■ MODERATE RISK ■ Public Service & Government
AI will write the gazetteer entries and scan the archives; it will not stand in a planning hearing and persuade a council that a Victorian warehouse matters more than forty apartments. Heritage consulting automates its paperwork and keeps its politics.
“AI catalogs historic buildings. But arguing to save them from developers needs a human.”
Our AI replacement risk score — how we score jobs
Heritage consultants sit at the collision point of history and development. The work: researching a site's historical significance through archives, maps, and fabric analysis; writing heritage impact assessments and conservation management plans; advising developers on what a planning authority will accept; negotiating between owners, councils, conservation officers, and community groups; and sometimes testifying at appeals when the negotiation fails. It's scholarship deployed as advocacy, billed by the hour, usually against a deadline set by someone's construction financing.
The research substrate is automating quickly. AI transcribes and searches digitized archives that once demanded days in a records office; image recognition dates architectural features and matches historic photographs; LiDAR, photogrammetry, and drone surveys document fabric in hours; and language models draft the boilerplate sections of assessments — site descriptions, policy summaries, statutory context — that consume most billable time. Heritage databases enriched by machine learning can flag significance patterns across building stock at a scale no human surveyor matches. The honest consequence: report-writing hours, the industry's bread and butter, are compressing.
But significance is an argument, not a data point. Deciding whether a building's alteration harms its heritage values, and persuading a skeptical planning committee, developer, or inspector of that judgment, is interpretive, contested, and jurisdiction-specific — exactly the work courts and planning systems require a qualified human to stand behind. Site visits still surprise: the archive says 1890s, the roof structure says older, and no model climbed into the loft. Relationships carry the practice — conservation officers trust consultants whose judgment has proven sound, and developers hire the ones who can land a consent. Our risk score of 40 reflects a profession whose research and drafting labor shrinks substantially while its advocacy, judgment, and negotiation core — the part clients actually pay premiums for — stays human.
Automatability: our editorial assessment of current and near-term AI capability
Research and drafting automation is arriving now — expect AI-assisted archival work and report generation to be standard practice by the late 2020s, squeezing junior researcher roles and fixed-fee report work first. Through the 2030s the profession consolidates around judgment, negotiation, and expert-witness functions. As long as planning systems require accountable human expertise — and heritage disputes stay political — the core role persists past 2040.
It's replacing their research hours, not their role. Archive transcription, feature dating, survey documentation, and assessment boilerplate are automating fast, which shrinks the drafting work that filled timesheets. What remains — judging significance, negotiating with planners and developers, standing behind an opinion at appeal — is contested, jurisdictional, and legally human. Smaller research teams, same essential expert at the center.
Stable at the judgment end, precarious at the report-mill end. Demand tracks development activity and heritage regulation, both persistent. But entry-level positions built on archival grunt work and boilerplate drafting are exactly what AI compresses, so the career ladder's bottom rungs are thinning. Aim early for site experience, planning-system fluency, and hearing exposure — the skills that convert research into defended opinions.
The evidence base is going digital and abundant: searchable transcribed archives, drone and LiDAR fabric records, image models matching historic photographs, and drafted policy sections on demand. Assessments that took weeks now take days. The unchanged part is interpretive — whether change harms significance and what a planning authority should accept — plus the site visit, where buildings routinely contradict their paperwork.
Three clusters: planning-system expertise deep enough to negotiate and testify credibly, since accountability can't be delegated to a model; digital documentation and AI-research fluency, so you direct the new tools rather than compete with them; and relationship capital with conservation officers, developers, and communities. The consultant who pairs automated research with trusted judgment wins the work the automation creates.