In partnership with

Meta's Muse and OpenAI's dots take different approaches to AI agents. Here's how to decide what's worth testing in your business.

PUT AI TO WORK

Two new agents, two ways to work

Muse for Small Business focuses on tools you already use. OpenAI’s dots are built to stay with a task over time. Here’s what each company says—and how to decide what’s worth testing.

AI tools for small business

Issue #16 · Wednesday, October 7, 2026

A new AI agent can sound impressive and still be the wrong fit for your business. The useful questions are simpler: Can it work with the apps you rely on? Can it handle a task you can check? And can you access it on terms that make sense for you?

Meta and OpenAI have each announced a new approach to AI agents. We’ll compare what they’ve actually described, then look at two Make updates that may affect existing scenarios.

In This Issue

  • Muse and dots: choose by the task
  • Start with a task you can check
  • Check access and price before you plan
  • Two Make updates to check
  • Quick Win: scope a small, verifiable test

Muse and dots: choose by the task

Meta introduced Muse for Small Business as an agent that can connect to tools businesses already use. Meta lists connections including Asana, Canva, Dropbox, QuickBooks, Klaviyo, Notion, Shopify, Slack, Stripe, Zoom, and Facebook and Instagram business accounts. Its announcement says Muse can use business context to help with tasks such as looking at sales information, preparing marketing materials, or surfacing emails that need a response.

OpenAI describes dots as always-on agents with their own cloud computer that can work toward goals over time and connect with more than 4,000 apps through plugins. The announcement says you can reach a dot through ChatGPT, Slack, or Teams. One early-tester example: a dot noticed a missed invoice, prepared it, and sent it after the user’s approval.

Those descriptions point to different starting questions. If the work begins in tools your team already uses, check whether Muse supports the specific accounts and actions you need.

If you want an agent to keep working toward a goal over time, examine how dots handle instructions, progress, and review. Neither description guarantees that a particular workflow will be available in your account or produce the result you expect.

Think of a product announcement as a map of intended capabilities, not a test report for your business. A listed app connection does not automatically mean every feature in that app is available.

Likewise, an agent described as always-on still needs a clear goal, the right access, and a way for you to check what it did. Before trying either approach on important work, identify the exact step you hope to improve and the evidence you would need to trust the result.

Start with a task you can check

Choose a task with a clear input and output, such as drafting a reply to a routine customer inquiry for your review. Be precise about what information the agent may use, what the result should contain, and what it should leave untouched. A narrow request is easier to assess than a broad instruction such as “run my marketing.”

Next, check the connection itself. Confirm that the service supports the app and account you plan to use, and whether it can only read information or can also make changes. If the workflow could send a message, update a record, or spend money, keep a person in the review step while you evaluate it.

Start with a low-risk sample and compare the result against your original records rather than relying on a confident-sounding answer.

Check access and price before you plan

Meta says Muse is free for most of what people need, with subscription plans for more. It is available in the US and Canada.

OpenAI says a first dot is included at no extra cost with Pro and Business Premium plans, with rollout in eligible markets. Enterprise users can try the beta when their workspace administrator enables it.

Those terms are not directly comparable, and the announcements do not list a single equivalent price for every paid option. Check the current offer and eligibility for your account before making a decision.

Availability and plan details can change, so verify them where you would actually sign up—not in a third-party comparison written for a different date or region. Also consider the effort around the subscription: connecting accounts, setting permissions, explaining the task, and reviewing output all take time.

A low or included subscription price does not make a workflow worthwhile if the setup and checking take longer than doing the task yourself.

The practical choice is the tool that can complete a specific job with acceptable effort and a result you can verify. You do not need to select a winner across every feature. You can test one workflow, keep the parts that help, and wait on the rest until there is a clear business reason to revisit them.

Two Make updates to check

This issue contains affiliate links. If you purchase through these links we may earn a commission at no additional cost to you.

If you run scenarios in Make, two items in its October 2 release notes are worth a look. They may affect scenarios using WhatsApp Business Cloud or Google Gemini AI’s video module. If you use either integration, review the relevant scenario and the current instructions from Make before making changes.

WhatsApp Business Cloud pricing

Make says Meta’s new pricing model for service messages took effect October 1. After 1,000 free monthly service messages, additional messages will not be delivered unless a valid payment method is configured in Meta Billing Hub.

If a scenario sends these messages, first identify which scenarios use the connection and whether they rely on the free allowance. Then confirm the account’s message volume and payment setup in Meta’s own billing area. Avoid assuming that a scenario will keep sending once it reaches a limit or lacks a valid payment method.

If you are unsure how the pricing applies, consult the current Meta and Make documentation before editing a live workflow. The goal is to catch the scenarios this affects—not to change ones that do not use this service.

Gemini video module

Make says the Generate a video module now uses the gemini-omni-1.1-flash model instead of the deprecated gemini-omni-flash-preview model. Scenarios using the old model continued to run until September 30.

If you use this module, open the scenario, review the model selection, and save the module with the current option as described in Make’s release notes. After updating it, test with an appropriate sample and verify the result before relying on the scenario again.

If you do not use the affected video module, there is no need to make a change just because the release note mentions it. These are targeted checks, not a reason to rebuild unrelated automations. Keep your test separate from important production work whenever possible.

Quick Win: scope a small, verifiable test

Pick one task with an easy-to-check result, such as turning a short list of sales figures into a weekly summary. Write down which apps the task needs and what a useful result should include. Confirm the agent supports those apps for your account, then try it on a low-risk example and compare the output with your original information. If it misses, note whether the problem was access, unclear instructions, or the task itself, adjust one thing, and test again. Track the time you spend setting up and reviewing, plus any corrections you make. If the task involves sending a message or changing a record, have a person approve that step. Then decide whether the output saved enough time to justify using the tool again.

That’s Issue #16.

AI agents and business tools keep changing. We’ll continue to focus on practical ways to evaluate what’s available and see whether it fits a real job in your business.

If this issue would help another business owner, forward it to them.

— Put AI to Work

Some teams never seem to stop moving. They're on Attio, the agentic CRM.

Every customer signal is captured in one shared context layer, always current and compounding. Agents and workflows build pipeline, chase every buying signal, and move deals forward, an always-on revenue engine running alongside your team.

With Attio, you’ll get:

  • Leads automatically prioritised and routed to the right rep

  • Expansion and risk signals caught the moment they land

  • Follow-ups written in your voice, already there when you arrive

Teams like Parallel, Turbopuffer, and Wordsmith build on Attio. Are you one of them?