AI made drafting cheap, and now review is the bottleneck
AI made drafting cheap, so more drafts get made, and the pile lands on whoever reviews them. The bottleneck moved from writing to review. It eases when drafts are made only where a decision needs one and written rules sort the routine ones.
AI was supposed to save time. The drafts do appear faster. And yet you, or whoever signs things off, are reviewing more than ever, with the same hours in the day.
What owners try
Owners generate more, because drafts are cheap now: more replies, more proposals, more reports. It feels like using the tool properly, and it feeds the pile.
Why the pile grows
When something gets cheaper to make, people make more of it, so the total effort can rise. Cheap AI drafts do exactly that. Writing takes seconds, so more gets written, and every draft still needs a person to read it before it goes out. The pile moves to the reviewer, who has the same hours as before.
The slowest step in a process sets the pace for everything else. With drafting this cheap, that step is review.
Signs review is the bottleneck
- Drafts wait for review for days.
- One person reviews almost everything.
- The pile grows even though drafting is quick.
- People ask for more drafts because they are cheap.
- Routine cases wait in the same line as the ones that need thought.
How to ease the pile
To ease the pile, treat the reviewer as the limit and plan around them. Prepare a draft for each case that needs your judgment, and let written rules prepare the routine ones, so approving them takes a moment.
| What arrives | What the system does | Does a person see it |
|---|---|---|
| A routine case a written rule covers | Sorted by the rule into a short approval list | Yes, as a quick approval |
| An unusual or high stakes case | Drafted and flagged first | Yes, before anything goes out |
| A case the rules do not cover | Held with the question attached | Yes |
Picture two inquiries on the same morning. A client asks for a copy of last month's invoice. A written rule covers that, so the reply is ready on the approval list and approving it takes a moment. Another client asks for a discount in the middle of a project. No rule covers that, so it is flagged first with the question attached and waits for your judgment. The inquiry desk sorts inbound inquiries this way.
See the inquiry desk →Decisions you keep
- Decisions about what counts as routine.
- What gets approved. A person approves everything that goes out.
- Which cases reach you first.
The system behind this: drafts made for the cases that call for a decision, a short approval list for the routine ones, and a person signing off on everything that goes out.
Questions
Why am I reviewing more now that AI writes the drafts?
Because cheap drafts mean more drafts, and each one still needs a person to read it. The work moved from writing to review.
How can I tell it is the review pile holding us back?
Drafts wait days for one person, and the wait grows even though drafting itself is quick.
Will letting the AI write more make it better or worse?
Worse for the reviewer. More drafts into the same review step make the pile longer. Draft only where a decision needs one.
What do I do first?
Look at last week's pile and mark which cases a written rule could have sorted. Those are the first to move into a quick approval list.
Related
- Why the AI tools haven't saved any timeAI tools save little time when they sit beside the old process as an extra step. The time comes back when the process is rebuilt around the AI step.
- The owner-approval bottleneck: when every yes waits for youReplies, proposals and discounts all wait for the owner's yes. Written rules send the routine to a named person and bring the owner only the exceptions.
- A drafted reply to each lead, approved from your phone before it sendsAI drafts replies to lead emails that pass a junk screen, from your business facts. Risky asks get flagged. Nothing sends until a person approves it by phone.
- Where AI belongs in a revenue systemWhich steps from lead to revenue go to AI, which run on plain rules, and which stay with a person. The hard rule: the AI never invents a price.