25 hrs/wkReclaimed
    Requests reviewedManual sampleAll of them
    Lead-match opportunitiesMostly missedSystematic

    An autonomous PR engine that only interrupts the team when a human is needed.

    0

    Scheduled inbox-monitoring sessions

    11.25 hrs/wk

    Operator capacity returned to client and craft work

    All

    Inbound requests handled end to end by the system

    0

    Opportunities lost to slow response time

    Three operators redeployed from triage to craft. A live data layer that doubles as proof in every new-business pitch. Zero placements lost to a slow inbox.

    “The foundation of any AI transformation work lies on a value stream map. Once you see how the value is generated, the technology is easy.”

    Said Lopez — Founder, LeverageAI

    Sound familiar?

    • Does someone on your team spend hours every week scrolling inboxes a system could watch instead?
    • Are opportunities slipping because the response window closed before the right human saw the request?
    • When you ask your team to describe their workflow, do the answers go vague the closer they get to automation?
    • Are you sitting on high-signal inbound data that nobody is mining for client and prospect matches?