Why digital productivity tools make you busier, not better

Executive overview

Adding tools to a workflow speeds up individual steps — but if those steps aren't the bottleneck, nothing improves. Eliyahu Goldratt's theory of constraints identifies the single slowest step in any system as the factor that governs total output; everything else just piles up behind it.

For knowledge workers, the bottleneck is rarely communication speed or information flow. It's cognitive capacity — the irreplaceable thinking, deciding, or creating that only you can do. Tools that automate or accelerate pre-bottleneck steps create the illusion of productivity while making the actual constraint worse.

Identifying your bottleneck — and protecting it — matters more than making everything else faster.

The theory of constraints

• Every system has a single slowest step; throughput is capped at its rate regardless of how fast everything else runs
• Goldratt discovered this while optimising a chicken coop assembly line: moving one worker to the slow step tripled output
• The same principle applies whether the system is a factory, a swimming training plan, or a writing process
• Goldratt's one-word summary of the entire theory: "focus" — the bottleneck shows you where to direct effort
• Common mistake: improving a fast step feels productive but only increases the pile-up at the real constraint

From assembly lines to knowledge work

• The Broad Institute applied theory of constraints to a genetic sequencing lab: switching from a push to a pull system eliminated the chaos of work piling up at slow steps
• The same team then mapped all active projects on Post-It notes and immediately saw they had far more in process than they could complete
• They created a limited-capacity funnel: nothing new enters until something exits — "stop starting, start finishing"
• A custom gearbox manufacturer had a 15-person design office switching tasks 50+ times a day; enforcing one design at a time cut delivery time from 12 months to 2
• Fewer things in process consistently leads to more things actually completed

Why digital tools often backfire

• Most productivity tools — Slack, Notion, AI assistants — speed up information flow and pre-bottleneck steps
• If the bottleneck is human review, approval, or deep thinking, faster upstream delivery just creates a larger queue
• David Epstein's own experience: elaborate Notion/Slack workflows for YouTube video production moved content fast, but it all stacked up waiting for his approval; adding a fact-checker targeted the actual constraint and unblocked the pipeline
• Cal Newport's podcast workflow deliberately keeps him away from computers; the bottleneck is his thinking and recording time, so every process decision is evaluated on whether it protects that time
• Research on programmer productivity with AI tools shows the same pattern: agentic code generation makes it easy to have 13 things in progress simultaneously, log-jamming the whole system

Multitasking and the stress tax

• Psychologist Gloria Mark's research: the more task-switching during the day, the lower end-of-day productivity and the higher stress
• Thermal imaging and heart rate variability studies show measurable stress spikes when email inboxes open
• The feeling of busyness from toggling is not correlated with actual output
• Multitasking isn't a response to having too much work — it's often a response to not having identified what the real constraint is

AI and the non-bottleneck trap

• AI tools are most easily applied to low-value, automatable steps — rarely the actual bottleneck
• In academic research, access to the right data takes months; writing the paper takes days. Speeding up plot generation is a minor convenience, not a productivity lever
• The real question is not "can this tool make something faster?" but "is what it makes faster the thing that limits our output?"
• A consultant working with companies on AI adoption observed: it's never been easier to do too much — companies implement AI broadly, generate "work slop" (high volume, mediocre output), and create new problems to manage
• The pattern mirrors email in the 2000s: a tool that made something genuinely faster but degraded the whole system

Applying constraints thinking to writing

• David Epstein's bottleneck in his first two books was organisation of information — a digressive research process meant he wrote 150% of a book and cut it back
• For his third book, he spent the first year only researching and interviewing, then distilled 100,000 words of notes into a single-page outline at a Franciscan monastery before writing a word
• Result: he finished the book early for the first time; the book is 20% shorter and no chapters were cut
• The constraint was the hard thinking required to decide what the book was about — everything else flowed once that was resolved
• Front-loading research is efficient because every interview and reading produces usable material; overwriting burns large amounts of effort on sentences that will be cut
• Tony Fadell used the same logic at Apple: forcing the Nest team to prototype the product box before the product existed made them decide what mattered — ultra-constraints slow you down but make execution much faster

Practical rules that target bottlenecks

• Stop starting, start finishing: limit work in process so the constraint isn't overwhelmed
• Identify what only you can do; design processes that maximise the percentage of time spent doing it
• Make all current commitments visible — teams that do this almost always discover redundancies and hidden work
• Do not evaluate a tool in isolation; evaluate it in the context of the whole process and where the actual constraint is
• "If you don't waste a few hours, you'll end up wasting a few years" — time spent thinking about the bottleneck pays back many times over