It does take two minutes. Then the same question comes back from a different client on Thursday, and you spend two more.
Some answers you have typed many times. How long onboarding takes. What happens if a client misses a call. Where the invoice link lives. The answer sits in you and nowhere else.
That is Unextracted Intelligence.
Slow replies have other causes too, like high volume and no triage. This post covers the one where the answer lives only in your head.
If you answer the same question fewer than five times a month, a saved-replies folder in your email is enough. It costs nothing. Write the answer once and paste it when the question comes.
A system earns its setup time in three cases. The answer changes by client. The reply needs a fact looked up first. Or a question arrives that your folder does not cover, and you want it to ask you instead of guessing.
This example is invented to show the pattern. It is not a client story.
An agency owner earns about $200,000 a year. She has one part-time helper. She asks AI to read her last 30 days of sent email and pull out every question a client asked. It finds 38 questions and sorts them into nine groups.
She reads the groups herself, because a quick sort can miss detail. Two groups hold questions that look alike but need different answers. “Can I pause my contract?” means one thing for a monthly client and another for an annual client. She splits them, which gives her 11 answers to write.
She writes each answer in her own voice. The drafts will sound like whatever she gives the system. That takes her an afternoon. After that, the system drafts a reply when one of those questions arrives. She reads each draft and sends it.
Sooner or later a question arrives that fits none of the groups. The system does not guess. It asks her.
She answers by voice note. The system saves that answer and uses it next time. In this made-up example, week four feels different from week one for that reason. The inbox shrinks because she taught it, one answer at a time.
Prices change. Policies change. Keep every answer in one document with a date on it. The owner reviews that document every quarter, and on any day a price or policy changes. A stale answer sent with confidence does more harm than a slow one.
Read the first 20 drafts next to what you would have written yourself. Fix the gaps in the answers, not in the drafts. Every reply also names a person the client can reach, so an annoyed client is never stuck with a machine.
If your business sells personal attention in every message, skip this for anything that matters to the relationship. Use it only for routine questions like invoice links and meeting times.
Some replies should never go out on their own. A refund. A change to the scope of work. A price exception. If a wrong answer could cost money or a client’s trust, the system drafts it and a named person approves it before it sends.
Every reply leaves one written line: what was asked, what was sent, and when. An owner reads a small sample each week. That is how a wrong answer gets caught before a client finds it.
Plan on an afternoon to write the first answers, and about 15 minutes a week to read a sample. That is a rough guess. If you already pay for an AI assistant, the tool cost is close to zero. Someone has to own the system. At first that is you. Later it can be a team member or a contractor who checks replies once a day.
Pull the last 30 days of your sent mail. Mark the questions you have answered more than twice. Pick the three that come up most and write your real answer to each. That is enough to start.
I wrote about which tasks are ready for this in The Same-Shape Test. A timer cannot do this job, and A Schedule Is Not Staff explains why. The full picture of how work moves is in the first post in this series.
Picture a client writing at 7 p.m. and getting a clear, correct answer in minutes. You see the thread the next morning, and nothing needs fixing.
The Profit Leak Scorecard shows which questions in your inbox you are still answering by hand.

It’s Handled™ | Operational architecture for founders who built the business, and are ready to stop being the infrastructure.
I’m Rachel Lavern, a founder and Operational Architect. I’ve spent 22 years inside the operations of Fortune 500 companies, and now I help founders step out of the middle of their own businesses. Here’s my story.
November 9, 2026
Be the first to comment