Agentic commerce: when AI agents shop for customers
What agentic commerce means, how AI assistants are starting to research, compare and buy on behalf of people, and how online stores can prepare.
How AI agents that browse and act on websites work, what they can do for users and businesses, the security risks and how to make your website agent-friendly.
AI browser agents can navigate websites, fill in forms, compare options and complete tasks on behalf of users. For businesses, they create opportunities to automate web-based work and new ways customers reach your site. Websites that are well structured, accessible, fast and clear about policies are easier for agents to use. Agents also bring security risks, including manipulation through hidden instructions on web pages.
AI assistants can now operate web browsers: clicking links, reading pages, typing into forms and completing multi-step tasks. Major AI companies have released agent features and AI-powered browsers, and capabilities are improving.
Most require user permission for sensitive steps, such as payments or submitting personal information.
Agents can handle repetitive tasks in web-based tools without APIs, such as updating supplier portals, under supervision.
Some customers will ask agents to find, compare or book services. Websites that agents can understand and use may benefit. Our guide to agentic commerce covers shopping agents.
Many practices that help agents also help people and search engines:
Accessibility work, such as labelling buttons and form fields, makes sites easier for both assistive technologies and AI agents.
Malicious pages can hide instructions that try to manipulate agents into revealing data or taking harmful actions. Agents and their users need protections.
Agents acting on behalf of users raise questions about authentication, authorisation and fraud.
Some agent traffic may resemble bots. Balance blocking abusive automation with allowing legitimate agents and assistants.
Our guide to AI cyber threats covers wider security considerations.
Analytics may show visits from AI agents and assistants. Track referrals and watch for changes in how customers arrive and convert.
Standards for how agents interact with websites, identify themselves and handle payments are developing. Follow updates from your platforms and providers.
An insurance broker spent hours each week copying policy details from several insurers' web portals into its own system. Some portals had no API. The broker piloted a supervised browser agent that logged into a test account, navigated to policy pages, extracted details and prepared updates for staff to review. The agent used a dedicated account with limited permissions, and every action was logged. After proving reliable, the agent handled routine updates, while staff dealt with exceptions.
Customers may ask agents to perform tasks on your website. Clear error messages and confirmation pages help both people and agents complete tasks correctly.
Some agents identify themselves, while others look like normal browsers. Monitor unusual patterns and adjust protections carefully.
Document which tasks agents perform, who supervises them, what accounts they use and how actions are reviewed. Revoke access promptly when agents are no longer needed.
Good candidates are repetitive, rule-based and low risk, such as gathering information or filling standard forms. Avoid tasks involving payments, legal commitments or sensitive data until agents are proven reliable and properly supervised. Start with read-only tasks before allowing agents to submit anything.
Browser agents can be slower and more expensive than direct integrations, because they process pages visually and step by step. Where a supplier offers an API, it's usually more reliable and cheaper. Use browser agents for gaps where no better connection exists, and compare costs against the staff time saved.
Our web development team builds accessible, structured websites that work well for people and agents, our AI solutions team designs supervised agent automations and our web app developers provide APIs so agents don't need to scrape.
An AI assistant that can navigate websites, fill forms and complete tasks on behalf of a user.
Block abusive automation, but consider allowing legitimate agents that bring customers.
Use clear structure, accessible forms, structured data and explicit policies.
This guide is general information, not legal, tax or financial advice, and details such as prices, features and rules can change. Product and company names mentioned are trademarks of their respective owners and are used only to identify those products; UK Technology Hub is not affiliated with or endorsed by them.
Tell us about your project and get a free, fixed quote.
Get a free quoteWhat agentic commerce means, how AI assistants are starting to research, compare and buy on behalf of people, and how online stores can prepare.
How criminals use AI for phishing, deepfakes, malware and automated attacks, and the practical defences small and mid-sized businesses should put in place.
How large language models like the ones behind ChatGPT, Claude and Gemini actually work, what they're good at, where they fall short and how businesses use them.