Most productivity reports drown in speculation, but this survey cuts through the noise with hard data from 1,750 tech workers. Lenny Rachitsky delivers a startling verdict: artificial intelligence isn't just a novelty; it is already overdelivering, saving the average worker half a day every week. For busy leaders trying to decide where to allocate resources, the distinction between a hype cycle and a compounding revolution is the difference between a costly mistake and a strategic advantage.
The Role Divide
Rachitsky's central finding challenges the assumption that AI is a uniform tool for everyone. The data reveals a stark hierarchy of value based on how different roles apply the technology. "Founders are getting the most out of AI," Rachitsky writes, noting that nearly half report saving over six hours a week. This isn't just about typing faster; it's about a fundamental shift in how work is approached. Founders are using AI as a "thought partner and sounding board," leveraging it for strategy and vision rather than just output.
In contrast, product managers and designers are still largely stuck in the production phase. While PMs are using AI to write product requirement documents and create mockups, Rachitsky points out that "AI is helping PMs produce, but it lags in helping them think." The top use cases are downstream tasks like documentation and communication, while upstream strategic work like user research sits near the bottom. This creates a paradox: the people closest to the "what to build" questions are using the tool the least for those very questions.
Founders are treating AI as a thought partner and sounding board, not just a tool for specific deliverables.
Critics might argue that this gap is simply a matter of time and learning curves, but the data suggests a deeper structural issue. The tools available today are optimized for generating content, not for navigating ambiguity. As Rachitsky notes, "Writing a PRD has a clear output; competitive research does not." Until AI can handle the "fuzzy problems" of strategy, the productivity gains will remain unevenly distributed, favoring those who can frame high-level problems over those who execute the answers.
The Engineer's Dilemma
Engineers present the most complex picture in the survey. They have fully accepted AI as a coding partner, yet they report the most mixed results regarding quality. "51% of engineers tell us that AI makes the quality of their work better, but 21% say it's worse," Rachitsky observes. This is the highest "worse" rating of any role, likely because the bar for correctness in code is binary and unforgiving. A slightly flawed paragraph in a memo is a minor annoyance; a buggy function can crash a system.
Despite the quality concerns, the demand for AI to handle the "boring but necessary" work is surging. Engineers want the tool to take over documentation, code review, and testing. Rachitsky highlights a massive demand gap here: while only a small fraction currently use AI for these tasks, the desire to do so is skyrocketing. This aligns with the concept of the "productivity paradox," where initial technology adoption often disrupts workflows before the true efficiency gains are realized. The industry is currently in that awkward middle phase.
Furthermore, the tool landscape for engineers is uniquely volatile. Unlike other roles where ChatGPT dominates, engineers are fragmenting toward specialized tools. "Cursor, ChatGPT, and Claude Code are all within 4 percentage points," Rachitsky writes. This indicates that switching costs are low and that purpose-built tools are winning over generalist chatbots. The market hasn't consolidated, and the winner is still being decided.
The Future of Work
The survey's most compelling insight lies in the gap between current usage and future desire. For almost every role, the biggest opportunity isn't doing what they do now, but doing what they haven't done yet. PMs want to use AI for user research; founders want it for market analysis; engineers want it for testing. Rachitsky frames this as the next wave of adoption: "The next wave of AI adoption will require not just better models but better workflows for human-AI collaboration on fuzzy problems."
This echoes the historical "diffusion of innovations" theory, where early adopters find value in novelty, but mass adoption requires solving the mundane, high-friction parts of a job. The survey suggests we are moving past the novelty phase. "AI has clearly cemented a place as work and productivity infrastructure," Rachitsky asserts. The question is no longer if AI will change work, but whether organizations can restructure their workflows to let AI handle the upstream thinking, not just the downstream typing.
If AI is already giving most people back at least half a day per week in late 2025, what does 2026 look like? What about 2027?
Rachitsky warns that the current models are merely the starting line. Quoting Kevin Weil, he notes, "The AI model that you're using today is the worst AI model you will ever use for the rest of your life." This is a critical reminder for leaders: the productivity gains reported in this survey are likely the floor, not the ceiling. The compounding nature of these improvements means that waiting to adopt is a strategic error.
Bottom Line
Rachitsky's survey provides the most concrete evidence yet that AI is a genuine productivity multiplier, but it also exposes a critical bottleneck: we are using a thinking machine to do production work. The strongest part of the argument is the data showing that founders, who use AI for strategy, are reaping the highest rewards. The biggest vulnerability is the assumption that tools will naturally evolve to handle the "fuzzy" strategic work without intentional workflow redesign. Leaders should watch for the emergence of tools that bridge the gap between generation and reasoning, as that is where the next massive leap in value lies.