The agent is good, but you are still the one who has to remember to use it.

There is a version that does not wait for you. You set a rule once: when a new invoice lands in this folder, pull the numbers into my sheet. When a lead fills out my form, research them and draft the first reply. When a deal goes quiet for 14 days, flag it and write the follow-up. Then you walk away. The agent watches for the trigger and does the work the moment it happens, whether you are at your desk or asleep.

This is the difference between an agent you operate and an agent that operates on its own. The first saves you minutes. The second removes the task from your plate entirely.

This week: how to build your first event-triggered agent, and why this month is the one where autonomous agents stopped being a demo and started being something every platform ships.

The Old Way:

Before AI, reactive work was all manual and all yours. A lead fills out a form, so you stop what you are doing, look them up, and write back. An invoice arrives, so you open the spreadsheet and key in the numbers. A support ticket comes in, so you draft a reply. A contract lands, so you read it and pull out the dates that matter.

None of these are hard. The problem is they interrupt. Each one drags you out of real work to do two minutes of mechanical processing, and the cost is the context switch, not the task. Knowledge workers lose a large share of the day to exactly this kind of reactive busywork, the small things that pile up because they each demand a response the moment they arrive.

Even scheduled AI does not fully fix this. An agent that runs every morning at 7 still leaves the lead waiting until 7 the next day. Reactive work needs a reactive agent.

The Replacement:

An agent with a trigger. You define the event and the response once, and the agent fires the moment the event happens.

The pattern is always the same three parts: an event happens, the agent processes it with your context, the agent writes the result back to your tools. A new ticket arrives, the agent reads it, drafts a reply, and saves it for your review. A lead submits a form, the agent researches the company and writes a personalized first email. An invoice hits a folder, the agent extracts the line items and updates your tracker.

This is not scheduled automation. A scheduled task runs whether or not there is anything to do. A triggered agent runs only when the thing it is waiting for actually happens, which means it handles the work in real time instead of in a daily batch.

The capability is real and available now across several platforms. Dust lets you set agents to fire on webhooks from tools like GitHub, Jira, and Zendesk, and write results back to Slack, Notion, and HubSpot, with every run logged so you can audit what happened and why. Any tool that can send a webhook can be the trigger.

What this replaces: the constant low-level interruption of reactive work, and the mental cost of switching tasks every time something lands.

Setup time: about 20 minutes per agent. Then it runs on its own.

Triggered agents are not brand new. What changed this month is that they stopped being a feature on one platform and became the direction the entire industry is moving at once.

Microsoft just introduced a new category it calls Autopilots, always-on autonomous agents that operate without waiting for a prompt, starting with one that runs across Outlook, Teams, and SharePoint. Its Work IQ APIs, which map how an organization actually works so agents can act on it, ship June 16 (source: Vybe). Dust added event-based triggers, write access to Google Drive, and the ability to loop an agent into a task by email. The common thread across all of them is the shift from agents that wait to agents that watch.

That is the bet to understand. The first wave of AI was reactive: you asked, it answered. This wave is proactive: you set the conditions, it acts. The people who learn to think in triggers now, who get used to defining "when X happens, do Y" instead of doing Y by hand every time, will have a head start on a way of working that is about to become standard.

The Pro section below includes:

What you get: the reactive busywork handled the moment it arrives, without you in the loop.

The Triggered Agent Setup: The full 20-minute build, 6 copy-paste recipes (invoice processing, lead response, ticket triage, contract review, deal follow-up, file organization), the guardrails that keep an autonomous agent safe, and how to pick what to automate first. → Complete walkthrough, copy-paste recipes, no coding required

The Autonomous Agent Playbook: What the shift from reactive to proactive agents means for how you work, the three things to automate first in any business, and how to stay in control when agents act on their own. → Full breakdown with specific actions for your situation

Monthly resource drops:

"I set up one automation from the toolkit last weekend and saved 6 hours in the first week alone. It feels like having an extra team member running in the background."

Elena M. DTC Founder.

🔒 Pro members get the full trigger setup, all 6 agent recipes, and the guardrails. Have your first one running this week.

This one is a look ahead, not a set-it-up-today tool, but it is the clearest example yet of where autonomous agents are going: handling real money under hard limits.

MetaMask launched Agent Wallet on June 8. It is a self-custodial wallet built so an AI agent can trade and interact with DeFi on your behalf without ever holding your keys. The interesting part is not the crypto. It is the security model, because it is the same problem every autonomous agent faces: how do you give software enough freedom to be useful without letting it cause damage when something goes wrong.

MetaMask's answer is a wallet with a leash. You set the rules before the agent runs: daily spend limits, an allowlist of approved protocols, a risk profile. The default Guard Mode enforces those limits onchain and requires human approval by two-factor for anything outside policy or flagged as malicious. The agent cannot opt out of the checks.

The honest caveat: this is early access only, roughly 200 users, command-line only, with general availability targeted for summer. Most readers cannot use it yet, and it is crypto-specific. But the design is worth studying because the set the limits first, enforce them automatically, require human sign-off on anything risky pattern is exactly the guardrail model the workflow above uses, applied to money.

🔒 Implementation checklist is for pro members. Get a breakdown plan to implement AI into your workflow.

🤖 Microsoft launches Autopilots, a new class of always-on agents. The first, Microsoft Scout, runs autonomously across Outlook, Teams, and SharePoint without waiting for prompts, with its own governed identity. Its Work IQ APIs ship June 16 (source: Vybe). Why this matters for you: the biggest software company on earth just made always-on agents a default part of its stack. This is the workflow in this issue, at enterprise scale.

🚫 The US government banned a frontier model three days after launch. Anthropic released Fable 5 on June 9. On June 12, citing national security, the government issued an export control directive suspending access for all foreign nationals, and Anthropic disabled it for everyone. Why this matters for you: it is the clearest signal yet that AI capability has crossed into territory governments now police directly. More in the Big Story.

🔗 MetaMask ships a wallet for AI agents. Agent Wallet lets autonomous agents trade DeFi under hard, onchain spending limits with human approval for anything risky (source: CoinDesk). Why this matters for you: it is the Tool of the Week, and the security model is the template for letting any agent handle high-stakes actions safely.

🧠 Anthropic's MCP crosses 97 million installs. The standard that connects agents to tools passed 97 million installs in March and is moving under Linux Foundation open governance (source: Crescendo). Why this matters for you: the plumbing that lets agents reach your tools is now industry-standard infrastructure, not an experiment. Every triggered agent you build rides on it.

For two years, AI agents have mostly been things you operate. You open a tool, give it a task, watch it work. Useful, but still reactive, still waiting on you. This month, several of the biggest companies in software shipped the next version at the same time: agents that act on their own.

Microsoft introduced an entire new category it calls Autopilots, led by an always-on agent that runs across your email, chat, and files without waiting to be asked, and backed it with Work IQ APIs that map how your organization actually works so agents can act on it. Dust shipped event triggers and email-based control so you can loop an agent into a task like a colleague. MetaMask gave agents a wallet that can move real money under hard limits. And the protocol that connects all of this to your tools, MCP, crossed 97 million installs and moved under open governance.

Then the most striking signal of all. Anthropic released its most powerful model, Fable 5, on June 9. Three days later, on June 12, the US government issued an export control directive citing national security, and Anthropic disabled the model for everyone. In internal testing the underlying model had found 181 working security vulnerabilities in Firefox where the previous model found two. A government pulled a working, publicly available AI model off the market in three days because of what it could do.

Put those together and the pattern is clear. Agents are moving from things that suggest to things that act, the capability is accelerating fast enough that governments are now intervening, and every major platform is racing to ship the proactive version at once. The way people work with AI is shifting from giving instructions to setting outcomes and standing rules.

🔒 The full breakdown (what the shift from reactive to proactive agents means for your work, the three things to automate first in any business, and how to stay in control when agents act on their own) is for Pro members.

New This Week:

Prompt Library Updates:

NEW Resources This Month:

Each resource lives permanently in your Pro account. Use them whenever you need them.

The agents you used last year waited for you. The ones shipping this month do not. For the next few months, building one that runs on a trigger is an edge. By the end of the year, it is how everyone works. The people who learn to think in standing rules now, who get comfortable setting an outcome and a limit instead of doing the task by hand, get the head start.

If you do one thing before next Thursday, pick the reactive task that interrupts you most and build one agent to handle it. Set it to draft-only, watch it for a week, then let it run. The first time it does the thing before you remember to, you will see your work differently.

👇 What’s behind the paywall:

Pro members deploy AI in their work an average of 5-8x more often than free readers (based on reply data from past issues). The difference is having the exact setup, not the concept.