Meta's new AI agent for small businesses needs your data

What Meta has announced

An AI agent is an assistant that does not just answer questions but can carry out tasks for you inside other software. On Tuesday 29 September Meta said it was extending its Muse agent to small businesses. It can read Instagram professional account analytics, Facebook Pages and Meta ad accounts, and connect to software such as Shopify, Stripe, QuickBooks, Slack, Canva, Notion and Zoom, among others, according to TechCrunch. The same report says it is free with usage limits, with paid plans for those who need more.

With that data, Meta says Muse can analyse sales, campaigns and social activity to draft a growth plan, review the accounts for unusual expenses, spot which ads are working and flag emails that need a reply, Help Net Security reports. The same outlet notes two caveats: Meta says nothing is published, sent or spent without the user's approval, and Muse launched in September in the US and Canada. None of the sources says when it will reach other countries.

A Tuesday in a pottery studio

Picture a small pottery studio that also sells online. Web orders sit in the online shop. Commissions arrive through Instagram messages and WhatsApp. The cost of clay, glazes and postage goes into a spreadsheet whenever someone remembers, and the receipts pile up in a box.

One Tuesday the owner asks a simple question: which pieces actually make money, and which just make work? Answering it means opening four different places, copying figures by hand and matching them up. It gets pushed to Sunday. On Sunday it does not happen either.

Why does a small firm's data end up scattered?

The scene is made up, but we have seen it in many guises. Small businesses grow through quick fixes: an online shop because they needed to sell, a spreadsheet to keep costs to hand at the end of each quarter, WhatsApp because that is where the customers are. Each tool does its own job and none of them talks to the others.

On top of that, much of what matters never gets written into any system at all: the special price for a loyal customer, the supplier who is always late, the order agreed over the phone. That knowledge lives in the owner's head.

What an AI agent cannot see

This is the cost the news brings into view. An agent like Muse works with whatever it finds in the connected tools. If sales are in one place, costs in another and customer agreements nowhere at all, the analysis it hands back will be as patchy as the data behind it. It could suggest a growth plan without knowing which products lose money.

And the cost does not start with AI. You are already paying it, in hours spent copying figures between systems, in decisions made on gut feeling and in questions left unanswered because joining up the data takes too long. An agent can speed up a lot of that work, but only on top of tidy data.

How can a small business get ready for an AI agent?

There is no need to wait for Muse, or any other agent, to reach your country before preparing. What helps with any agent is having the business's information in tools that can be connected. This order tends to work:

• List where each piece of data lives: sales, payments, costs, customers, orders and stock. Note what exists only on paper or in your head.
• Pick one main home for each type of data and stop copying it into stray spreadsheets.
• Check that your software can connect to other tools. If one cannot, that is your weak spot.
• Bring the essentials together in a dashboard you look at every week, even if it starts with three figures.
• Once that works, try an AI agent on that data, with one specific question, and review every suggestion before acting on it.

For the pottery studio, that might mean bringing the online shop, the cost spreadsheet and the commissions into a single tool. Then the question of which pieces make money can be answered in minutes, whether the owner asks it or an agent does.

How we help at digizone

At digizone we build custom CRM and ERP software: management tools that bring customers, sales, hours and costs together in one place and connect to the software you already use. We start by understanding how your team works and build the tool around that.

One example is ETL. Information on hours, clients, roles and invoicing was scattered across different systems, which made it hard to tell which projects were profitable. We built a custom web ERP that imports the data they already held in Access, records hours by client and project, works out the costs and links them to invoicing. Management now sees client and project profitability on live dashboards. It is the kind of well-ordered foundation on which any analysis, with or without AI, makes sense.

Start with a first version

Getting a business's data in order does not have to be one big project. The MVP idea, a minimum viable product, applies here too: build the smallest thing that already solves a real problem, such as a dashboard that puts sales and costs side by side, and use it from day one.

For the pottery studio, that first version would be a dashboard that reads the online shop and the cost spreadsheet and shows, piece by piece, what is left after materials and postage. Then the Instagram and WhatsApp commissions are added, and later the stock. Every figure that finds its proper home is one more an agent can read, whether that is Muse or another.

If you would like to see how it would work for you, our free workshop, in Spanish, walks through it step by step.