AI agents for marketing: what they can really do for a small business
AI agents can already take real work off a small marketing team. They can answer a new enquiry in seconds, turn one article into a week of posts and write your Monday report. They can’t yet run your marketing alone: a person still sets the goal, checks the work and answers for the result.
Small businesses are adopting AI fast. In September 2026, 30% of UK businesses used at least one AI technology, up from 9% when the Office for National Statistics first asked in 2023. Firms with fewer than 10 staff were close behind, at 29%. In Ireland, the Central Statistics Office found that 20.2% of enterprises used AI in 2025, up from 8.1% in 2023. Across the EU, marketing and sales is the most common use: 34.7% of firms that use AI apply it there, according to Eurostat.
The word “agent” now appears on almost every marketing tool, often as a new label on old software. Here is what is real, and how to start safely.
What an AI agent actually is
An AI agent is software you give a goal rather than a script. It decides the steps itself and uses your tools to carry them out: your inbox, calendar, CRM or ad account. Anthropic puts the difference simply in its guide Building effective agents. Workflows follow “predefined code paths”. Agents “dynamically direct their own processes and tool usage”.
| Type | What it does | Marketing example |
|---|---|---|
| Chatbot | Answers questions when asked | Tells a visitor whether you design logos |
| Automation | Follows fixed rules you set | A form arrives, a welcome email goes out, sales gets a task |
| Agent | Works towards a goal, choosing its own steps and tools | Reads a new enquiry, checks your calendar, replies with times and logs it in the CRM |
That freedom has a price. Anthropic notes that agents trade “latency and cost for better task performance”, with “the potential for compounding errors”. Its advice is to find “the simplest solution possible”. For most small businesses, that means a simple automation with one AI step often beats a fully autonomous agent.
Six marketing jobs agents can do today
The best jobs for an agent are narrow, repetitive and easy to check, like these six.
| Job | What the agent does | Tools that do it today | Where you stay in charge |
|---|---|---|---|
| Answer new enquiries | Replies in seconds, day or night, answers common questions, qualifies the lead and books a call | Meta Business Agent on WhatsApp, HubSpot’s Customer agent | Prices, refunds and promises go to a person |
| Repurpose content | Turns one article into social posts, a newsletter and short video scripts | ChatGPT, Claude or Gemini | Edit for your voice and check every fact |
| Write the weekly report | Pulls figures from Google Analytics, ad accounts and your CRM, and summarises what changed | Zapier Agents and similar tools | Check its numbers against your dashboards for the first month |
| Research leads | Researches target companies, fills gaps in your CRM and drafts personal outreach | HubSpot’s Prospecting agent, Zapier Agents | Read every email before it goes out |
| Run paid search | Finds new search terms, writes ad variations and picks the best landing page | Google’s AI Max for Search | You set the budget, goals and exclusions, and review search terms weekly |
| Research your market | Reads dozens of websites and writes a briefing with sources on competitors’ prices, offers and reviews | ChatGPT agent, Gemini or Claude research modes | Open the sources before you act on them |
The first job is the easiest win. In a Harvard Business Review audit of 2,241 US companies, firms that contacted a web lead within an hour were nearly seven times as likely to qualify it as firms that replied an hour later. They were over 60 times as likely as firms that waited a day, and the average reply took 42 hours. An agent replies in seconds.
Treat vendor claims as a starting point. Google says AI Max typically brings 14% more conversions at a similar cost per acquisition, based on its own 2025 data. Test it against your current campaigns.
Where agents still go wrong
Agents fail in three predictable ways, and each can be avoided.
- They state wrong things with confidence, and you are liable. In February 2024, a Canadian tribunal ordered Air Canada to pay damages after its website chatbot invented a refund rule for bereavement fares. The airline argued the chatbot was “a separate legal entity that is responsible for its own actions”. The tribunal disagreed: the company was responsible for all the information on its website.
- Cheaper is not the same as better. Klarna said its AI assistant handled 2.3 million conversations in its first month in 2024, the work of 700 agents. By May 2025, its chief executive admitted the focus on cost had lowered quality. Klarna started hiring people again so that “there will always be a human if you want”.
- Many projects never pay back. Gartner predicts that over 40% of agentic AI projects will be cancelled by the end of 2027, due to “escalating costs, unclear business value or inadequate risk controls”. It also warns of “agent washing”: of the thousands of vendors selling agents, it counts only about 130 as real.
The lesson: give an agent a narrow job, accurate facts and a person to hand over to.
The rules in the UK and Ireland
An agent doesn’t change your legal duties: whatever it says or sends counts as your business speaking. Four sets of rules matter most.
- Customer data (UK GDPR). An agent that reads enquiries, emails or your CRM is processing personal data. Say so in your privacy notice and check where the vendor stores the data. Don’t let it make decisions that significantly affect people on its own: the ICO says human review of such decisions “cannot be token gestures”.
- Marketing emails (PECR). An agent sending outreach follows the same rules you do. Individuals must have opted in, unless they are existing customers who bought something similar and were offered an easy opt-out. You can email companies, but sole traders count as individuals.
- Reviews (DMCC Act). Since 6 April 2025, the UK bans any review that “purports to be, but is not, based on a person’s genuine experience”. Never let an agent write reviews for you. The CMA can fine up to 10% of global turnover.
- Chatbots that talk to people in the EU (AI Act). Since 2 August 2026, chatbots must tell people they are dealing with an AI, unless that is obvious. The duty sits mainly with the tool’s maker, so check that yours does it, especially if you serve customers in Ireland. Fines reach €15 million or 3% of worldwide turnover.
This is a summary, not legal advice, and the ICO is updating its guidance after the Data (Use and Access) Act 2025.
How to start in 30 days
Start with one job, not a strategy. A month is enough to tell.
- Week 1: pick one job and measure it as it is. Choose the job that eats the most hours or loses the most leads, usually enquiries or reporting. Write down today’s numbers: reply time, hours spent, enquiries won.
- Week 1: write the agent’s rulebook. List the facts it may use (services, prices, your booking link) and what it must never do (quote custom prices, promise dates, agree refunds). Decide when it hands over to a person, and make sure it says it is an AI.
- Week 2: test the tools you already pay for. HubSpot and WhatsApp Business now include agents, some on paid plans only, and Google Ads has AI Max built in. Before you buy anything new, ask what the tool does without a person pressing “go”. If the answer is “nothing”, it is an automation, which may be all you need.
- Weeks 2 and 3: check every output. Read each reply, post or report before it goes out, and add a rule for every mistake.
- Week 4: compare and decide. Put the new numbers beside the old ones. Keep it, extend it to a second job or switch it off.
If you’d like a hand, our AI marketing automation service sets up agents like these with clear rules and people in charge.
Frequently asked questions
Will an AI agent replace my marketing person or agency?
No. It replaces tasks, not judgement. Deciding what to say, to whom and why still takes people who know your business. That is the lesson Klarna learned in public.
How much does an AI agent cost?
Often less than a new tool. Many come inside software you already use and are charged by usage: HubSpot’s use HubSpot Credits, and Meta’s is part of some WhatsApp Business Premium plans. Set a spending cap first.
Do I have to tell customers they’re talking to an AI?
For customers in the EU, yes: chatbots have had to say so since 2 August 2026, unless it is obvious. The UK has no AI-specific rule yet, but telling people is the honest choice and builds trust.
Sources
- Business insights and impact on the UK economy: 8 October 2026, Office for National Statistics
- Business in Ireland 2025: digital sustainability and artificial intelligence adoption, Central Statistics Office, 24 March 2026
- Use of artificial intelligence in enterprises, Eurostat, data extracted December 2025
- Building effective agents, Anthropic, 19 December 2024
- Meta’s AI agent for WhatsApp Business is now available globally, TechCrunch, 3 June 2026
- Use HubSpot’s AI agents, HubSpot Knowledge Base
- Zapier Agents, Zapier
- AI Max for Search campaigns, Google Ads Help
- Introducing ChatGPT agent, OpenAI, 17 July 2025
- The short life of online sales leads, Harvard Business Review, March 2011, with the audit figures as quoted by SmartCompany, July 2011
- BC tribunal confirms companies remain liable for information provided by AI chatbot, American Bar Association, February 2024
- Klarna is hiring customer service agents after AI couldn’t cut it on calls, Entrepreneur, 9 May 2025
- Gartner predicts over 40% of agentic AI projects will be canceled by end of 2027, Gartner, 25 June 2025
- How do we ensure individual rights in our AI systems?, Information Commissioner’s Office
- Electronic mail marketing, Information Commissioner’s Office
- No more faux five-stars: the DMCC Act bans fake reviews, CMS, 17 April 2025
- EU AI Act Article 50 transparency obligations go live, Baker Botts, 10 September 2026
Written with the assistance of AI, then reviewed and refined by the Brielgio team.