I first realized a big company’s valuation can make small teams nervous
Last Wednesday night, I was sitting in a shared office in Hangzhou. A client was waiting for the second draft of a copy deck, but I was staring at funding news about Anthropic: the outside world was debating whether its valuation could reach $1 trillion, while its previous year-end revenue target of $100 billion was now closer to $80 billion. My first reaction wasn’t to study the financials. I quickly asked myself: if I rely on it each month to write drafts and handle customer service, and it raises prices or changes its service, can I switch to another provider and keep working?
That was when it hit me: for people like us, the most direct impact of this kind of news may not be “what will happen to the AI industry.” It may be: don’t put every client deliverable on one provider.
What the news actually says—and why it matters to us
Anthropic is often compared with OpenAI, and it develops the Claude model family. The reports discuss a possible IPO path, its valuation, and the assumptions behind those numbers. Many of the conclusions are still market estimates, not settled outcomes.
My friend Lin Yue, who provides accounting and tax consulting in Hangzhou, works with more than a dozen small restaurants. She used to send document organization, email drafts, and FAQ replies all to one AI service. When she started worrying about a changing budget, she split her workflow into three parts: a low-cost model for document classification, her own judgment for complex decisions, and manual checks for important files. She didn’t replace every tool, but she reduced her dependence on a single provider.
The cost to replicate this today: start with a “task ledger,” no tech required
Money: $0; a ready-made spreadsheet is enough. Time: 30 minutes. Technical barrier: you need to know how to type and copy-paste; no coding required. First step: open the AI service you already use, create a spreadsheet, and write down the three tasks you repeated most often in the past week—for example, organizing customer questions, generating first drafts, or summarizing meeting notes. Beside each task, record only four things: how often it took time, how long each run took, whether customer privacy was involved, and how difficult it would be to switch services.
Then pick the lowest-risk task for a trial. If you receive 20 routine inquiries each week, run 10 through each of two available options and compare output quality with what you actually pay. I made this mistake myself: I bought an annual plan right away, then discovered that only one part of my workflow was genuinely high-frequency. In the early stages, pay-as-you-go usage with a small budget is safer.
If I were considering a tool switch, this is how I’d stage it
If I’m just getting started: I’d choose one free or low-cost option that can handle the main task and get delivery working. There’s no urgency to test everything now; protecting the client’s trust comes first.
If I already have one or two clients: I’d create a task ledger and keep anything involving contracts, money, or privacy under human review. I’d also prepare a backup option, without needing to buy two services at the same time.
If I’m scaling: I’d divide tasks into tiers, set monthly caps for frequent, standardized work, and track the cost and failure rate of each option. That way, when I switch providers later, I’ll be relying on my own operating data—not the mood of the news cycle.