Honeyjar is an AI workspace for press and communications, built by two longtime PR professionals, Michelle Masek and Nadia Jamshidi. It builds media lists, writes announcements and pitches in the customer's voice, scans news and social media, and tracks coverage. The customer is the one who presses send. It is in early testing with companies including Nextdoor, and its waitlist filled with 10,000 names in the first day.
The mechanical steps are already cheap
Finding reporters by topic is not new. Muck Rack, the long-used database of journalists, added an AI agent in February 2026 that suggests which reporters to add to a pitch, and any chatbot can draft the email. What Honeyjar adds is an attempt to build in the judgment of senior PR people, including the judgment to tell a customer to hold off. Its founders are training it on the instincts of other experts for that. This is the one step in the chain that was still hard to buy.
A free pitch makes a bad pitch expensive for everyone
When sending costs nothing, the sender's own filter is the only one left. Reporters have their own, and it is harsh. Yet pitches still work: most journalists say they have found at least some stories through one. So the door stays open for a pitch that fits, while the inbox fills with ones that do not. Cheap drafting makes more of the second kind.
For example, a school superintendent has launched a new reading program. Left alone, she asks a chatbot for a press release and sends it to every education reporter in the region, perhaps sixty names. A tool with real judgment would ask what changed for families. It might tell her that a program launch is not news, but a board vote on cutting bus routes reaches every household. It would then suggest two sentences to the two reporters who covered the last vote. She sends two emails instead of sixty, and those two have a good chance of being read.
The same holds wherever sending is free
Any work that starts with an unsolicited message has this problem now: a recruiter's first note to a candidate, a sales email to a stranger, a letter asking a local business for sponsorship. AI makes the draft free in all of them, and the recipient's patience was never free. So tools in each of these areas will end up competing on what they refuse to send, and the first question to ask of any of them is whether it ever says no.
If you handle announcements for your business on top of another job, you can ask an assistant three things before it drafts anything: who would care about this, what changed for them, and why now. If the answers come out thin, wait.