A national study of entry-level hiring skips the jobs many Clinton graduates take

Counter service, one of the job categories the study set aside. Photo: Vitaly Gariev / Unsplash

Every study of the labor market has to decide what counts. The one behind much of this summer’s coverage of artificial intelligence and first jobs decided to set three kinds of work aside, and they happen to include the categories that employ the most people in and around Clinton.

The research comes from Coresignal, a workforce data company, and it has been cited steadily since June in national coverage of what AI is doing to junior hiring. It is unusually open about its own method, which is what makes the omission worth reading rather than complaining about.

Researchers gathered 2,748,270 American job postings aimed at entry-level workers and cleaned them down to 2,626,430, removing senior roles that had been mistagged and ads that were off topic. The core year-on-year comparison is narrower still: 346,532 postings from January through May 2025 against 905,658 from the same months of 2026.

That difference in size is not a flaw, and the company addresses it directly. Because far more ads were collected in the second year, it reports every result as a share rather than a count, and it re-ran the analysis using only employers who advertised in both years to check that the pattern was not an artifact of new companies entering the data. The direction held.

Then a study that classified junior job ads by title sorted each posting into one of four groups. The first is AI-exposed, meaning office and knowledge work that current software can already do part of: bookkeepers, paralegals, data entry clerks, junior accountants, administrative assistants. The second is in-person and manual: food service, retail, warehouse, trades, driving. The third and fourth are the ones that matter here. The third is described as generic or ambiguous, and the company says it deliberately left it out. It contains sales titles, customer service titles and advisory titles. The fourth is simply unclassified. Neither appears in the headline findings, which is why the study notes that its two published lines do not add up to 100 percent.

Sales and customer service are not marginal categories in the Jackson metropolitan area. Federal Bureau of Labor Statistics estimates for the metro, which takes in Hinds County and seven others, put sales and related occupations at 9.0 percent of all employment, the third largest occupational group in the region, at a mean wage of $20.41 an hour. Office and administrative support is the largest at 12.2 percent, and transportation and material moving sits between them at 9.8 percent.

So of the three biggest occupational groups around Clinton, one falls squarely inside the study’s AI-exposed category, one falls inside its in-person category, and one was held out of both. None of this makes the findings wrong. Holding out ambiguous titles is a defensible choice, arguably a conservative one, because a category that mixes retail clerks with account executives would blur any result it touched. The company’s reasoning is published alongside the numbers. But it does change what a reader here should take from the coverage.

What the study can support is a statement about composition. Among the junior ads it did classify, office and knowledge roles grew as a share and in-person roles shrank, and that held up when the analysis was restricted to a fixed set of employers. What it cannot support is a claim about the whole entry-level market in a place where a large slice of that market sits in the bucket it excluded.

There is a second limit the company states outright. These are advertisements, not hires. Coresignal cannot tell why an employer posted a role, whether the position was filled, or who filled it. Its own write-up describes the output as patterns in advertised work rather than a census of employment, and warns that small states and small job boards are noisier than large ones.

Mississippi is one of the states it treats carefully. The published state table places it near the bottom, and marks it as a place climbing from a very low starting point rather than one where the shift has taken hold. The company warns against reading a fast percentage rise as evidence that a state has become a hub, since a small number growing quickly remains a small number.

The practical reading is narrower than the headlines suggest. This is good evidence that the mix of advertised junior office work is tilting toward roles where AI tools are already in use. It is not evidence about how many jobs exist. And on the question facing a graduate leaving Mississippi College this spring, whether the sales floor or the service desk is a worse bet than it was a year ago, the study is silent by design.