
Meta, OpenAI and Google have all launched AI agents for business in the space of a few weeks. Here is where agents genuinely help a small business today, what experts and regulators warn about, and a simple, safe way to start.
A chatbot answers a question. An AI agent takes a goal, works across the apps you connect to it, and carries out several steps to get there: pulling sales figures, drafting a campaign, preparing an invoice or sorting customer enquiries. Karen Kerrigan, President and CEO of the Small Business & Entrepreneurship Council (SBE Council), puts the shift simply: “Now, AI is moving from simply helping business owners do the work to actually getting work done on their behalf.”
That promise is easy to oversell. The useful question for an owner is not whether agents are the future, but which jobs in your week an agent can take on now, what it needs from you to do them well, and where a person still has to sign off. This guide sticks to what the vendors, analysts and regulators have actually said.
On 29 September 2026 Meta expanded Muse, the personal AI agent it launched on 8 September, into Muse for Small Business. As TechCrunch reported, owners can connect Instagram professional account analytics, Facebook Pages and Meta ad accounts, plus tools such as Shopify, Intuit QuickBooks, Stripe, Canva, Slack, Notion, Klaviyo and Zoom. It is free with usage limits, with paid plans for more use. Meta's own framing: “They told us they're short on hours, not ideas. So we built Muse for Small Business to help get work done with the tools they already use.” According to Meta, as quoted by SBE Council, nothing publishes, sends or spends without the owner's approval.
The same day, OpenAI introduced dots: always-on agents powered by GPT-6 Astra, each with its own cloud computer, that can connect to more than 4,000 apps and be messaged in ChatGPT, Slack or Teams. They are rolling out to ChatGPT Pro and Business Premium users in eligible markets. OpenAI's most small-business-friendly example is modest and telling: an early tester's dot noticed he had forgotten to invoice a publication, prepared the invoice and sent it after his approval.
Google followed at Gemini at Work 2026 with the Gemini agent, which can run as a personal assistant or as a persistent coworker agent with its own Workspace account and email address. Google says these capabilities are not publicly available yet, but one detail is worth noting for any business budget: hard, real-time spend caps that pause an agent automatically when they are reached.
Small firms are not waiting. SBE Council's forthcoming Small Business Check Up Survey, previewed on 30 September, found that 58% of small employers already use AI agents regularly or occasionally, and another 14% are considering or planning to. Its 2026 technology survey found that 82% of small business employers have adopted AI tools, with the typical business using five.
Depth is another matter. The Intuit QuickBooks 2026 AI Impact Report, built on more than 34,000 survey responses and anonymised payment records from more than 5.3 million businesses, found that roughly 7 in 10 businesses say they use AI regularly, but only about 1 in 10 pay for dedicated AI tools. The top barriers it lists are about trust: data privacy, accuracy and limited knowledge of what AI can do. Cost is a factor, but not the biggest one.
Analysts also question which agent a small business should pick. Mark Beccue of Omdia told TechTarget he is sceptical that owners will choose Muse over “more proven cross-systems players” such as Microsoft and Google, adding: “Muse might make sense for SMBs that are consumer-oriented retail online businesses.” He also observed that small firms are “more concerned with cost-effective solutions that require fewer resources” than with handing data to an agent.
Ask an agent to pull sales, ad and social figures from the tools you connect and explain what changed and why. SBE Council lists analysing sales and marketing results, reviewing ad performance and flagging unusual expenses among the jobs agents can already take on. It is low risk because the agent reads and reports; you decide what happens next.
Agents can study what worked, research trends and draft posts, emails or a campaign plan in your brand voice. The safe pattern, which Muse and dots both build in, is that drafts wait for your approval before anything is published or sent.
Overdue invoices, unanswered emails, calendar clashes and project follow-ups are where an always-on agent earns its keep. OpenAI's invoice example is the model: the agent spots the gap and prepares the work, and a person approves it.
An agent can read incoming messages, group common questions and draft replies. Decisions with a real effect on a customer, such as refunds, refusals or individual pricing, should stay with a person.
Many owners keep customers in one app, orders in another and finances in a third. Connecting them is the real value, which is why Google's Darren Mowry writes: “An agent is only as useful as the systems it can reach.”
Agents act, and actions can go wrong. TechTarget noted that The Guardian reported Meta's consumer Muse giving a Facebook Marketplace buyer a seller's home address and arranging a meeting without the seller's knowledge; Meta said it was investigating. OpenAI's own launch post is plain about the limits: “Dots can still make mistakes, so always review consequential work.”
Gartner analyst Wayne Hankins argues the fix is not only technical: “Agentic AI is ultimately a governance challenge.” He recommends clear limits on the data, tools and decisions an agent can access, and says: “As agent autonomy increases, organizations should implement human approval workflows for sensitive actions.”
Regulators agree. The UK Information Commissioner's Office says that “AI agency does not mean the removal of human, and therefore organisational, responsibility for data processing,” and that organisations “should not give agentic AI systems access to information just because it might be useful in the future.” The ICO opened a call for evidence on agentic AI on 8 October 2026, running until 20 November 2026, ahead of final guidance.
Agents work with whatever systems you give them. A website with a working contact form, clean product and service pages, and data held in proper tools (a CRM, a store platform, accounting software) gives an agent something reliable to read and act on. A business run from scattered spreadsheets and inboxes gets far less from the same agent.
That is where we focus at CoreCodify. We design and build websites, mobile apps, UI/UX and brand identities, and we build AI solutions such as assistants, enquiry triage and internal tools that connect to your real data with the right permissions and approval steps. If you want to explore where an AI agent could save your team time, or get your website and data ready for one, tell us about your business through our contact page.