The two mistakes qualified professionals make when they enter this world

Two mistakes come up again and again in the first months of a professional entering expert AI work. Both are entirely fixable. Both are worth naming plainly before you make them. Neither is about the technical work itself.

Mistake one: applying below your level

The first mistake is applying for roles that ask for less experience than you have. It sounds absurd. It is common.

The mechanism is straightforward. A senior specialist looks at a role labeled "expert annotator" or "domain evaluator" and reads it as an entry-level position. They apply cautiously, undersell their qualifications in the resume, and get placed at a rate calibrated for someone with less depth than they have. They then spend the next three months doing work well below what they are capable of, at a rate below what their expertise should command.

The fix is to be direct about your seniority in your application materials and to apply for roles that specify the level of expertise you actually have. The catalog usually includes several tiers in each discipline. A consultant physician does not need to apply for a role that could be done by a resident. A partner-level attorney does not need to accept an entry-level legal-annotation position. The higher-tier roles are there, they pay more, and you are qualified for them.

If in doubt, apply for the more senior role. Labs prefer to place a strong candidate one level above their listing than to place a weaker candidate at the listed level. That is the entire economics of expert work at the frontier: capability is the scarce input, not credentials.

Mistake two: treating it like a second job

The second mistake is treating expert AI work like a second job. That is, sitting down for four-hour sessions with the assumption that the more time you put in, the better you will do.

This is wrong. Expert AI work is the opposite. It rewards short, high-quality sessions from a well-rested mind. It punishes long, low-quality sessions from someone trying to grind through volume.

The reason is that the work is cognitively dense. Every item you are asked to evaluate demands the full attention of a working professional. A four-hour session at the end of a clinical day, or after a courtroom appearance, produces work an evaluator would be embarrassed by if they could see it side-by-side with their fresh output.

The professionals who do this work well tend to have a pattern. Two-hour block. Early in the morning, or on a weekend, when they are rested. Coffee. No calendar interruptions. They do focused work, submit, and stop. Two of those blocks a week is often enough to add meaningful income without any degradation of quality.

The professionals who burn out do the opposite. Six-hour weekend sessions. Late-night grinding. Volume over quality. Rate reductions when the quality reviewer flags too many rushed responses. Then dropping out entirely.

The counterintuitive move is to do less, but do it well. Rate stays high. Quality reviewer sees consistent output. Placement expands to more roles and better rates. That is the whole shape.

What this means for how you set up

If you are entering this work, plan around these two mistakes explicitly.

First, apply for roles at your actual level of expertise, not one below. Read the role specification carefully. If it asks for a senior specialist, and you are one, apply. If it asks for a resident, and you are a consultant, skip it. The right role for a senior expert exists in the catalog.

Second, calendar the work as high-quality short blocks, not as a second job. Two hours, twice a week, well rested, is a better setup than one six-hour weekend session. Your quality reviewer's read of your work matters far more than the total hours you log.

Both mistakes are entirely fixable. Both are worth avoiding. Neither is about the technical work itself, which is why professionals doing this for the first time often do not think about them until they have already made them.

If you are just starting, browse the current catalog, pick a role at your actual level, and calendar it deliberately.