I first realized AI wasn’t “just asking everyone to write more copy”
At 10 p.m. last Tuesday, I was revising a client proposal in a shared office in Hangzhou. All three of us had AI subscriptions, but the group chat had a message that stopped me: “It writes faster than I do. Do you still need me?” That was when I realized the hard part wasn’t providing the tools. It was letting AI take over the work without making my colleagues feel replaceable. I’d made this mistake before: sharing account credentials in the chat and assuming digital transformation was done once everyone knew how to use them.
Proton’s dispute exposed the real problem behind an “AI team”
According to The Neuron, Proton faced internal disagreement over how AI should be used and managed. My take: the issue wasn’t whether employees were willing to work. It was whether the company had clearly defined which tasks AI would handle and which decisions people would remain accountable for. My friend Lin Yue, a personal IP consultant in Chengdu, uses AI to organize interview recordings, checks the facts herself, then sends the finished draft to the client group. What she saves is transcription time; she doesn’t hand over judgment wholesale. An AI team is a workflow, not several extra subscriptions.
Replicating this today doesn’t have to be expensive
I’d start by copying one minimum viable process rather than training the whole company. At roughly $20–$30 per month for a typical individual subscription, ¥100 should cover a one-week test. The technical bar is being able to explain the inputs and check the output; no coding required. Budget two hours: about 40 minutes on day one to choose one repetitive task, such as meeting notes, paste a redacted sample into the tool, and generate a first draft; about 30 minutes on day two to verify the facts and rewrite it; then 50 minutes turning the steps into a team checklist. I wouldn’t connect automatic email sending, and I wouldn’t rewrite ten workflows at once.
Match the setup to the stage you’re at
If I were just starting, without a stable weekly client base yet, I’d use free allowances to organize proposals or summarize contracts, run it for a week, and not worry if I postponed trying it. With one or two clients already, I’d create a repeatable “AI first draft plus human confirmation” workflow and spend the saved time communicating with clients. If I were scaling, I’d assign permissions and review responsibilities by role, then decide whether a team plan was worth buying. I’d use AI to reduce repetitive work, not to add another expense that merely looks fashionable.