The First Time I Saw AI Talk, but No Budget Follow
Last Wednesday afternoon, I joined a meeting at an eight-person design company in Hangzhou. The boss eagerly asked whether we wanted to use AI for proposals. I opened a tool and tried it for ten minutes; everyone said, “Pretty useful.” By month-end, I checked the accounting sheet: total related spending for the month was just 180 yuan.
I assumed I had not pushed hard enough. Later, I realized many companies are stuck in the same place: not because they do not want AI, but because they are unsure which step to tackle first and worry a paid membership will sit unused.
AI Spending Is Rising, but the Money Is Concentrated
The Neuron cited Ramp data showing that enterprise AI spending hit a new high in October, yet most companies spent very little, with a small group accounting for most of the total. Meanwhile, adoption is climbing and model prices are falling.
For small teams like ours, that is actually good news: we do not necessarily need a big upfront investment. At 9 p.m. last Wednesday, Shanghai-based freelance photographer Lin Ran tried using AI in a shared document to organize client requirements. Ten minutes later, she had turned six feedback comments into a list of items awaiting confirmation. She said she would consider a fixed subscription only as a next step; for now, she only handles repetitive text occasionally and does not want to pay long-term for features she will not use.
The tool itself is not complicated. It is more like an assistant you can talk to, use for writing, revise drafts, and organize materials. The real choice is not the “most powerful” option, but the task you repeat every day and are willing to check.
To Replicate This Today, We Do Not Need a Whole System
We can keep the cost down: start with the free version of a new tool at 0 yuan; test it for two weeks, roughly 2–3 hours, with half that time spent comparing our own repetitive tasks. The technical barrier is simple: copy text, paste materials, and know that a person must verify the final result. The first step is not registering for a pile of accounts. It is opening one familiar AI tool, starting a new conversation, and entering one real client requirement with identifying details removed.
I got stuck here too: I once tried three tools at the same time and wrote down a page of features, but ended up using none of them. Later, I picked just one task—organizing meeting notes—and used it for five consecutive days before deciding whether to pay.
At Our Current Stage, There Is No Need to Match Everyone Else's Spending
If we are just starting out, I would use the free tier for two days and test low-risk tasks such as organizing requirements or drafting replies. If we are not ready to try now, that is fine too.
If we have 1–2 clients, I would track our repeated weekly actions and run a small test on the workflow that takes the most time. Names, phone numbers, and other sensitive details must be removed from client materials first.
If we are scaling up, I would wait until at least two people on the team feel they need it, then compare monthly fees, export options, and the cost of manual review. We could set the initial budget at 100–300 yuan per month rather than buying an entire suite.
How hot AI spending is does not matter that much to us. What matters is this: take one small workflow, run it for two weeks, and see what it actually saves us.