We gave your shop an org chart a while back: separate agents for support, marketing, orders, and the rest, each with its own lane. It worked, but it put the work of figuring out who should do what back on you. If you wanted a sale announced, a few products described, and the inventory checked, you had to go to each agent yourself and ask.
That changes today. You can now tell Sinan what you're trying to get done, in plain language, and get back a plan: a short list of tasks, each one assigned to the agent that's actually built for it. You see the whole plan before anything happens, and you can tell us if a step is wrong or missing.
What a plan actually looks like
Say your goal is to get ready for a weekend sale. Sinan might split that into a task for the agent that drafts your Instagram posts, a task for the agent that writes customer announcements, and a task for the agent that keeps an eye on stock. Each one works only on the piece that's theirs.
Every task still ends the same way it always has: a draft, sitting in front of you, waiting. The agent that writes the announcement writes the announcement. It does not send it. The one checking stock reports what's low. It does not reorder anything. That hasn't moved an inch, and it isn't going to — a plan just tells your agents what to work on, not what they're allowed to send without you.
Why this is worth having
The honest reason we built this is that "go ask the marketing agent, then the support agent, then the orders agent" is a chore, and chores don't get done at 11pm when you actually have five minutes. A goal is something you can type in one line between customers. The breakdown into tasks is the part we think Sinan should be doing, not you.
It also means the plan is a record, not just a conversation that scrolls away. You can look at it later and see which tasks are done, which drafts are waiting on your review, and which haven't started. That's useful on its own, separate from how fast any one agent works.
It follows you between the phone and your desk
The plan and its progress show up the same way whether you're on the app or at your computer. Start a plan on your phone on the way to open the shop, check it from your laptop at the counter, and it's the same plan, not two separate ones you have to compare. If you close the app in the middle of reviewing a draft and come back to it later, you land right back where you left off instead of starting over.
We're not pretending this replaces knowing your own shop. Sinan breaks a goal into tasks based on what it knows your agents can do, and that's only as good as the setup behind it. If a plan asks the wrong agent to do something, or misses a step entirely, tell us — that's exactly the kind of mistake we want reported, not worked around.
What this doesn't do yet: it doesn't learn a goal you've typed before and suggest it again, and it doesn't let one agent hand work to another without a plan behind it. Both are reasonable things to want. Neither is built.
For now, the change is a smaller one than it sounds: instead of four conversations with four agents, you have one sentence and one plan, and the same approval step at the end of every task that was always there.