Octopus Energy's AI email assistant Arlo scored higher satisfaction than human replies

UK energy supplier Octopus trialled Arlo, an AI that answers routine customer emails. In a three-month trial it handled about 8,000 emails a week and scored 76% satisfaction, versus 72% for comparable human replies.

Organization: Octopus Energy
Industry: Energy
When: Jul 2026
Evidence: Company reported
Agents and tools: Arlo Kraken

Reported outcome

76% customer satisfaction for AI replies vs 72% for comparable human replies, on ~8,000 emails a week (4% of UK customer emails).

76%
CSAT, AI replies
72%
CSAT, human replies
~8,000
Emails per week
4%
Share of UK emails

Caveats

Stated by the company that uses the agent (blog, earnings call, interview). Not independently audited.

Arlo only handles routine questions. Vulnerable customers, sensitive cases, and complex complaints go straight to people, so the comparison is not like-for-like across all emails.
Earlier press reports gave much larger figures (about a third of emails, '250 people's worth of work'). Those do not match this later trial data.

What the agent does

Arlo drafts and sends answers to simple customer emails, such as tariff renewals, payment dates, and account details. It runs on Octopus's Kraken platform, so it can see the customer's account. Customers can always reach a person.

Why the numbers differ

In 2024, the CEO said AI was doing the work of about 250 people. The 2026 trial reports a smaller, more careful scope: 4% of emails. Read both, and trust the one with a clear method.

What to copy

Start with email, not live chat. Email gives the agent time and gives you a review step.
Compare CSAT on the same type of question for AI and humans.
Write down which customers the AI must never handle.

Starter prompt

Written by us, not by Octopus.

Customer email: {email}
Account data: {account_json}

If the email is about tariff renewal, payment dates, or account details,
draft a reply under 120 words using only the account data.

If the customer mentions financial hardship, health, a complaint, or
seems distressed, do not draft a reply. Output: ESCALATE + one-line reason.