■ HIGH RISK ■ Legal
The desk half of investigation — records, databases, social media, background checks — is being automated into a subscription product. The field half endures: courts still want a licensed human who personally witnessed the subject carrying the couch he claims he can't lift.
“AI scans social media better than your stakeout ever could.”
Our AI replacement risk score — how we score jobs
Modern PI work is mostly research with occasional loitering. The caseload runs: insurance-fraud surveillance (filming the 'disabled' claimant doing CrossFit), background investigations and due diligence, skip tracing debtors and witnesses, domestic and custody cases, process serving, and corporate work — employee theft, IP leaks, pre-litigation fact-finding. Deliverables are reports and evidence packages that must survive a lawyer's scrutiny and, sometimes, cross-examination.
The research layer is automating at speed. OSINT platforms aggregate public records, court filings, property data, and social media into minutes what once took a week of courthouse visits; facial recognition and AI photo analysis connect identities across platforms; skip-tracing databases score likely addresses algorithmically. A background check that justified billable days is now largely a query. Clients know it too — law firms and insurers increasingly run first-pass research on their own subscriptions and hire investigators only for what the databases can't reach. The desk-work revenue that sustained many one-person agencies is compressing hard.
What the databases can't reach is the defensible core. Surveillance still means a human parked down the street, because video evidence needs a testifying witness and pretexting a subject requires improvisation. Interviews — getting a reluctant neighbor or a nervous ex-employee to talk — are craft. Licensing law keeps investigation a regulated human profession, and courts weigh sworn human testimony in ways they don't weigh a scraper's output. Privacy law adds friction to full automation of people-tracking, arguably protecting the licensed middleman. Our 56: the profession keeps its field boots and its witness stand, but loses much of the billable desk time in between.
Automatability: our editorial assessment of current and near-term AI capability
The research compression is happening now — OSINT platforms and skip-tracing databases have already shortened what clients will pay days for, and AI aggregation keeps improving. Expect desk-research revenue to keep shrinking through around 2030, while surveillance, interviews, and testimony hold as the licensed human core. Solo practices built on background checks feel it first; litigation-support and field-heavy shops feel it least.
It's making the desk research obsolete — backgrounds, records, social media sweeps, and skip tracing are increasingly automated products clients can buy directly. The field trade survives: surveillance needs a witness who can testify, interviews need human craft, and the profession is licensed by law. PIs who moved up from lookup work to fieldwork and analysis are doing fine.
Anything requiring presence or persuasion: sitting on a subject for eight hours and filming the money shot, knocking on doors until a neighbor talks, serving papers on someone dodging them, and taking the stand to defend what you saw. Courts credit sworn human testimony, not scraper output — that legal reality is the profession's backbone.
Yes, but enter it as a field-and-litigation professional, not a database jockey. The billable hours that used to come from research are compressing, so build surveillance skill, interviewing craft, courtroom credibility, and a specialty like insurance fraud or corporate work. Also become excellent with the AI tools — they're your leverage, not just your competition.
Radically, and mostly before the current AI wave: online records, social media, and skip-tracing databases replaced the courthouse crawl years ago, and AI aggregation is now finishing that job. Simultaneously, tech created new casework — digital infidelity, online fraud, cyber-adjacent corporate investigations. The job description rewrote itself once already; it's rewriting again.