Rob Seamans

A few high-level notes on the paper: 1. We ask where the marginal value of work AI cannot yet do may rise as AI speeds up the rest of the job. - Such work can become a bottleneck, especially where more of the surrounding tasks in a job can be done by AI. We call this "bottleneck pressure." 2. Using a task-based framework, we show that for an AI-infeasible task, this pressure is simply the exposure of the occupation posting it. - So you get a new, theory-derived measure that potentially shows where AI could raise the marginal value of work it cannot yet perform. 3. The magic happens at the task level. The same task has different bottleneck pressure in different jobs. 4. To take this measure to the data, we map 288 million U.S. job postings from 2018 to mid-2026 to O*NET tasks and use their GPT-4 feasibility ratings from Eloundou, @sj_manning, Mishkin, and @danielrock's "GPTs are GPTs" (2024). 5. O*NET can be a very coarse proxy for advertised work. - About 90% of the task mentions matched from postings fall outside the posting occupation's own O*NET list, although those statements themselves come from O*NET. 6. Indeed, about 53% of mention-weighted variation in pressure on AI-infeasible work lies within tasks, across occupations. 7. In every year, AI-infeasible tasks within occupations faced more pressure, on average, in senior than junior postings, about 6.5 percent more in the first half of 2026. 8. Over the full sample, mean occupation pressure rose from 17.6 to 18.7, or by 6 percent, about 75% from task mentions shifting toward higher-pressure occupations. (Thank you to @robseamans for pushing me to do 7 and 8!) Overall, reporting bottleneck pressure alongside exposure shows where improving the work AI cannot do may pay off more as capabilities expand.

Delighted to see the engagement by @RonnieChatterji on this new measure to capture "bottleneck pressure" in tasks and jobs. He and researchers including @Alex_M_Richmond, @daveholtz, @sonnytambe are regularly doing some of the coolest and most novel work in the space of AI and organizational economics. I would really like to start emphasizing here the amazing work done by so many people at The Burning Glass Institute. I'm hesitant to mention names because I worry I will miss someone :) None of this would be possible without their diligence and care. Doing the kind of mapping we employ in the paper is a huge lift and I am amazed to see the care and expertise with which this has been done. Our paper starts off with a somewhat counterintuitive framing: how can AI raise the value of something it cannot even do? The answer lies in complementarities! (check out @lugaricano's book on "Messy Jobs" too, as you will find echoes of this theme there as well!) The same task can have very different pressure depending on tasks that surround it (Figure 1 from the paper attached). And, how bottlenecked a task is rises in proportion to how much the rest of the job is exposed to AI. There are several advantages that immediately become clear from the detailed mapping we leverage: (1) First, we can measure bottleneck pressure at the task level and aggregate up to jobs (2) Second, we find that about 90 percent of labeled matched task mentions lie outside the posting occupation’s own O*NET task list. This really makes the postings to task mapping a powerful tool to go beyond what could be achieved by just the O*NET catalog. Even leaving aside the new work and tasks that AI is creating (which is important to study, and we are!), just the occupation-specific catalog itself misses so much of what the jobs have been asking for. As AI permeates work, I am hopeful that economists and AI researchers will find this measure useful to report alongside exposure measures, as together they can show: (1) the scope for AI to perform work and (2) where improving the work it cannot yet do may pay off more as AI capabilities expand. PS - The paper ranks many tasks as well as all occupations on the O*NET catalog by this "bottleneck pressure" index. Please check it out!