At a glance
Jarrod Kanizay sold UniJobs to TES Global in 2015. Now he runs AcademicJobs.com, selling AI candidate matching into a sector cutting jobs, hiring carefully, and facing new EU rules on recruitment algorithms.
Universities are cutting staff and hiring at the same time, and both things appear on the same job boards. Australia's higher education sector added about 3,300 full-time equivalent positions in 2025, a 2.3 per cent rise, in the same period that Melbourne, Deakin, Western Sydney and UTS each flagged hundreds of potential job losses. The ads keep going up. The pool behind each one has thinned.
That's the market Jarrod Kanizay is walking back into. He founded UniJobs in Melbourne in 2006, sold it to Times Higher Education's then-owner in 2015, and now runs AcademicJobs.com. His pitch this time is narrower than the one he made two decades ago. Not more reach. Better matching.
Reach was the 2015 story. A job board could credibly claim to put a lectureship in front of academics in 40 countries, and more than 700 institutions paid for that. Matching is a different promise: that the eleven people who can actually teach coastal engineering or run a Phase II oncology trial will see the ad, and the other nine hundred won't lose an afternoon to it.
The first exit, and the ownership chain that followed
TES Global bought UniJobs in April 2015 for an undisclosed sum and described the deal as the first step in building a global higher education business. Rob Grimshaw, then TES chief executive, said the ambition was to change how academic and professional support roles get filled, "matching global talent with global universities".
Kanizay stayed on as Asia-Pacific managing director until 2020. The site didn't stay as it was. It was rebranded THEunijobs and travelled with Times Higher Education through two further ownership changes: TPG Capital sold TES Global to Providence Equity Partners in January 2019, and THE was carved out and sold to Inflexion, a transaction that closed at the end of February 2019.
Read that sequence and the second venture starts to make sense. A listings business inside a rankings-and-data company is a small revenue line attached to a much larger brand. The board that began as UniJobs now sits beside THE's data, consultancy and events arms, and its job is to sell subscriptions to institutions that already buy rankings data.
Matching, and the numbers worth questioning
AcademicJobs.com launched in 2023. The product at its centre is a sourcing tool, AI Candidate Connect, which scans public research profiles to find academics who never open a job board. The named sources are Google Scholar, ResearchGate, Scopus and LinkedIn. Kanizay's framing is connecting jobs to people with intent, active and passive seekers alike.
The scale claims need care. The site's own listing page showed 70,235 university jobs worldwide in November 2026, and the company says it carries more than 72,000 live listings across faculty, research, administration and executive roles. THEunijobs listed around 1,200 over the same period. Those are different products carrying different jobs, and neither figure has been audited by anyone outside the business. Treat a listing count as a marketing claim until your own search confirms it.
Inflexion, for its part, set out its growth plan for THE in 2019 as cross-selling to existing clients, winning new ones and launching add-on products, with expansion into the US and Asia.
The second half of the AcademicJobs.com pitch doesn't look like a job board at all. THE Network runs ad-free rooms organised by specialty rather than an engagement feed, borrows the kudos mechanic from fitness apps to mark milestones such as a defended thesis or a submitted grant, and adds collaboration tools for research teams, academic calendars, research news, a crossword and a daily weather forecast. Kanizay's line is that innovation has to be continuous because every other higher ed board is running off-the-shelf software.
The risk with a daily-habit product is that engagement quietly becomes the objective. A crossword doesn't improve a shortlist. What improves a shortlist is density: whether the people you actually need have joined the room. That's field-by-field work, and it's the same cold-start problem Kanizay solved the first time, slowly, by owning one national market before he owned any others.
Photo by Alissa De Leva on Unsplash
The compliance clock nobody priced in
Anyone selling AI matching into Europe is now selling a regulated product. The EU Artificial Intelligence Act (AI Act) lists AI systems used in recruitment and selection as high risk under Annex III, covering tools that target job advertisements, filter applications or evaluate candidates. Transparency duties under Article 50 have applied since 2 August 2026. The heavier obligations, which include risk management, data quality, logging, technical documentation, human oversight and conformity assessment, were deferred to 2 December 2027 under the Digital Omnibus agreed in 2026.
The delay isn't relief. Emotion recognition in the workplace has been banned since February 2025, and AI literacy duties have applied since the same month. Fines for breaching the high-risk obligations reach €15 million or 3 per cent of global annual turnover. Prohibited practices reach €35 million or 7 per cent.
Spain's data protection authority has shown how early this bites. The Agencia Española de Protección de Datos issued a preventive warning to an organisation preparing to deploy an AI tool that screened and evaluated job applications, before the system went live. Employers deploying these tools, rather than the vendors that built them, carry most of the operative duties: retain automated logs for at least six months, tell candidates a high-risk system is in use, and staff the oversight role with people who have the competence and the authority to override an output.
Then there's the part no procurement contract fixes. Reported analysis of Indeed's work with OpenAI put roughly 70 per cent of its sponsored applications through AI ranking rather than keyword search. The figure is Indeed's own and hasn't been independently audited, but the mechanism applies to every matching layer: a search committee sees a shortlist, not the universe it was drawn from, and not the criteria that trimmed it.
A vendor that cannot explain a ranking to a rejected candidate is a liability to the university that bought it. Ask for that explanation during the trial, not after the rejected candidate's lawyer does.
What the sector can actually afford to hire for
The Australian staffing data cuts against the headlines of pure contraction. Total staff full-time equivalent reached 145,882 in 2025, with academic staff up 4.0 per cent to 53,459 and professional staff up 2.6 per cent. Continuing positions rose 7.3 per cent to 86,385. Casual full-time equivalent fell to an estimated 19,402, dropping its share from 14.0 per cent to 13.3 per cent, after Fair Work Act changes in February 2025 gave casuals with six months of regular service the right to ask for permanent work.
Teaching-only roles grew 18.7 per cent to 7,584 full-time equivalent, much of it casual conversions rather than new money. Andrew Norton's analysis of the same figures found casual numbers fell at 13 universities, rose at 14 and barely moved at 15, with institutional swings from a 50 per cent reduction to a 24 per cent increase. One national average hides a lot.
The revenue side is where the constraint sits. New international student commencements were forecast to fall 18 per cent, a hit the assistant minister for international education, Julian Hill, put at about A$2 billion. Melbourne identified about 179 roles for possible abolition from a workforce of 13,800. Deakin projected a revenue decline of A$200 million to A$250 million by 2029, opened voluntary redundancies and now requires vice-chancellor approval for appointments to new or vacant non-academic positions. Charles Sturt plans to cut 101 full-time equivalent roles from February 2027, with 305 staff competing for 204 jobs. Adelaide faces a revenue gap near A$90 million and has offered voluntary early retirement.
Put those together and you get the hiring behaviour every committee recognises. Fewer requisitions. Each one has to be defended to a dean, a finance director and a union. A search that pulls 200 applications used to look like a success; now it looks like a fortnight of reading nobody has time for.
What a hiring team should do this cycle
None of this requires a verdict on which platform wins. It requires a decision about how the next search gets run.
- Test with one vacancy, not one demonstration. Hand over your hardest role and ask for ten qualified names with a one-line justification each. You'll know by Friday whether the pool is real or a repackaged keyword search.
- Rewrite the ad before buying more distribution. Most academic ads are written for the approval committee: every duty, every selection criterion, no answer to what the job is like or why this department. Better copy changes the applicant pool more than an extra thousand impressions does.
- Ask the vendor for the explanation you'd owe a rejected candidate. If they can't produce it for a sample of ten names, the ranking is a black box with a licence fee.
- Measure hires by source, not clicks. AI-referred traffic won't show up as a separate channel on most boards, so track applications, interviews and appointments.
For the candidate side of the desk, the questions differ. If a system surfaced you, ask which system and what it matched on. If you apply and hear nothing, remember that the first filter may have been a model scoring your publication record against a spec written by a committee. It's worth building a CV and cover letter that read well to a machine as well as a professor, and building a network that means someone recommends you before the algorithm does.
Two decades on, the same problem with a new instrument
The first UniJobs grew because Australian universities wanted one place to advertise and one person to ring when the ad didn't work. Personal service was the product. Kanizay's own account of why he came back starts there: call volumes, response times, the sense that centralised support had drifted away from the campuses paying for it.
That diagnosis is easy to make and expensive to prove. There were roughly 70,000 listings on his board and around 1,200 on the one he sold, and the gap says as much about product strategy as about market share. A board optimised for subscription revenue inside a data company keeps its inventory curated. A board optimised for reach lists everything and lets search sort it out. Neither is automatically better for a department trying to hire a senior lecturer in paediatric nursing by August.
What's changed is the cost of the filter. Two decades ago the filter was a keyword field. Now it's a model that decides which of the 70,000 listings a candidate sees, and which of the candidates a committee sees. That shifts where the real hiring decision happens, and it puts a new obligation on institutions that used to just buy an ad.
The test that costs nothing
Reach without a shortlist is noise. Any university running a search this term can check that in a week: one vacancy, ten names, one line each on why. If the tool can't do it, it's a listing site with a new label. If it can, the conversation about price gets much shorter.
- academic-hiring
- higher-education-recruitment
- academicjobs.com
- academic-job-boards
- ai-recruitment

