In 2024, Oxford University Press tried a small change to a familiar ritual. Instead of broadcasting another call for papers, it approached people already reading about the relevant subjects. Two biology journals became the test. The invitation appeared while qualified readers were on the publisher’s own website. A research habit became a clue about a possible future author.
- Hum connects what publishers publish with what their audiences actually read.
- Its Alchemist tools cover content tagging, audience engagement, and manuscript screening.
- The useful lesson: start with a specific decision, then connect the data needed to make it.
Hum supplied the machinery. Its software analyzed content, assigned topic tags, and used reading behavior to assemble audiences. People who had already visited the call-for-papers page were excluded. The point was to introduce an opportunity to someone plausibly interested who had not encountered it. One campaign produced almost 260 clicks; another exceeded 1,300. Those were visits toward submission, rather than a count of published papers.
01 / The person between the platforms
This is an oddly revealing problem for publishing. An organization can know an extraordinary amount about a research field while knowing surprisingly little about the person reading its journals. Subscription records, event registrations, email activity, and website behavior live in different systems. The institution paying the bill may be easier to recognize than the individual doing the reading.
Hum began in 2020 with associations and a customer data platform: software for bringing those scattered records together. Its roots were a strategic partnership with Silverchair, the scholarly publishing technology business. Today its team page lists Dustin Smith as co-founder and CEO and Niall Little as technical co-founder and CTO. Publishing expertise and data engineering sit unusually close together here.
The organization describes a remote-first team spread across nine time zones. Little’s personal motto is “Make it work, then make it work better.” That is a useful description of Hum’s product history. Before the manuscript assistant came the less glamorous business of getting systems to exchange information.
02 / Fewer guesses in the inbox
Consider PCMA, the convention-management association. Before its January 2022 meeting, more email was colliding with pandemic-era digital fatigue. Hum joined CRM records to behavior across PCMA’s platforms. A visit to travel information could prompt one message; interest in health and safety could prompt another, pointing toward the virtual conference.
Hum’s published case study reports a 59% higher click rate for segmented emails than broad sends. It also identified eligible members missing from campaigns. The first thing to falter was the assumption that one list could stand in for many different intentions. The repair began with identifying those intentions.
That sequence helps explain Hum’s place in the market. A general customer data platform can collect events. An email tool can deliver a campaign. Hum’s specialty is linking a publisher’s content to audience interests and putting the resulting intelligence into publishing work. Its buyers include scholarly publishers, societies, and B2B media organizations.
03 / The engine leaves the garage
In June 2023, Hum publicly announced Alchemist, its AI engine inside the data platform. By March 2025, it was offering Alchemist-powered tools separately. Hum’s account of the change is practical: building the CDP had already required content preparation, behavioral data, and predictive capabilities. That infrastructure could serve more than marketing.
Taxonomy organizes a publisher’s backfile and classifies new material. Engage builds behavioral profiles and targets campaigns. Review analyzes manuscripts, checks citations, and assesses journal fit. Hum’s September 2025 suite also named Insight and Search. The shared ambition is to make specialist applications work from connected information, rather than require every publisher to assemble its own AI infrastructure.
This is enterprise software sold through conversations and demonstrations. A buyer’s practical comparison spans general data platforms, editorial screening tools, and internal engineering. Hum’s argument rests on subject expertise and integration: understanding journal workflows is part of the product, alongside understanding the text.
04 / Two returns. One caution.
Karger offers a useful test of the economics. Its author-marketing program combined calls for papers, events, institutional campaigns, and recommendations. Hum’s December 2025 account reported a 4.5x return when counting a broader set of campaign touchpoints. A narrower calculation tied to corresponding authors submitting to the promoted journal returned 1.2x.
Same program. Different attribution rules. These are reported returns, not subscription prices.
The distinction deserves attention. Marketing can influence several steps before submission, and researchers can move between journals. Counting those connections changes the answer. Someone copying the approach should choose an outcome, define which interactions count, test the message, and compare results consistently. Buying the software does not settle the measurement question.
05 / The editor still has a job
Review takes Hum upstream, before publication. Developed with GroundedAI, it supports manuscript assessment rather than making the final editorial decision. A March 2026 announcement expanded its two-way ScholarOne integration. In June, AIP Publishing announced a partnership to bring it into conference-proceedings quality control.
Putting the tool inside an existing submission system matters. An editorial team can examine the manuscript intelligence alongside the work it already does, instead of carrying every paper into a separate experiment. The integration is a distribution advantage for Hum and a workflow advantage for the publisher. Its value still depends on whether the information proves useful when an editor checks it.
There is a tension here. Better targeting asks who deserves an invitation. Better screening asks which submission deserves attention. Both depend on models interpreting evidence, and both need people prepared to question the interpretation.
The valuable connection is the one that helps a person make a better decision.
Hum’s October 2026 recap of a publishing panel confronted the difficult case: legitimate research that passes integrity checks yet contributes little, and unconventional research whose importance emerges later. An automated filter cannot resolve that tension by itself. Likewise, audience tools need usable behavioral signals and permission to act on them. Hum is most interesting where those conditions hold: a publisher with valuable material, scattered information, and a concrete decision waiting to be improved.