■ MODERATE RISK ■ Agriculture & Environment
Not entirely, but the measurable half of grading is already machine work. Optical systems size and color-match roe faster than any human eye; the salting decision and the flavor call still belong to a master.
“AI grades eggs by size and color. Your palate is 'experienced.'”
Our AI replacement risk score — how we score jobs
Grading happens in minutes under pressure. Roe comes out of a sturgeon, gets screened through a mesh to separate eggs from membrane, and is assessed immediately for grain size, uniformity, color, shell strength, and — critically — how it behaves. A grader rolls eggs between fingers to feel firmness, tastes for nuttiness versus muddiness or metallic off-notes, and decides the salt ratio on the spot. Malossol grading means low salt, which means less preservation margin and less room to be wrong. That call sets whether a batch sells as premium or as an also-ran, on a product where the price gap between grades is enormous.
Machine vision handles the objective dimensions well. Automated systems already size eggs, quantify color against reference standards, count uniformity, and reject damaged grains at speeds no human matches, and near-infrared spectroscopy can estimate fat and moisture content non-destructively. Sorting, weighing, tin filling, and traceability documentation are all automated in modern processing houses. Sturgeon farms increasingly use ultrasound and biopsy to time harvest, which is another judgement call moving toward instruments. For high-volume farmed roe of consistent origin, machine grading is largely adequate.
The resistance is flavor and consequence. Texture — the way a grain pops rather than smears — and the taste profile that determines a tin's market tier remain sensory judgements, and the salting decision is irreversible. Instruments can correlate with quality but the buyers, restaurateurs and luxury retailers, still trust a named master grader's palate, and that trust is itself part of the product's story. It is also a tiny, secretive profession where knowledge passes person to person and there is no dataset to speak of. Our 29 reflects a job whose measurable parts are gone and whose reputational core is not.
Automatability: our editorial assessment of current and near-term AI capability
Optical sorting and instrumented quality checks are already installed at scale producers and will be near-universal in farmed operations by the early 2030s, removing the routine sizing and inspection work. What persists is a small cadre of master graders making the salt and tier decisions and lending their names to the product. Expect the profession to shrink toward that apex over the next two decades and stay there — a craft credential attached to a luxury good, not a headcount role.
For size, color, uniformity, and damage, yes — optical sorting outperforms human consistency and runs far faster. Spectroscopy adds fat and moisture estimates without opening a tin. What machines do not do is taste. The texture judgement and the salt decision that separate a premium tier from a mid-grade batch are still made by hand and palate, and they are irreversible once made.
It is a very small, largely inherited profession, and automation is compressing the routine end of it. Entry usually comes through sturgeon aquaculture or luxury food processing rather than through a grading qualification. If you want in, treat grading as one skill inside a broader production role — harvest management, curing, food safety, and export compliance — rather than as a standalone job.
Much of it already has at scale. Optical sorters, automated filling, and digital traceability are standard in large farmed operations, and the trend accelerates through the late 2020s as aquaculture consolidates. The remaining human step sits at the top: deciding cure and tier for premium batches. Expect that to persist as a specialist function long after routine grading is fully mechanized.
Sensory judgement on an irreversible decision, plus reputation. Salting cannot be undone, and getting it wrong destroys a high-value batch, so producers want an accountable expert rather than a threshold in software. Luxury buyers also purchase provenance, and a named grader is part of that. Graders who also understand the instruments and the production chain are the most durable.