Give Muse one business question to answer
If your Muse account shows the relevant connector, start with one question: which posts from last week deserve another try? Bring the numbers, define your goal, and request a draft you can check.
Meta's September 29, 2026 business update names Instagram professional account analytics, Facebook Pages, and Meta ad accounts, alongside tools such as Canva and Shopify. Business announcement.
Muse is the personal agent product. Muse Spark is the model family, and Meta AI is the assistant surface. They aren't interchangeable names for this connector update. Muse launch, Spark introduction.
Confirm the account and connector before starting
As checked October 2, the September 29 business announcement names Muse availability in the US and Canada. It directs you to Muse app settings for the current connector list. That doesn't promise a connector for every personal Instagram account. Business access and connectors.
The September launch describes iOS, Android, and web access, with free core use and paid expansion. Exact current paid prices weren't verified for this guide. Launch details.
Open the connector you intend to use, check the account identity and requested access, and confirm the right business appears. If it doesn't, stop and resolve the connection rather than letting Muse guess from another account.
Start with the posts and the question
For a small candle shop, a first review might cover seven days of Instagram posts and ask which format to repeat next week. Choose a measurable goal, such as saves or profile visits, before asking what's working.
1. Pick one date range and one account.
2. Provide the current offer and a short description of the audience.
3. Request the available metric names, definitions, and source dates before analysis.
4. Limit the first job to reading results and preparing a draft. Leave customer records, payment tools, and publishing permissions out of this task.
If your connection doesn't expose the needed metrics, provide a reviewed, anonymized export as input where supported. Missing data should stay missing. A connector logo isn't proof that Muse read the whole period.
Copy this weekly review brief
This is an original proposed exercise, not a Nyla hands-on result. Replace every field: [BUSINESS] and [AUDIENCE] describe the shop; [ACCOUNT OR INPUT] identifies the exact source; [DATE RANGE] fixes the period; [GOAL] names the metric; [OFFER] gives the current product and constraints.
Business: [BUSINESS]
Audience: [AUDIENCE]
Source: [ACCOUNT OR INPUT]
Period: [DATE RANGE]
Goal: [GOAL]
Current offer and constraints: [OFFER]
First confirm the account, available dates, metric definitions, and what you can read. List missing data or permissions. If the source or period is wrong, stop.
Review the posts within this period. Separate paid and organic results where the source identifies them. Show the source figures and arithmetic for comparisons. Do not treat reach, views, saves, inquiries, and purchases as interchangeable. Do not claim sales impact without attributable evidence.
Return: a source-linked results table, two observations supported by it, unanswered questions, and a draft three-post plan. Label each recommendation as a hypothesis and name the metric to check next time.
Do not publish, schedule, send messages, spend money, or modify shared content. Stop after showing the draft for my review. Ask before expanding the sources or permissions.Check the arithmetic before the recommendation
Here are fictional inputs, not a Muse output: a candle-pouring Reel reached 1,200 accounts and received 24 saves. A care-tip carousel reached 800 accounts and received 32 saves.
If saves are your goal, the ratios are 24 ÷ 1,200 = 2 saves per 100 reached and 32 ÷ 800 = 4. The Reel reached more accounts; the carousel had the higher saves-to-reach ratio. Neither figure establishes how many purchases a post caused.
A reasonable proposed next test is one more care-tip carousel. Keep the offer similar, record the new date range, and review the same metric afterward. It is a hypothesis to test, not proof that carousels always win.
If Muse mixes views with reach, correct the denominator and request the calculation again. If it labels a paid post as organic, separate them before deciding what to repeat.
Approve a draft after you've checked its sources
Before using the plan, confirm the account, date range, metric definitions, paid-versus-organic treatment, and calculations. Check that the proposed posts fit what you sell and don't invent customer results. If the evidence is too limited, keep the recommendation as a test idea.
Meta says publishing, sending, and spending require your approval. Approval boundary. Its launch also says you can change access or disconnect services. Access controls. Treat those controls as something to inspect, not a guarantee that every result is correct.
Keep the first job finished at the reviewed draft. Ask for another run only after deciding its source, period, and limits. More connected accounts can wait until they answer a specific question.