The Remote Labor Index started showing up everywhere last week. LinkedIn first, then Hacker News, then half the remote-work subreddits I lurk in. It scores common remote work tasks on a scale from 0 to 100 based on how automatable they are with current AI tools. Not theoretical AI. Not AGI. Stuff you can install and run right now.

Erik Brynjolfsson has been telling anyone who’ll listen that the AI productivity takeoff is “finally visible in the data.” And this index feels like someone decided to get specific about where exactly that takeoff lands first.

What scores 85 and above

Email triage: 92. Form filling: 89. Basic web research: 87. Calendar scheduling: 85. Data entry from web pages: 83.

If you work remotely, you probably touched at least three of those today. But they’re not edge cases or specialized tasks. They’re the connective tissue of knowledge work, the repetitive overhead sitting between you and whatever you were actually hired to do. And the index assigns lower scores to things requiring judgment calls, relationship nuance, or creative synthesis. Strategic planning: 31. Complex negotiation: 28.

So who’s actually capturing the gains?

Brynjolfsson’s macro data shows real, measurable productivity improvements from AI adoption. But McKinsey’s late-2025 survey pegged the median knowledge worker’s weekly AI time savings at under 30 minutes, which means all those chatbot subscriptions and enterprise AI rollouts are producing roughly enough saved time per person to watch half a sitcom episode without the laugh track.

So something does not add up. Or maybe it does, and the gains are concentrating in the hands of people who automated the high-scoring tasks, while everyone else is still copying text between browser tabs and rebuilding mental context after every alt-tab. The AI productivity paradox was always more of a distribution problem than an existence problem.

Already solved

The tasks scoring 85+ on the Remote Labor Index are not future capabilities we need to wait for. Browser agents handle them today. A browser agent like dassi lives in your browser’s side panel, sees the page you’re on, and acts on it directly. No copying text into a chat window. No switching tabs to paste a response back.

Because the difference between a browser agent and a chatbot for these tasks is the difference between an assistant sitting next to you at your desk and one locked in a room down the hall who only helps if you walk over carrying a printout of what you need.

I watched a demo last month where a browser agent triaged 40 support emails in roughly the time it took the presenter to explain what the tool was doing, processing real customer complaints and shipping questions and account issues, generating contextually appropriate draft responses that a human then approved or tweaked with a single click. And when you look at the seven tasks people still do manually in their browser, they map almost perfectly to the top of this index.

So the bottleneck is not technology. It’s that most people still equate “AI for work” with typing prompts into ChatGPT and pasting the output somewhere else, which is roughly like owning a dishwasher but hand-washing everything and using the machine as a drying rack.

The uncomfortable part

People whose entire job consists of 85+ tasks should be worried, and pretending otherwise is dishonest. If your role is 80% email triage, calendar coordination, and data entry, the math is brutal.

But most remote workers are not in that position. And for most of us, those high-scoring tasks are the crap that keeps us from our real work — the strategic thinking, relationship building, and creative problem-solving that scored below 40. So automating the overhead is a damn relief, not a threat.

Remote work already filtered once for self-direction and digital fluency, and now the index suggests it’s filtering again for whether you’re willing to offload the mechanical parts of your workflow so you can spend your actual hours on things that require a human brain making human judgments in real time.

Quarterly ratchet

The index updates every quarter. And scores across the board will climb. Tasks scoring 60 to 75 right now will cross 85 within a year.

dassi already handles the top of the index from your browser’s side panel, with your own API key and your data staying local. The bottom of the index is safe for a while. But “a while” keeps shrinking.