01
Sets a voice, not just a politeness dial
You choose how the assistant writes, from formal and precise to warm and conversational. The setting shapes how answers are worded, across every answer it gives.
AI Engine
A brand voice you set, with guardrails that tone cannot override.
Set how your assistant sounds to customers, from formal and precise to warm and chatty. Each channel can sound different, so the chat window on your site need not read like the summary your team picks up.
What it does
Built for the questions customers ask before they contact you.
02
Sounds different where it should
How it sounds is set per channel, because the audiences differ. The chat window on your site can sound like your front-of-house, while you keep the handoff channel plain and factual for the team picking the enquiry up.
03
Changes the wording, not the claim
Facts come from your own pages whatever tone you pick. A friendlier voice rewrites the sentence; it does not add an offer, a price or a policy that is not in your approved content.
04
Covers the first message, not just the replies
The opening line a customer sees when the chat window appears is part of what the assistant says, so it shows up in the review step alongside everything else you approve. If it sounds wrong, that is where you catch it, before any customer reads it.
Setup
How you switch it on.
01
Connect your website
Point Adoomi at your website, FAQs and approved content. Every fact the assistant states comes from there whichever voice you pick, so the source is the thing worth getting right first.
02
Set how it should sound
Decide how you want the assistant to come across, then set the voice in the bot's settings. Deciding it before you start makes the review quicker, and a note to yourself is enough to work from: friendly and brisk, first names, short sentences, never jokes about safety. Where the reader is different, set that channel differently, since your team reads handoff summaries and your customers do not.
03
Review the answers before go-live
Voice settings are part of the same owner review as the content itself. Read the answers the assistant proposes, in each language you serve, and change the voice or the underlying page wherever a phrasing lands wrong.
04
Publish, then keep adjusting
Add one line to your website to go live. The whole setup, from connecting the site to publishing, takes about 10 minutes, and the Free plan starts at £0 with 50 message credits. The voice can be changed later and reviewed the same way. Start free and hear it for yourself.
Works with
Named integrations — and their limits.
Website chat window
The voice you set applies to every live answer in the chat window on your site. The limit: it shapes wording only, so a warmer tone cannot make a thin page say more than it says.
WhatsApp handoff
A conversation can move to WhatsApp carrying the customer's original question and the contact details lead capture has taken. The limit: the summary your team reads is plain because you set that channel to sound plain during review, not because the customer-facing voice is dropped automatically.
Human handoff
When the assistant is unsure, or the question is a sensitive one, the conversation routes to a named person. The limit: no tone setting moves that threshold, so a confident-sounding voice still hands over at exactly the same point.
Multilingual answers
Customers can ask in their own language and get an answer in it, 24/7. The limit: do not assume a tone survives translation, so review the wording in each language you serve as part of the same review step, especially anything humorous or very informal.
Worked example
One question, two voices, one source page
CustomerAre you open after work? I don't finish until 6.
AdoomiFormal voice: Thank you for your enquiry. We open later on Thursday evenings, so an after-work visit is possible. If you would like a time confirmed, I can pass your details to the team.
AdoomiFriendly voice: Yes, Thursdays we're open late, so after work is no problem. Want me to take your details and have someone confirm a time for you?
Your teamHandoff summary, set to sound plain and factual for the team: customer asking about after-work visits, opening-hours page shows later opening on Thursday, details taken after the customer agreed, wants a time confirmed.
Both answers were written from the same opening-hours page. Only the voice setting differs, and the fact does not. Neither version invents an evening the page does not state: if the page were silent on late opening, both voices would hand the question to a human.
Honest limits
What it doesn't do.
The assistant will not impersonate a real person
You can give the business its own voice. You cannot have the assistant present itself as a named individual, so customers are never left believing they are speaking to a specific member of staff.
Tone cannot rescue weak source content
A polished voice makes a vague answer sound polished and still vague. Where a page does not state your prices, hours or policy clearly, the fix is the page, because no voice setting turns thin content into an authoritative answer.
Changing the tone does not widen what it may answer
Voice and permission are separate settings. Making the assistant sound more confident does not extend its scope: out-of-scope, unsupported and sensitive questions still route to a human exactly as before.
In regulated sectors, some tones are off the table
Where a sector governs how something may be phrased, not only what may be said, you may find some phrasings are not appropriate for some answers. Review is where you find that out: you may need to reword it, or route that answer to a person.
Check how it sounds in every language you serve
Humour, slang and very casual wording are the least likely to land the same way in another language, and nothing guarantees a tone survives translation intact. Read those answers language by language rather than assuming.
FAQ
Questions buyers ask about voice and tone.
Can a friendlier tone make the assistant say things it otherwise would not?
Tone and permission are separate settings, so no: the boundary is the same whichever voice you choose. Nor does tone change where your data sits. Hosting stays in the EU, GDPR export and deletion are built in, and the chat window sets no tracking cookies.
Can the chat window and the handoff summary sound different?
Voice is set per channel, so yes, and the plain summary your team reads is a choice you make rather than an automatic one. The usual arrangement is a customer-facing voice in the chat window and plain, factual wording on the handoff, because the two readers need different things.
How does the voice work in other languages?
Multilingual answers go through the same review step: read the wording in each language you serve before go-live and change anything that lands wrong. Humour and very informal wording is worth checking hardest, since casual phrasing does not always survive translation intact.
Can the assistant write as a named person from our team?
Adoomi will not impersonate a named real person. You can give the business its own voice, but the assistant cannot present itself as a specific individual a customer might believe they are talking to.
Will a better tone improve answers on pages that are thin or vague?
Voice settings change phrasing, not substance, so a vague page answered in a polished style produces a polished but still vague answer. Where customers keep asking something your pages do not state clearly, update the page, because the assistant hands over rather than filling the gap.
Does a chattier voice cost more to run?
Adoomi bills by message credit, and one credit is one AI answer, so the tone and length of the wording do not change what an answer costs. The pricing page lists the ladder: Free £0 for 50 credits, Local £8 for 200, Growth £20 for 500, Scale £75 for 4,000.
Related
Useful next product pages.
Launch
Put this feature behind real customer conversations.
Adoomi starts with your public business knowledge and routes the moments that need a person to your team.