Now
Andrew Rochford's current work is on the governance of deployed artificial intelligence. As AI systems begin to make consequential decisions at a scale and speed beyond case-by-case human review, the question he is working on is a practical one: how to verify, continuously and at scale, that machine decisions remain equivalent to the standard a qualified human professional would hold, and how to govern, on evidence, which decisions a machine should be permitted to make at all.
In June 2026 he published a research paper setting out an architecture for exactly this, Above-the-Loop Governance: A Statistical Architecture for Verifying and Allocating Decision Authority in Deployed AI Systems. It applies statistical methods proven over more than half a century in clinical research, regulated manufacturing, and population-grade measurement to a problem most current oversight leaves open.
He is Founder and Director of Trakked, an Australian company building oversight infrastructure for enterprise AI systems.
The gap this work addresses
Almost every organisation now deploying AI into a consequential decision faces the same bind. The technology offers efficiencies a competitor will capture if it goes unused, while producing outputs that, if wrong, expose the organisation to regulatory action, harm to the people it serves, and litigation.
The usual answer is to keep a qualified human in the loop, reviewing decisions one at a time. That has been the right approach for its era, and it remains right for many decisions. What has changed is scale. Human review does not keep pace with the volume and speed at which AI now decides, and the cost of doing it properly rises in step with deployment. Removing the human without a defensible substitute is an exposure no serious organisation can accept. Keeping human review at full scale is a cost few can bear.
Beneath that sits a second problem. As AI shifts from recommending to deciding, it holds real authority over outcomes. At present that authority is granted by default, through configuration and gradual expansion, with no continuous check that it remains appropriate or any mechanism to withdraw it when the conditions that justified it no longer hold.
The work treats both as a single question: how to keep qualified human judgement as the standard AI is measured against, while allowing AI to operate at the scale its applications require, and proving, continuously and on evidence, that each decision is held by the party that should hold it.
A thirty-year career across medicine, enterprise and media
Andrew Rochford trained and worked as a doctor in emergency medicine, holding registrar positions at major teaching hospitals including Royal North Shore Hospital in Sydney and Princess Alexandra Hospital in Brisbane. That clinical grounding, making consequential decisions under uncertainty and pressure, runs through everything that followed.
He went on to build and lead in healthcare enterprise. He founded Docta, a telehealth and digital health company that he led as Chief Executive and scaled internationally into India and Indonesia, before it was acquired by the global healthcare provider Aspen Medical. As Chief Executive of Aspen Medical's Indonesian operations, he led the establishment and construction of the group's first hospital, a 200-bed facility in Depok, West Java. He is President Director of PT Aspen Medical Group Indonesia and Chair of the Aspen Medical Hospital Advisory Board.
Alongside medicine and enterprise, he spent close to two decades in public communication of health and science across national television and radio, work built on a single ability: taking complex clinical and scientific material and making it clear, accurate and useful to a non-specialist audience. That same discipline now shapes his research, which is written to be legible to the regulators, executives and professionals who must decide whether to trust it.
Background
Andrew Rochford holds a Bachelor of Medical Science, majoring in Anatomy and Neuroscience, and a Bachelor of Medicine and Bachelor of Surgery with Honours, both from the University of Sydney. His early research was published in the journal Neurosurgery. He is trained in emergency medicine.
He founded Docta, which he led as Chief Executive until its acquisition by Aspen Medical. He is President Director of PT Aspen Medical Group Indonesia and Chair of the Aspen Medical Hospital Advisory Board.
Collaboration and independence
The current research was developed with the support of, and benefited from review by colleagues at, the Centre for AI, Trust and Governance at the University of Sydney. The author is an independent researcher and holds no institutional appointment.
Disclosure
The author founded and serves as Director of Trakked Pty Ltd, which operationalises the architecture described in the paper as a deployable governance system. The evaluation framework set out in the paper is offered as a neutral instrument applicable to every system claiming the category, and the author holds that any system he builds should be tested against it as rigorously as any other. A provisional patent application covering certain methods referred to in the paper was filed in May and updated in June 2026 and is pending. The research received no external funding.
Contact
For research, media or professional enquiries: andrewrochford@gmail.com