TownHow workflows work
Your assistant knows you, and it has access to the tools you use every day. Now let us teach it to use them on its own.
You probably have routines. Every morning, you check your calendar and scan your inbox. Every time an invoice arrives, you log it somewhere. Every time a meeting starts, you wish you had prepped.
Workflows are how you teach your assistant to handle these routines for you. Instead of asking every time, you set it up once and it runs on its own.
Think of it as standing instructions
Remember our delegation analogy? On day one, you tell your new hire: “Send this email.” That is a one-off task.
But after a week, you might say: “Every morning, pull my calendar, check for urgent emails, and send me a summary by 8 AM.” That is a standing instruction. You are not asking them to do it each day - you have set the expectation once.
A workflow is a standing instruction for your assistant. And because your assistant already knows you - your role, your contacts, your priorities - and has access to your integrations, its standing instructions are personalized from the start. Your morning briefing highlights the people and topics that matter to you, not a generic summary of everything.
Starting with suggestions
You do not have to design your first workflow from scratch. Remember the suggestions from Chapter 1?
Based on what your assistant learned about you during onboarding, it recommends workflows that fit your work patterns. If you handle a lot of scheduling, it might suggest an auto-scheduling workflow. If you receive invoices regularly, it might suggest an expense logger. If your calendar is packed, it might suggest meeting prep briefs.
These suggested workflows come pre-configured - the right trigger, the right tools, instructions tuned to your role. You can enable one with a click, see how it performs, and customize from there.
As you use Town more, new suggestions surface. Your assistant notices patterns in your work and proposes automations you might not have thought of. This is an evolving part of Town - the suggestions get more specific and more useful as the product develops.
The four pieces
Every workflow has the same structure. Let us walk through them.
When it runs. This is the trigger. Maybe it is every incoming email. Maybe it is a daily schedule. Maybe it is when a calendar event starts. You decide what kicks it off.
What it does. These are the instructions, written in plain language. “Read each new email, decide if it needs a label, and archive anything that is a newsletter.” You are describing the task the way you would describe it to a person.
What it can access. These are the tools - the systems your assistant can use to complete the task. Reading email, updating a spreadsheet, searching the web, sending a Slack message. You choose which ones to enable for each workflow. The integrations you connected in Chapter 2 are what make these tools available.
How much you trust it. This is the mode, and it is the most important concept in this chapter.
A dial, not a switch
Town gives you a dial for trust, not an on/off switch. You control exactly how much autonomy your assistant has, and you can adjust it at any time.
There are three settings:
- Read-only means your assistant can look at things but cannot change anything. It can analyze your email, research a topic, and report back, but it cannot send, label, or archive. Think of it as shadowing. Your new hire watches and learns, but does not act.
- Approval-required means your assistant does the work, then pauses and shows you what it wants to do. You see the proposed action - the draft email, the calendar invite, the label - and approve or reject with one tap. You can review these approvals from the web app, Slack, or your phone.
- Autonomous means your assistant just handles it. No waiting, no asking. The task runs and completes on its own.
Most people start with approval-required. It is the default for good reason - you get to see what your assistant would do before anything actually happens. After you have reviewed a few runs and trust the behavior, you can promote to autonomous. For more on how safety controls work, see town.com/features/security .
Mixing trust levels
You do not have to pick one mode for everything in a workflow. You can set different trust levels for different actions.
Say you build an inbox triage workflow. You want it to label and archive emails instantly - that is low-risk, happens dozens of times a day. But if it is going to send a reply on your behalf, you want to see it first.
You can set the workflow to autonomous, then override just the email-sending tool to require approval. Labeling and archiving happen silently in the background. Replies wait for your OK.
This per-tool control is how most experienced users set things up. Trust the routine stuff. Review the things that matter.
Building a workflow in conversation
You do not need to fill out a form or write code. Just describe what you want.
Here is a real example. Say you want a daily briefing every morning.
Create a workflow that runs every morning at 8 AM Pacific. Pull my calendar for the day, check for any urgent unread emails, and send me a summary. Keep it short - bullet points, not paragraphs. Skip newsletters and marketing emails.
Your assistant generates the workflow: the trigger (daily at 8 AM), the instructions (what you just said), the tools it needs (email, calendar), and a suggested mode. You can adjust any of it.
Now compare that to a vague request:
Make me a morning briefing.
Your assistant can work with this - and because it knows your role and your patterns, it will make reasonable guesses. But the more specific you are, the less guessing it has to do and the better the first run will be.
Same pattern from Chapter 1. Context leads to better results.
Iterating
Workflows are not set-and-forget. You refine them over time, the same way you would with a colleague.
Maybe your morning briefing is too long. Say: “Make the briefing shorter - just the top 3 things I need to know today.” Maybe it is missing something. Say: “Also include any Slack messages where I was mentioned overnight.”
Your assistant updates the workflow based on your feedback. Tomorrow’s briefing is better. This is the “gets sharper over time” principle applied to automation.
Stock workflows
Town ships with pre-built workflows that work out of the box. These cover the most common patterns:
Auto-label reads every incoming email and categorizes it. Morning Briefing sends you a daily summary of your schedule and what needs attention. Auto-draft writes replies in your voice, ready for you to review and send.
You can enable any of these with one click and start seeing results immediately. And because your assistant already knows you, even stock workflows are personalized - your morning briefing highlights what matters to you, not a generic summary.
What’s next
You now understand what workflows are and how the trust dial works. In the next chapter, we will get into the patterns that make the biggest difference - how to write better instructions, common workflows, and what power users do differently.