Key takeaways
- An AI agent does a piece of work end to end, such as drafting a customer reply from your records, instead of only answering questions.
- 43% of Australian small and medium businesses reported some AI use from December 2025 to February 2026 (National AI Centre).
- Agents are still early: the National AI Centre says agentic AI remains largely untapped by SMEs.
- Automating repetitive tasks is where SMEs think AI could help most: 35% picked it in NAB's survey of about 670 SMEs.
- The government's advice is to start with AI built into tools you already use and build custom only when the work justifies it.
- Trust is the biggest barrier, so the safest first agent is one where a person approves anything that reaches a customer.
What can an AI agent do for a small business?
An AI agent does a job rather than just answering questions. It reads an email, looks up your records, drafts the reply or updates the system, and asks a person before anything important goes out. For small businesses the common first jobs are customer replies, lead follow-up, data entry and admin that someone repeats every week.
Where Australian SMEs think AI could help most
| Area | Share of SMEs picking it |
| Automating repetitive tasks | 35% |
| Marketing and sales optimisation | 31% |
| Improving decision-making | 23% |
| Enhancing customer experience | 22% |
| Cost reduction and efficiency | 18% |
| Risk management, cybersecurity and fraud detection | 12% |
| Talent and HR support | 6% |
| None: AI will not assist my business | 16% |
Automation matters most in process-heavy sectors: 52% of Finance and Insurance and 52% of Business Services SMEs picked it, against 22% in Retail.
What that looks like as agent jobs
- Customer email: Zapier's ready-made Support Email Agent "drafts customer replies using your knowledge base".
- Website enquiries: the National AI Centre lists "adding an AI chatbot to a website to answer customer enquiries" as a typical way to connect AI to your systems.
- Sales and stock: it also lists "connecting AI to sales or inventory system" and analysing data pulled from several systems.
- Leads and meetings: Zapier's templates include a lead enrichment agent and a meeting prep agent.
The difference from a chatbot or ChatGPT is that an agent works inside your tools and finishes the step. If you want the longer version, read what a personalised AI agent is and how it differs from ChatGPT.
Sources: NAB Behavioural & Industry Economics, Embracing AI: Adoption & Key Opportunities Identified by SMEs (April 2026; about 670 non-farm SMEs surveyed 10 November to 9 December 2025); National AI Centre, Choose a solution (published 22 April 2026); Zapier, Zapier Agents (page undated). All fetched 7 November 2026.
How many Australian small businesses use AI, and are they using AI agents yet?
About four in ten. The National AI Centre's monthly survey found 43% of Australian small and medium businesses reporting some AI use from December 2025 to February 2026, and NAB's survey of about 670 SMEs found 42%. Agents are a different story: the National AI Centre says agentic AI remains largely untapped.
What the two surveys found
| Finding | Figure | Source |
| SMEs reporting some level of AI adoption, Dec 2025 to Feb 2026 | 43% (44% in February 2026) | National AI Centre |
| Same measure, previous quarter (Sep to Nov 2025) | 45% | National AI Centre |
| Adopters using AI for content generation / data analytics | 54% each | National AI Centre |
| Adopters using AI for cybersecurity and threat detection | 48% | National AI Centre |
| SMEs using AI today / not using / planning to | 42% / 44% / 14% | NAB |
| Highest-using sectors | Property Services 69%, Finance & Insurance 64%, Business Services 61% | NAB |
| Lowest-using sectors | Retail 22%, Transport & Storage 21% | NAB |
On agents specifically, the National AI Centre says more complex uses "such as agentic AI, supply chain optimisation, and AI-assisted HR – remain largely untapped". Neither survey publishes a separate figure for agents. So a small business putting an agent to work today is early, not late.
The National AI Centre also reports that broad adoption, where AI is used across several parts of the business, reached its highest level in seven months, while limited one-off use is declining: businesses that start tend to expand.
Sources: National AI Centre, AI adoption insights: December 2025 to February 2026 (first published 7 May 2026, updated 3 June 2026; SME AI Pulse by Fifth Quadrant, at least 400 SME owners and decision-makers a month, weighted by industry, state and size); NAB, Embracing AI (April 2026). Fetched 7 November 2026.
What are the best AI agents for a small business?
There is no single best agent. The best one is the simplest that does the job. The National AI Centre's advice is to start with AI already built into software you pay for, then standalone tools, then AI connected to your systems, and to build a custom solution only for complex, higher-impact work.
The four ways to bring AI into a small business
| Approach | Best for (National AI Centre) | Example |
| 1. AI built into software you already use | Everyday tasks where setup needs to be quick and risk is low | Accounting software that categorises transactions or flags unusual activity; Gemini inside Gmail, Drive and Sheets on Google Workspace Business Standard |
| 2. Standalone AI tools | Testing AI, low-risk tasks, or use by a few staff | A chat assistant you log into to draft emails, research or summarise |
| 3. AI integrations | More capability without a fully custom build | A ready-made agent on an automation platform; Zapier Agents connects to 9,000+ apps and has a free tier of 400 activities a month |
| 4. Customised AI solutions | Complex, higher-impact uses where control, data handling and accountability matter | An agent built around your workflows, data and approval rules |
The National AI Centre's own wording: "Start with the first option and move to later options if you need more control or capability." It also says to check what you already have, because AI features in your existing systems "can reduce cost, complexity and risk".
A note on "best AI agent" lists: many of the ones we checked on 7 November 2026 were published by agencies or tool sellers ranking themselves. Use them to learn what exists, not to decide. On Google Workspace, agent management sits in the Enterprise plan (Gemini Enterprise), not in Business Standard. Zapier bills agents in US dollars on "activities, not tasks".
Sources: National AI Centre, Choose a solution (22 April 2026); Google, Google Workspace AI (Australian page, undated); Zapier, Agents pricing reference (updated 6 November 2026). Fetched 7 November 2026. Plans and allowances change; check before you buy.
Should a small business use an off-the-shelf AI agent or get a custom one built?
Start off the shelf. Ready-made agents cost little to try and suit simple, low-risk steps. A custom agent makes sense when the work needs judgement, runs across several of your systems, touches customer data and repeats often enough to pay back the build. The National AI Centre's rule: choose the simplest option that meets your needs.
Off-the-shelf vs custom, side by side
| Off-the-shelf agent | Custom agent |
| Setup | Hours to days, often by you | Weeks, by a builder |
| Fits your process | You adapt to the tool | Built around how you work |
| Knows your business | What you upload or connect | Your records, rules and past decisions, by design |
| Control and approvals | Whatever the tool offers | Set by you, step by step |
| Ongoing support | The vendor's help desk | Monitoring, updates and fixes (what the government calls MLOps) |
| Lock-in risk | Workflows live in the vendor's platform | Depends on the contract: ask who owns the agent and the data |
The honest trade-off: off-the-shelf is cheaper and faster, and for a single simple step it is usually enough. It struggles when the job crosses your inbox, CRM, accounts and documents, needs judgement, or must follow your rules every time. That is where a custom agent earns its cost. The National AI Centre notes that "more control comes with higher cost and effort" and that "in most cases, building a custom solution comes later if needed".
For prices on both, see how much an AI agent costs in Australia. If you are weighing doing it yourself against hiring help, read whether you need an AI agency or can do it in-house.
Source: National AI Centre, Choose a solution (22 April 2026; fetched 7 November 2026). The comparison table is our summary of the trade-offs, not a government table.
Why aren't more small businesses using AI agents?
Mostly trust, not cost. In the National AI Centre's December 2025 to February 2026 survey, around 65% of SMEs not adopting AI cited distrust of AI decisions or wanting to keep human control. 54% said AI is not relevant to their business, and 19% of all SMEs said they don't know how to use it.
- Trust and control. The National AI Centre calls trust "the dominant barrier" and says businesses need to understand "how they remain in control". An agent that drafts and waits for approval answers that directly.
- Relevance. Fewer than 30% of Construction and Agriculture businesses are adopting AI, while more than half in Health, Education and Services are. The Centre puts this down to a lack of "visible, relatable examples", not capability.
- Not knowing where to start. 19% of SMEs, up 2 points on the previous quarter, said they don't know how to use AI in their business.
- Doubt about value. In NAB's survey, 16% of SMEs said AI will not assist their business at all.
Sources: National AI Centre, AI adoption insights: December 2025 to February 2026 (updated 3 June 2026); NAB, Embracing AI (April 2026). Fetched 7 November 2026.
How should a small business start with AI agents safely?
Pick one repeated, low-risk task and have a person approve anything the agent sends to a customer. About half of Australian businesses already using AI check outputs before they reach customers. Before you subscribe, check where your data is stored, whether it is used to train the AI, and how you would switch the agent off.
A five-step start
- Name the task. The National AI Centre says to define what you want to improve first, "being specific about the outcome you're aiming for".
- Check what you already pay for. Your accounting, email or CRM software may already include AI features.
- Keep a person in the loop. Have staff approve customer-facing output until you trust it. About half of businesses already using AI do this.
- Read the data terms. Check what data the AI uses, where it is stored, whether it trains the model and who owns the outputs. The privacy regulator's best-practice advice is not to enter personal information, especially sensitive information, into publicly available AI chatbots.
- Plan the off switch. Know "how you would pause, fix or switch off the system if needed", and who fixes it. Your business stays responsible for how the AI is used.
For the security side of giving an agent access to your systems, see is it safe to give an AI agent access to my business systems.
Sources: National AI Centre, Choose a solution (22 April 2026) and AI adoption insights (updated 3 June 2026); OAIC, Guidance on privacy and the use of commercially available AI products (published 21 October 2024, updated 17 January 2025). Fetched 7 November 2026.
What does Productbox Labs build for small businesses?
Productbox Labs builds custom AI agents for the work your team repeats every week. We design, build and wire the agent end to end, it is live in about 30 days, it gets monthly skill upgrades after that, and you own the asset. If a ready-made tool covers your job, start there; the table above shows when custom is worth it.
Not sure an agent is worth it for you yet? The AI ROI calculator estimates what one is worth to your team before you spend anything.