Wednesday afternoon, I was again hunting for an AI assistant for a small team
Last Wednesday afternoon, I was sitting in a shared office in Nanshan, Shenzhen, staring at three candidate tools. Some required account and development-environment setup; others offered plenty of features, but I could not tell whether they could handle everyday work. My needs were simple: organize customer feedback, draft quotes, and avoid sending files to the wrong people. I also did not want to learn a pile of technical jargon just to test one tool.
I got stuck here before: I mistook “can chat” for “can do the work.” I signed up, rushed to recommend it to a client, then discovered I had not figured out team data, permissions, or costs. So this time, I started by seeing whether it could handle one real, small task.
What Inkling 2 Small is, and who is using it
Inkling 2 Small is a smaller-version AI assistant from Thinking Machines. Rather than helping individuals build complex systems, it is positioned as an easier starting point for smaller teams. More precisely, it suits relatively routine work, such as assisting between the team’s existing documents and tools.
I know Zhouran, an independent consultant. At a client retrospective in Binjiang, Hangzhou, in November 2025, she used it to turn scattered requirements in chat logs into a checklist, then draft talking points for the next round of communication. She does not write code, and her team has only two people. What attracted her was the chance to run a limited trial before rebuilding the entire workflow. Still, a demo is not the same as long-term reliability. Testing with our own redacted materials is safer than feeding it a client’s original files.
What it takes to replicate this today: money, time, barrier, first step
Based on the public page’s current information, getting started usually requires no additional software purchase. The main cost is the team’s time: about one hour to register, learn the interface, and run a test. I would initially budget RMB 0 for a tool subscription, but before official use, check the latest pricing, quotas, and whether business features cost extra. The technical barrier is low: knowing how to use a browser and which materials can be uploaded. No coding is needed, but when customer privacy is involved, confirm how data is handled first.
For the first step, click the trial or “Get Started” link in the top-right corner of the official site. After creating an account, do not immediately connect customer materials. Run a fictional feedback sheet through it: classify and summarize the entries, then manually check for omissions and invented claims. If registration leads to a team invitation, seat limit, or payment page, stop and confirm the details with official documentation first. There is no rush to try it; this is not a tool every side-hustle operator must set up.
Match it to your stage; there is no need to go all in at once
If I were just starting out, I would first use it for personal tasks, such as organizing a week’s topics or turning interview notes into an outline. I would show it only public or fictional materials and judge whether its output saves time. With one or two clients, I would build one fixed flow: customer materials go in, the assistant produces a draft, and I review it before sending. The focus would be reducing repetitive organization, not chasing “full automation.” While scaling, I would list permissions, costs, data retention, and member collaboration before deciding whether to connect more business systems.
My principle is simple: test one real but low-risk task for seven days, and track time saved, rework, and actual cost. I have also made the mistake of installing so many tools that my work became more fragmented. If I do not have a clear pain point right now, I can wait. An AI assistant is worth trying, but not worth handing over a client workflow just to chase what is new.