AI follow-up in 2026: texts, WhatsApp, AI calls and when to hand over

You are a customer service lead for a UK [trade] firm. Write the first message [company] sends within a minute of a new quote request, so the person answers when we call.

About us: we cover [area], we call from [number], the person who calls is usually [name], and we text between [hours and days].

Write:
1. A text for working hours: their first name, my name and company, that I'll call in the next few minutes, and the number I'll call from. Under 160 characters.
2. A text and an email for requests outside our texting hours: we've got it, and when exactly we'll call.
3. A version to submit to Meta as a WhatsApp template, with the variables marked like {{1}}, for people who ticked a WhatsApp opt-in that names us.
4. A "sorry we missed you" text for after a missed first call.

Rules: keep every message about their request. No offers, discounts or links to deals (one offer turns it into marketing that needs consent). No fake urgency. Plain UK English.

You are setting up an AI assistant that replies to new enquiries for [company], a [trade] firm in [area], by [text / WhatsApp]. Write the full instructions I'll give the assistant.

It must:
- say in its first message that it's an automated assistant for [company],
- answer only these questions: [areas covered, how long a survey takes, what happens next, opening hours],
- ask at most [number] short questions: [e.g. home owner? what they need? best time to call?],
- offer a call or survey from these slots: [slots],
- hand over to a person straight away when someone asks for one, is upset or confused, mentions price or finance, sounds like a complaint, or may be a vulnerable customer,
- stop messaging at once if they say stop, and confirm it has stopped.

It must never: pretend to be a person, quote prices, promise savings or grants, give advice on finance, or message anyone who hasn't opted in.

Then write 10 test conversations to run before going live, including: "are you a real person?", "how much?", "stop", someone angry about a past job, and someone who seems confused. Show the reply you'd expect for each.

This is not legal advice: check the setup with a compliance adviser before you switch it on.

You are a careful reviewer of UK marketing consent wording (PECR and UK GDPR). Here is the consent wording and tick boxes on our quote form, and the page it sits on: [paste wording and page URL or text].

We want to contact people by: [calls / texts / email / WhatsApp / automated or AI calls]. Our business name is [company], and [we are the only firm who will contact them / a partner will also contact them: name].

Tell me, for each channel:
1. Whether the wording clearly covers it, and why or why not.
2. Whether it names the business that will make contact.
3. What's missing or vague (e.g. "selected partners", no mention of automated calls).

Then suggest clearer wording, as separate optional ticks where a channel needs its own consent, and list what we should store with each consent (time, page, wording version).

This is not legal advice. Flag anything I should check with a compliance adviser.

You are an operations adviser for a UK [trade] firm. Here's how we follow up new enquiries today: [describe every step, who does it, and when]. We get about [number] leads a week, and [number] people can call. Our hours are [hours].

Produce a table with every step, and for each one: "let AI do it", "keep a person" or "never automate", with a one-line reason.

Use these rules:
- AI can send the first reply, answer simple questions, offer call times and send reminders.
- A person handles the first real conversation, price, finance, objections, anyone upset or asking for a person, complaints and vulnerable customers.
- Never: AI voice calls without consent for automated calls, a bot pretending to be a person, messaging after someone says stop.

Then tell me the three changes that would cut our time to first contact the most, in order, and what each needs (tools, consent, people).

Here is a conversation between our AI assistant and someone who asked [company] for a quote: [paste].

Mark it, in short points:
1. Did it say it was automated in the first message?
2. Did it stay on their request, with no offers or pressure?
3. Did it hand over to a person at the right moment, or too late?
4. Did it say anything that could read as a promise (price, savings, dates, grants)?
5. Did it stop when asked?

Then rewrite any weak replies, and suggest one change to the assistant's instructions so it doesn't happen again.