Forty requests before lunch
Every morning the request emails arrive. A features writer wants a small business owner who changed supplier because of energy costs. A trade title wants comment on new warehouse regulations by 3pm. A podcast producer needs a founder who started a company after fifty. They come from ResponseSource or a similar service, from the #journorequest hashtag, from newsletters and from journalists who email the agency direct.
Someone junior is meant to read them and forward the relevant ones. On a quiet day that works. On a busy day they skim the subject lines, forward a handful, and the request that was perfect for your logistics client sits unread until the next morning, by which time the piece has been filed with someone else's quote.
Why good fits slip through
The difficulty is matching, not reading. The person triaging has to hold every client's topics, spokespeople, restrictions and current campaigns in their head, then compare each request against all of them, fast.
- Requests rarely use the words a client would use. A request for someone on cold chain logistics fits a client who describes itself as a temperature-controlled distributor.
- Client briefs about which topics are allowed sit in separate documents.
- Spokesperson availability changes and the triager does not know who is travelling.
- Deadlines are buried in the text and range from an hour to a fortnight.
- Forwarded requests go to one person who may be in a meeting.
The cost of a late reply
Reactive requests are one of the cheapest ways to win coverage for a client, because the journalist has already decided to write the piece. Missing them means working harder for proactive pitches to get the same result. Clients notice when a competitor keeps appearing in the pieces they expected to be in.
There is a quality problem too. Rushed responses written ten minutes before a deadline are weaker than ones prepared with an hour's notice, and journalists remember the agencies that send useful comment quickly.
How we build the triage feed
What we build takes in every source you use, reads each request once, and routes it to whoever can act.
- Requests are collected from the subscription emails, a monitored inbox for direct requests, and any feeds you follow, into one queue.
- Each request is parsed into structured fields: title, journalist, topic, type of contributor wanted, and deadline converted to a date and time.
- Every client has a short profile: topics, words they use, spokespeople and their areas, and topics that are off limits. The account team maintains it.
- A language model such as OpenAI or Anthropic Claude compares each request with the client profiles and gives a short reason for each suggested match, so people can see why it was suggested.
- Likely matches are sent to the account team's channel in Slack or Microsoft Teams with the deadline at the top, and a button to claim the request.
- Claimed requests show who is responding and whether a response went out, so two teams do not answer the same request for different clients without discussing it.
| Field extracted | Why it matters |
|---|---|
| Deadline | Sorts the queue so the next closing request is on top |
| Contributor wanted | Spokesperson, case study, data or product |
| Title and journalist | Links to your journalist record and history |
| Suggested client and reason | Lets the team judge the match in seconds |
| Off-limits flag | Stops a match that breaks a client's own rules |
The model suggests. People decide. Anything it is unsure about stays in the general queue for a human to read, and nothing is sent to a journalist automatically.
A morning with the feed in place
At 9.10 the logistics account manager gets a Teams message: a trade title wants comment on warehouse regulation changes by 3pm, suggested for your logistics client because their operations director spoke on this topic before. She claims it, checks the spokesperson's diary, and has an approved comment to the journalist by lunch.
The junior who used to triage now reviews the unsure items and the requests nobody claimed, which is a much smaller job, and looks for patterns in what journalists keep asking for.
Recognise this?
- Journalist request emails pile up faster than anyone can read them.
- The right request is often spotted after its deadline.
- Client topic lists and spokesperson areas are not in one place.
- Two teams sometimes answer the same request.
- Clients mention competitors appearing in reactive pieces they missed.