At 10 p.m., Another Client Asked About the Quote

Last Tuesday at 10 p.m., I was still manually organizing client information: emails, quote sheets, and meeting notes scattered across three places. Lin Yue asked about my project timeline, so I dug through our chat history, found the answer, and copied it into a spreadsheet. I felt pretty defeated in that moment: the client assumed I was busy delivering work, while I was just moving information from one place to another.

If repetitive work often drags you down, this isn’t a call to immediately build a complex system. I got stuck too, and I initially made automation sound harder than it was. Later, I realized a one-person company usually doesn’t lack tools; it lacks a short path from “receiving information” to “organizing it and sending a reminder.”

Not a Chatbot, but an Assistant That Follows Fixed Steps

I think of this kind of AI automation workflow as an assistant that handles a set sequence: when a client email arrives, extract the name, requirements, and deadline; draft a quote; put the summary in a spreadsheet; and remind me to review it at the agreed time. It doesn’t make the final decision, but it can handle the research, copying, and formatting first.

I know a design consultant, Chen Hao, who prepares monthly content packages for three regular clients in Beijing. At 4 p.m. on Friday, he runs a workflow that gathers links to assets from client group chats, flags missing images, and generates a to-do list. He used to work until 8 p.m., but now he usually wraps up earlier. Chen Hao says he didn’t cancel any work; he just stopped copying the same batch of links five times.

The most important part isn’t making the AI “intelligent.” It’s defining the fixed steps clearly. The clearer the steps, the fewer the errors. My earlier mistake was connecting email, calendar, contracts, and payments all at once. With too many permissions, I became afraid to use it. Now I connect only the single most annoying, stable part of the process first.

What It Costs to Replicate Today: ¥100 and One Simple Workflow

Money: Start with a free plan or an entry plan costing a few dozen yuan per month; I’d keep the first-month budget under ¥100. Time: Expect about 60 minutes for the first setup, then 20 minutes of maintenance each week. Technical barrier: You don’t need programming skills, but you do need to be willing to write down what comes in, what happens, and where the result goes. First step: Open the automation platform you plan to use, find “Create Workflow” or a similar option, and start with an email-triggered template. If you’re unsure which template to choose, test it with one sample email.

Not everyone needs a tool. If you only receive two or three emails a day, doing it manually may be faster. If the material includes sensitive contracts, check the provider’s data terms before going live. There’s no rush to try this either. For now, write down the action you repeat most often during a week; the workflow often starts there.

I’d Test It Based on Where I Am Now

Just starting out: If my clients aren’t stable yet, I’d automate only “draft a quote from the email.” I’d confirm every price, commitment, and payment term myself rather than letting the tool make promises to a client.

With 1–2 clients: If weekly messages already pile up enough that I miss replies, I’d add a “no reply for more than 24 hours” reminder. It solves missed messages without jumping straight to full automation.

Scaling up: If I’m working with 2–5 people, I’d connect “collect assets, generate tasks, and flag missing items,” while keeping a human review step. The bigger the team, the more important permissions, records, and handoffs become; otherwise, the time I save turns into time spent repeatedly checking everyone’s work.

My rule now is simple: automate the thing I hate doing every day, run it for a week, then decide whether to expand. Building a workflow shouldn’t become another project. First, it should help me avoid one more late night.