If 2026 was the year AI became normal, the next few years will be about AI that does things, not just answers questions. Instead of asking a chatbot how to rebook a delivery, you'll ask an assistant to rebook it — and it will. This article looks at where AI is heading in everyday software, what people actually say they are comfortable with, and what it means for businesses.
From answers to actions: the rise of agents
An AI agent is software that can plan and carry out a task across several steps: checking information, filling in forms, comparing options and completing an action, usually with your approval at key points.
We're already seeing early versions:
- Developers delegate multi-step coding tasks to agents — Stack Overflow's 2026 survey found about a quarter of developers already use automated agent workflows.
- Search is becoming task-based. At I/O 2026 Google said AI Mode had passed one billion monthly users and announced search agents and agentic shopping features.
- Shopping assistants can now find, compare and, with permission, buy products.
Shopping by conversation
Retail is one of the first places agents will be felt by ordinary people. UK research this year gives a useful picture of the appetite — and the caution:
- In research by Commerce and PayPal, 64% of UK online shoppers said they were interested in trying agentic AI shopping assistants, but only 21% currently use AI for shopping. Their biggest worry was AI buying without approval.
- Checkout.com found that many Brits would be comfortable letting an agent handle routine, "boring" purchases — the weekly essentials rather than the special ones.
- A multi-country survey by Global Payments found consumers now expect AI to make 15% of their purchases within five years — yet a third want to approve every transaction before payment is made.
The message is consistent: people like the idea of delegation, as long as they stay in control. Expect agents that propose, and humans that approve — at least for now.
AI inside every app
The other big shift is quieter. Rather than visiting a separate AI tool, assistance will be built into the software we already use:
- Email and documents that draft, summarise and organise automatically.
- Accounting and admin tools that read receipts, chase invoices and flag anomalies.
- Business systems that answer questions about your own data in plain English.
- Phones and cars with assistants that understand context across apps.
For most people this won't feel like "using AI" at all. It will feel like software that finally does the boring part for you.
What changes at work
AI is unlikely to remove the need for people who understand a business, its customers and its trade-offs. What it changes is where time goes:
- Less time on retyping, searching, formatting and first drafts.
- More time on judgement, relationships, checking and exceptions.
- New skills: describing tasks clearly, checking AI output, and knowing when not to use it.
The businesses that benefit most will be the ones that redesign processes around this — not the ones that bolt a chatbot onto an unchanged workflow.
Trust, safety and rules
As AI acts on our behalf, trust becomes the central question:
- Approval and limits — spending caps, confirmation steps and clear audit trails.
- Data protection — UK GDPR still applies: what data an AI system sees, where it is processed and how long it is kept.
- Transparency — knowing when you're dealing with AI, and why it recommended something. Research from Commerce and PayPal found many UK shoppers object to brands paying for preferential treatment inside AI tools.
- Regulation — the UK has taken a principles-based approach through existing regulators, while the EU's AI Act is phasing in obligations. Businesses selling into Europe should keep an eye on both.
How businesses can prepare now
- Make your information machine-readable. Clear product data, prices, stock, opening hours and policies help both search engines and AI agents represent you correctly.
- Offer simple, secure ways to transact. Express payments, clear delivery options and straightforward returns make it easier for people — and their assistants — to buy from you.
- Automate one internal process. Start where staff retype or search the most, keep a human in the loop and measure the result.
- Set ground rules. A short AI policy covering approved tools, data and responsibility.
- Keep the human touch where it matters. Complex sales, complaints and relationships still win on people.
Summary
The next stage of AI is software that acts: agents that complete tasks, assistants built into everyday apps and shopping by conversation. People are interested but want control, so the winners will combine automation with clear approvals, good data and trust.
If you're thinking about how AI could fit into your own systems, read how AI changed software in 2026, explore our custom SaaS solutions or get in touch.
Research figures as published up to October 2026.