Introduction to AI workers

You already know how to use AI for one-off tasks—like drafting, summarizing, brainstorming, or answering questions. Every time you use AI, you need to ask for the same task over and over. This approach has some obvious limitations:

• You need to do extensive chatting in order to get something done

• You will need to repeat the same chatting every time, even if the task is the same

• Every request you make will be slightly different, leading to inconsistent outputs

Toolhouse overcomes these limitations by letting you create AI workers. AI workers are a new type of agents designed to be used for repeatable workflows—work you’d otherwise do manually, re-explaining the steps each time, and copying information between tools.

If you’re new to agent building, let’s focus on the core concepts first so when you start building, you’ll know how to set up your AI worker for consistent results.

What is an AI worker?

Generally speaking, an AI worker is a system that carries out a task with three components: a trigger, a process that may include specialized skills, and tools or systems it can connect to.

Trigger	What starts the worker	A schedule (“Every weekday at 9am”) or a manual run (for example by sending an email to your worker).
Process and skills	The steps the agent follows to complete the task the way you expect	Reviewing inputs, checking for missing information, drafting an output, and handing it off or taking the next action.
Integrations	The approved tools and systems the agent can use to gather information and, if allowed, take actions.	Slack, a CRM, internal documentation, a ticketing system, or a shared document.

AI Workers are most useful when the work is:

• Repeatable: The same task comes up regularly

• Structured: There’s a clear format for the output (so you can tell if the agent is doing a good job)

• Time-based or event-driven: It runs on a cadence or is triggered by an event

• Tool-based: It requires reading from or writing to systems your team uses

For open-ended thinking, brainstorming, or exploratory writing, regular chat is often a better fit—especially for one-off tasks.

Agents are also different from traditional API workflows you may have built in the past. Traditional workflows in other tools are often deterministic, meaning each step is explicitly defined, and the system follows the same path each time unless you change the logic. Agents are more probabilistic. They still operate within instructions, tools, and guardrails, but they use a model to interpret context, make bounded decisions, and adjust how they move through the work.

Building an AI Worker

A helpful way to create an AI Worker is by breaking your usual workflow into separate tasks. Imagine your AI Worker being a new teammate. You need to define roles and responsibilities, as well as thing they can and cannot do.

Toolhouse makes this process straightforward thanks to Agent Builder. Agent Builder makes you explain your tasks in plain language, and it will derive precise instructions for your worker. Agent Builder will automatically add objectives, triggers, processes, and integrations for you. It will also add guardrails and governance to your worker. Agent Builder can ask follow on questions when the task is not clear, and it will allow you to add knowledge files (such as brand guidelines or internal process guides) to your worker.

In Agent Builder, you can start with a simple prompt like this:

I want to automatically draft an email campaign and schedule weekly. The content should be based on our approved brand guidelines.

Agent Builder will understand this simple request and and some follow up question:

• It will ask you to upload your brand guidelines

• It will ask you if you have a preferred time and day for scheduling

• It will ask you what email system you use

You can then reply with your details, for example:

• You can drag and drop your brand guidelines

• You can say you want campaigns to be generated Monday at 10am PT

• You can specify you are using Mailchimp

With this details, Agent Builder will create an AI worker for you with the following details:

Title	Marketing Campaign Scheduler
Objective What is the overall goal of the worker?	Draft an email campaign
Personality How should the worker respond?	Always create an artifact, create a professional recap in bullet points
Trigger What starts the work?	Scheduled trigger: Mondays at 10am PT
Process What steps should it follow?	Read brand guidelines, understand content, create message and draft a Mailchimp campaign without sending it.
Tools What systems or applications do you need access to?	Mailchimp integration
Governance What controls do you need? When should it stop and escalate?	Draft recommendations; approval required before sending

Examples of workers

In this section you will find common, repeatable patterns of work that show up across teams and functions. Each one represents a type of AI worker, each one performing a clearly scoped task. The specific tools, data, and outputs may vary, but the underlying pattern stays consistent.

Briefing Pull information from multiple places, distill it, and package it into something decision-ready.	1) Collect inputs 2) Compare and extract key signals 3) Summarize for audience 4) Share as doc, memo, or briefing	Sales: Build account briefing from CRM, calls, Slack, and news. Marketing: Compile campaign or competitor recaps from analytics, social, and docs. Exec: Produce daily industry briefs with recommendations
Triage and routing Process inbound items and ensure they reach the right next step	1) Review inbound items 2) Categorize and prioritize 3) Route or create next-step artifact 4) Notify owner and requester	Support: Turn feedback into bugs, feature requests, or follow-ups. IT/Ops: Handle internal requests in Slack and route by urgency. Recruiting: Screen inbound candidates and move them into the right path.
Analysis and recommendation Interpret data or evidence, form a point of view, and turn that into a first deliverable.	1) Pull source data 2) Analyze patterns, gaps, and tradeoffs 3) Form recommendation 4) Draft memo, deck, spreadsheet, or email	Finance: Reconcile budget vs. actuals and draft a manager memo. Product/Research: Analyze user feedback and propose priorities. Procurement: Compare vendor quotes, create decision matrix, and draft PO email.
Content creation Create or update content, then tailor it for a specific audience or channel.	1) Generate first draft from notes or source material 2) Edit for tone and accuracy 3) Tailor to audience and channel 4) Publish or send	Marketing: Turn a brief into campaign assets and circulate for feedback. Manager: Turn notes into onboarding plan or training module. Sales: Write follow-up emails, QBR summaries, or stakeholder recaps.
Planning and coordination Turn goals into scheduled work and system updates	1) Build plan from constraints 2) Check dependencies and availability 3) Take action or prepare action 4) Update systems and notify people	PM/Chief of Staff: Block work, schedule meetings, and update trackers. Events: Coordinate off-sites, registrations, or travel plans. Admin: Manage forms, reminders, orders, bookings, and follow-ups.

Testing and training an AI Worker before it works with you

You can and should train your worker before making it work with you. When you create your worker, you will see the Testing Workbench appear. The Workbench allows you to test your worker with real data, and to see its output. You should test the worker in the Workbench a few times in order to understand if its output matches your expectations.

• If the output looks good: your worker is ready. You can work with it right away.

• If the output needs improvement: tell Agent Builder that went wrong and ask to fix. You can also copy the output and paste it into Agent Builder to provide context as to what needs to change.

Using AI Workers in your day-to-day

You may start by using workers your organization has already built. Workers can be shared with your team and they will behave in the same way.

Interaction type	Human interaction?	Description
Schedule	No human interaction	Worker will start automatically on a schedule you defined when creating the worker.
Event	No human interaction	Worker will start on a trigger, for example when they receive a notification from a webhook or from another service.
Chat	Human interaction	Worker will start when you send a message or a document.
Email	Human interactio	Worker will start when you send an email to its address.

Editing your AI worker

As you test your AI worker, it’s expected that the first version won’t be perfect. When something feels off, there are two effective ways to improve it.

• Coach it in natural language: When the issue is less clear, go to Agents, select your agent, then use the chat view to work through it. You can point out what it missed, ask open-ended questions, or have it explain its reasoning to surface gaps in the instructions.

• Update the instructions directly: In the Agents edit page, click the Details page to make small, specific changes in the System Prompt field or in the Integrations list. You can clarify steps, adjust the output format, or add a constraint.

After making changes, test again in the Testing Workbench to confirm the update worked.

Scaling your AI workers to your team

AI workers are designed for shared, repeatable work. Sharing a worker gives your team a consistent way to complete a task instead of reinventing the process each time.

Shared agents work best when they’re tied to a specific, recurring workflow your team already understands.

Remember that workers share the same access. If you share a worker that connects to Google Drive, it will use your Google Drive connection. If your teammates need their own permissions for the agent to work with systems like Slack, Gmail, or other tools, you can make a separate copy of the agent and transfer it to them.