Monday morning, my client files were scattered across five apps again
Last Monday morning, I took on an urgent request. The spreadsheet the client wanted was in my inbox, the reference examples were in a chat log, and the previous version was sitting in cloud storage. I copied and pasted three times, and nearly sent an outdated quote. I felt defeated: I was already paying for AI tools, yet I still felt like a porter moving files around. The problem wasn’t my ability. My tools and workflow simply weren’t connected.
AI assistants are shifting from chat windows to workspaces
Put simply, AI assistants used to feel like colleagues you called whenever you had a question. The newer workspaces want them to work directly with email, documents, code, and tasks. Google, Anthropic, and OpenAI are all moving in that direction.
The real shift isn’t another button inside the chat box. It’s whether I can finish a task inside the workspace, check the result, hand it to a colleague, and pick up from there next time.
I know Lin Yue, who runs an independent storefront. She operates a small photography studio in Gongshu District, Hangzhou. In April 2025, she brought client briefs, old emails, and delivery checklists into an AI workspace. Now, whenever a retouching request comes in, she asks the assistant to find the relevant history, then personally verifies the price and deadline herself. She told me the most valuable feature wasn’t “automated copywriting.” It was opening four fewer pages and asking the client for the file one fewer time.
Before reproducing this, I calculated the actual cost
I started with the free version, so the expected cost was ¥0. Building a workflow that could handle real business took roughly 2–4 hours. The technical barrier wasn’t high, but I needed to understand the basics of email and files. No programming was required.
My first step wasn’t buying a plan. I opened the “connectors,” “workspace,” or “projects” entry point in the AI product I already used, connected only one low-risk source—such as a test folder in my personal inbox—and asked it to summarize one email and one document.
I got stuck here too. At first, I fed the assistant the client’s entire history, and old and new versions got mixed together. The answer sounded extremely confident, and I nearly replied using incorrect information. After that, I created one fixed folder containing only confirmed files, then labeled every item with its date and client name. None of this required the technical jargon that gives people headaches. The core was to give the AI less room to guess and preserve my final review step.
Whether it’s worth continuing depends on my current workload
If I were just starting out and most client files still lived in one chat window, I wouldn’t bother setting this up yet. Using existing templates and keeping each client’s files organized would be more practical than learning another tool to feel “more advanced.” There’s no urgency to try now. Once repetitive copying starts wearing me down, that would be my cue to explore this entry point.
If I already have one or two stable clients, I’d pick the most repetitive step and test it for two weeks. For example, I could connect “receive the brief—find past records—build the delivery checklist.” I’d track the time saved and number of rework incidents each week. If the setup merely looked impressive while I still had to cross-check everything, I’d stop there.
If I’m scaling, I’d pay more attention to permissions, records, and handoffs: who has viewed client files, which quote is the final version, and whether a colleague can reuse the workflow. At that point, I might compare whether each workspace supports team sharing, third-party integrations, and audit logs before deciding whether to pay. Getting one low-permission workflow running was safer than migrating every file at once.