Where team output actually leaks
The hours don't disappear in the work. They disappear around it. That's why the multiples are large.
When a team's output disappoints, the default explanations are about people. Not enough of them, not skilled enough, not motivated enough. Those explanations feel right and they're usually wrong. I've spent years tracing workflows end to end, in compliance, trust and safety, and operations, and the hours almost never vanish inside the work itself. They vanish between the work.
Four leaks account for most of it.
1. Handoffs
Every time work moves between people, it stops. It sits in a queue, gets re-explained, gets reformatted for the next person's tools, and loses context that the next person has to rebuild. A workflow with six handoffs can spend more calendar time waiting at the seams than being worked on. Nobody sees this, because everyone is busy the whole time. Handoffs don't show up on anyone's timesheet. They show up in the gap between effort and output.
2. Rework
Checking, re-checking, and fixing things that came through wrong. Rework compounds quietly because every past failure leaves a scar in the process: one more review step, one more approval, one more spreadsheet reconciliation. Each scar made sense when it was added. Years later the process spends more time verifying work than doing it, and the checks apply to everything equally, including the 95% of items that are never wrong.
3. Waiting on judgment
Most work in a skilled team is routine, punctuated by moments that need a real decision. When those moments arrive, the work stops and queues for whoever is allowed to decide. The decision takes four minutes. The queue takes two days. Multiply that across every ambiguous case in a quarter and you find entire salaries' worth of time spent waiting for judgment that, once it arrives, is fast.
4. The bottleneck person
Every team has one: the person who knows how the spreadsheet works, who has the context, who everything routes through. They're usually the best person on the team, which is exactly the problem. Their capacity is the team's ceiling. Work scales until it hits their calendar and then it stops scaling, and because they're heroically busy, the constraint reads as dedication instead of as a design flaw.
The team isn't slow. The structure around the team is slow.
Why AI changes this, specifically
Notice what's not on the list: typing speed, raw effort, intelligence. That's why giving everyone a chatbot produces single-digit percentage gains. It speeds up the work, and the work was never where the hours went.
AI matters because it changes the economics of the four leaks directly. Handoffs shrink when a system carries context between stages instead of a queue. Rework shrinks when verification is automated and aimed where errors actually occur instead of everywhere. Waiting shrinks when routine judgment is encoded and only genuine ambiguity routes to a person. And the bottleneck person stops being a ceiling when the knowledge in their head becomes a system anyone can run, which also frees them to do the work only they can do.
This is where the large multiples come from. I rebuilt a Fortune 500 legal compliance workflow this way: the team's throughput went to twelve times its old rate while quality rose sixteen points. Nobody worked faster. Nobody worked more. The leaks closed. The same mechanism is how ~95% of an audit workflow gets automated while humans keep every judgment call: the volume work and the judgment work get separated, and each goes to the thing that's best at it.
The test
If you want to know whether this applies to your team, don't ask how hard people are working. Pick one piece of work and follow it from request to done, with timestamps. Count the handoffs. Count the hours it spent in queues versus in someone's hands. Find what fraction of the cycle was someone deciding versus someone waiting to decide. The ratio is usually embarrassing, and it's also good news. It means the capacity you want already exists. It's just leaking.