From a message to done
this+that turns an incoming message into finished work along one path: a message lands, the AI pulls out the task, a workflow acts on it, you approve anything that needs a person, and the work is done. Here is the whole map.
A designed flow diagram is on the way. For now, the text below is the map.
What happens when a message arrives?
this+that reads your messages across Gmail, Outlook, Slack, Microsoft Teams, Google Chat, and Telegram (Instagram and WhatsApp are coming soon). The moment one lands, the AI reads it and decides whether there is real work inside. When a message has real work in it, a request, a decision, a deadline, a commitment, the AI pulls that out. The rest stay in your inbox as normal messages.
How does a message become a task?
When the AI finds work, it extracts it into DoBox, a task manager that fills itself. A task carries the context from its source thread, and you can assign it to a teammate, set a due date and priority, add a checklist, comment, or attach a file. The message became a tracked task on its own.
What can a workflow actually do with it?
The workflow is the part that acts. You describe what you want in plain language and this+that builds it, with conditional logic, loops, and AI decisions made mid-run. A workflow can draft a reply in your voice, route a lead, create a calendar event, update an external tool through one of 13 built-in MCP servers (or any MCP-compatible tool), or hand the task to a person. One incoming message can start a multi-step sequence rather than a single canned reply.
When does it ask for approval?
Whenever you want it to. You can run a workflow fully automatically, or put an approval gate in front of any action so nothing goes out until you say yes. Most teams start with approval on everything and remove the gates as they come to trust what the AI does. The autonomy is yours to set, one step at a time.
Where does the brain come in?
Anything the AI does can be grounded by the brain, an editable knowledge layer that holds the facts, decisions, and processes you want it to rely on. It is what makes a drafted reply sound like your team and a routing decision match how you actually work. Agents and workflows also write back to the brain as they run, so it grows as the work happens. (Auto-updating the brain from your whole message stream is coming.)
That is the spine.
And every branch off it
Real work does not always run straight down the spine. The same machinery handles the variants:
- Message → task → assigned to a teammate, where a person picks it up
- Message → workflow → action in an external tool, through MCP
- Message → drafted reply → approval → sent
- Message → workflow → agent writes the result back to the brain
- A schedule or an event rather than a message → workflow, like a weekly digest
- Message → filtered as noise, no task and no action
Same pieces, different connections. The product is really one loop, and you decide how much of it runs without you.
The three pillars, each with its own page:
See it on your own inbox
Connect the analyzer and watch what this+that would pull out and act on. Free during beta.