The promise embedded in almost every AI productivity tool is simple: do more in less time, and reclaim the hours you used to lose to repetitive work. What that promise tends to leave out is the cost of getting there, and what happens to the time once it is saved. A first-person account from a neurodivergent entrepreneur building an AI-powered content pipeline for a podcast network offers a rare, granular look at what that gap actually feels like from the inside.
The Hidden Labor of Making AI Work
Building an automated social media pipeline for more than 25 shows is not a small undertaking. The process described in the source involved connecting multiple tools, including Claude Code and NotebookLM, into a single workflow that could feed content to YouTube, Instagram, and TikTok simultaneously. The setup took roughly a week of concentrated effort. Once running, it compressed a full day of manual work into a background process.
That sounds like a clear win. The problem is that it did not stay running. When the system broke, rebuilding it from scratch consumed the time that had been saved. A second agent, designed to automate press releases for new show launches, had to be paused mid-build. The work session stretched to 3 a.m. before running out of credits.
This is what most coverage of AI productivity misses. The hours saved by a working automation are real. So are the hours spent building, debugging, and rebuilding that automation. Neither side of that ledger typically appears in a dashboard metric.
The Automation Tax and the 79% Problem
There is a useful name for this dynamic: the Automation Tax. It refers to all the extra work required to make an AI tool function as advertised, plus the implicit expectation that because output is now faster, every available working hour should be filled with more output. The speed increases. The workload does not decrease.
This matters in general, but it carries particular weight for neurodivergent entrepreneurs. A report published in June by the U.K.’s Lilac Centre, based on a national survey of more than 600 neurodivergent entrepreneurs, found that 76% started their businesses specifically to create working conditions that suited them. Yet 79% report workload management or burnout challenges as a result of running those businesses. The same report describes AI and digital tools as “accessibility scaffolding,” a framing that captures something important: for people who face executive-function barriers, the value of AI is not purely about speed. It is about being able to start tasks at all, to convert scattered thinking into structured output, to get past cognitive walls that have nothing to do with intelligence or effort.
The tension is that the scaffolding itself requires maintenance. Learning where each tool tends to fail, knowing what to double-check, managing the connections between systems: none of that shows up in an “hours saved per post” calculation. The metric looks clean. The underlying reality is not.
What the Stopwatch Cannot Measure
Business culture has, as the source puts it, one instrument for measuring AI: the stopwatch. Minutes saved, output multiplied, projects completed. Every productivity dashboard asks how fast the organization is moving. Almost none ask whether the people inside it are working less.
This is a structural blind spot. When AI compresses the time needed to produce one newsletter, the rational response under most workplace logics is to produce five. Output goes up. So does the workload, because five newsletters still require fact-checking, scheduling, and accountability if something in them is wrong. The volume of decisions, reviews, and potential errors scales with the volume of output. The human in the loop does not disappear. The loop just gets bigger.
The more useful question is not “how much faster can this tool make me?” but a set of three: What exact task disappears? What new work shows up in its place? And does the total number of things to manage actually decrease, or does it just change shape?
A small tool that removes the dread of staring at a blank screen before a deadline can deliver genuine relief even if it only saves five minutes. A tool that multiplies output fivefold may increase the cognitive load of managing that output by an equivalent factor. The distinction matters, and it is one that aggregate productivity metrics are not designed to capture.
In Short
AI tools can genuinely reduce friction, remove specific tasks, and help people work in ways that suit them better. What they do not automatically do is reduce the total volume of work. The Automation Tax is real: building, maintaining, and managing AI systems takes time, and faster output tends to fill available hours rather than free them. The most honest question to ask before adopting any AI tool is not how much it can produce, but whether it leaves fewer things to manage at the end of the day, or simply different ones.
Based on reporting from Fast Company - Work Life.