Two thirds of UK online shoppers expect a response within two hours. 65 per cent of e-commerce support queries can be resolved by an AI agent without a human ever touching the ticket. And 80 per cent of routine customer interactions are expected to be handled by AI by the end of 2026. If you are running a UK e-commerce business on five to fifty staff and still answering every returns query by hand, you are spending money you do not need to spend. The tools exist. The question is which one, at what cost, and what the Consumer Rights Act says you have to get right.
What AI Customer Support Actually Means in 2026
AI customer support in 2026 is not a chatbot that says "I'm sorry, I didn't understand that" and routes you to a human. The current generation of AI agents — Freshdesk's Freddy AI, Zendesk's AI agents, and Intercom's Fin — read your help centre articles, your order data, and your return policies, then resolve tickets end to end. They draft replies, process refunds, update order statuses, and escalate to a human only when they genuinely cannot handle the query.
The pricing models have changed too. Zendesk now charges per resolution rather than per seat for its AI agents. Intercom charges $0.99 per resolution. Freshdesk bundles AI sessions into its paid tiers with optional top-ups. The economics are completely different from the old per-agent model, and for a small e-commerce team, that matters.
How the Three Platforms Compare on Pricing
| Feature | Freshdesk (Freddy AI) | Zendesk AI Agents | Intercom Fin | --- | --- | --- | --- | Base plan (per agent/month) | Free (up to 10 agents), Growth £12, Pro £39, Enterprise £65 | Suite Team £15, Professional £95, Enterprise £140 | Starter £32/seat, Growth (custom) | AI pricing model | Sessions (bundled + top-up) | Per resolution ($1.50 committed, $2.00 pay-as-you-go) | Per resolution ($0.99) | AI sessions included | 500 with Pro and Enterprise | None — separate AI add-on | None — per resolution from first query | AI top-up cost | $100–$1,000 per 1,000 sessions | Advanced AI add-on $50/agent/month | Included in per-resolution fee | UK data residency | EU (Frankfurt) | EU (Frankfurt) | EU (Dublin) | Claimed resolution rate | Not published | Not published | 60% average, 82% with Fin 2 | SOC 2 certified | Yes | Yes | Yes | GDPR DPA available | Yes | Yes | Yes |
The pricing difference matters more than it looks. A 20-person e-commerce team handling 3,000 tickets a month with a 60 per cent AI resolution rate would see 1,800 tickets handled by AI. On Intercom Fin, that costs $1,782 per month in resolution fees alone, before the base plan. On Zendesk committed pricing, the same 1,800 resolutions cost $2,700. On Freshdesk Pro with 500 bundled sessions and a 1,000-session top-up, the cost is roughly £39 per agent plus $500 — but you only cover 1,500 sessions, so you need a second top-up.
The cheapest option depends entirely on your ticket volume and resolution rate. There is no universal answer.
What the Consumer Rights Act 2015 Requires
The Consumer Rights Act 2015 applies to AI-assisted customer support the same way it applies to human support. Digital services — and that includes an AI chatbot handling returns — must be as described, fit for purpose, and delivered with reasonable care and skill.
In practice, this means three things for your AI agent:
First, the AI must not mislead customers. If a customer is talking to an AI agent, they need to know it. The Online Safety Act now covers AI chatbot providers serving UK users, and the Competition and Markets Authority expects transparency about whether a response is generated by a human or a machine.
Second, the AI must give accurate information about consumer rights. If a customer asks about their right to a refund within 14 days of an online purchase (the Consumer Contracts Regulations 2013 cooling-off period), the AI must get this right. Training the AI on your actual returns policy is not optional — it is a legal requirement.
Third, there must be an escalation path to a human. An AI agent that refuses to connect a customer to a person, or that loops endlessly without resolution, does not meet the standard of reasonable care and skill. Every platform listed above supports human handoff. The question is whether you have actually configured it and staffed the human queue.
The Resolution Rate Problem
Intercom claims a 60 per cent average resolution rate for Fin, rising to 82 per cent with Fin 2. Zendesk and Freshdesk do not publish comparable figures.
These numbers need context. A "resolution" in outcome-based pricing means the AI handled the conversation and the customer did not reopen the ticket within 72 hours (Zendesk's definition). That is not the same as the customer being satisfied. It is not the same as the issue being fixed. It means the ticket stayed closed.
For a UK e-commerce business, the queries that AI handles well are predictable: where is my order, how do I return this, what is your refund policy, do you ship to my postcode, can I change my delivery address. These account for 60 to 70 per cent of inbound volume in a typical online retail operation.
The queries that AI handles badly are also predictable: complaints about damaged goods where the customer is angry, disputes about partial refunds, warranty claims that require judgement, and anything involving the phrase "I want to speak to a manager." These must go to a human. Configuring your escalation triggers properly is worth more than any model upgrade.
Freshdesk Freddy AI — What You Get
Freshdesk splits its AI into three components. Freddy AI Agent is the customer-facing bot that resolves tickets. Freddy AI Copilot sits beside your human agents and suggests replies, summarises conversations, and drafts responses. Freddy AI Insights gives managers analytics on resolution rates and agent performance.
The Free tier supports up to ten agents with basic ticketing but no AI. The Growth tier at £12 per agent per month adds automation rules. The Pro tier at £39 per agent per month includes 500 AI sessions and the Copilot. Enterprise at £65 per agent per month adds custom objects and advanced security.
Freshdesk's strength for UK e-commerce SMBs is cost. A five-person support team on Pro costs £195 per month plus AI session top-ups. The weakness is that Freshdesk's AI is session-based, not resolution-based — you pay per conversation started, regardless of whether the AI actually resolved it.
Zendesk AI Agents — What You Get
Zendesk moved to outcome-based pricing in 2025. You pay per automated resolution — $1.50 if you commit to a volume, $2.00 on pay-as-you-go. A resolution is confirmed when the customer's issue is handled and the conversation closes without reopening within 72 hours.
The base Suite plans still require per-agent licensing. Suite Professional at £95 per agent per month is the minimum tier that supports the full AI agent capability with the Advanced AI add-on at $50 per agent per month.
Zendesk's strength is how well it connects with e-commerce platforms — Shopify, WooCommerce, Magento — and its ability to pull order data directly into the AI conversation. The weakness is cost. For a small team, the per-agent licensing plus per-resolution fees add up faster than the alternatives.
Intercom Fin — What You Get
Intercom Fin charges $0.99 per resolution on top of base plans starting at £32 per seat per month. Fin reads your help centre, your product documentation, and your custom data sources, then answers customer questions in a conversational style.
Intercom claims Fin handles one million customer issues per week across its user base. The 60 per cent average resolution rate means four out of ten queries still go to a human. Fin 2, released in late 2025, claims 82 per cent — but that figure comes from Intercom's own reporting and has not been independently verified.
Intercom's strength is the conversational experience. It was built as a messaging platform first, and the AI feels native to chat. The weakness for UK e-commerce is that Intercom is aimed at SaaS and tech companies — its e-commerce integrations (Shopify, WooCommerce) exist but are less mature than Zendesk's.
Ten-Point Checklist for Getting Your First AI Support Agent Live
1. Audit your ticket categories — identify the five to ten query types that account for 70 per cent of your volume. 2. Write or update your help centre articles to cover those top query types completely and accurately. 3. Check your returns policy against the Consumer Contracts Regulations 2013 — your AI will parrot whatever you train it on. 4. Choose your pricing model: per-session (Freshdesk), per-resolution (Zendesk/Intercom), or blended. 5. Set up human escalation triggers — angry customer detection, refund disputes over a threshold, and any query mentioning legal action. 6. Add a clear disclosure that the customer is talking to an AI agent — this is a legal requirement under the Online Safety Act. 7. Test with your 50 highest-volume real tickets before going live — measure resolution accuracy, not just resolution rate. 8. Set a weekly review cadence: check what the AI got wrong, update training data, adjust escalation rules. 9. Confirm your data processing agreement is in place — all three platforms offer GDPR DPAs, but you need to sign yours. 10. Start with one channel (live chat or email) and expand only after you have a week of clean data.
