Why is my AI receptionist reading my business name incorrectly when it answers calls?

Your AI receptionist is mispronouncing your business name because the speech engine reads the spelling — not the sound. It doesn't know your brand. You have to teach it.

Here's the fix, platform by platform.

Why This Happens in the First Place

Most people assume the AI is "smart enough" to figure out how to say the name. It won't.

The default speech engine processes text the same way a stranger reads it cold — no context, no brand history, no grace. Acronyms get spelled out letter by letter when they should be read as a word. Non-English words get anglicized. Unusual spellings get punished. If your name has any of those three things going on, you're going to hear something wrong on every single call.

The system isn't broken. It just hasn't been told.

Fix #1 — Phonetic Spelling in the Name Field

This is the fastest move and it works on most platforms.

Instead of entering your legal business name exactly as it's registered, enter how it sounds — letter by letter, syllable by syllable.

How to do it:

Go to your AI receptionist's company profile or greeting setup field.

Replace the correct spelling with a phonetic approximation.

Use hyphens between syllables. Add punctuation where you need a pause.

Example: A setup documented with the name "Bedrock Digital Agency" used letter-by-letter spelling plus added punctuation to force the system to process the greeting correctly during call transfers. The AI stopped improvising once it had nothing to improvise from.

Works well for:

Names with silent letters

Acronyms that should be read as words (or words that should be read as acronyms)

Founder-name-based brands with non-standard spellings

Fix #2 — Use the Platform's Dedicated Pronunciation Field

Some platforms build this correction directly into the settings. Don't ignore it.

Platforms like RingCentral ACE include a specific field for defining custom pronunciations for individual words. That field exists for exactly this reason — they know the speech engine gets names wrong.

How to do it:

Look in your platform's voice or greeting settings for a "pronunciation" or "custom word" field.

Enter your business name and the phonetic version side by side.

Save and test with a live call.

If your platform has this field and you haven't used it, that's the first place to go. Don't build a workaround when the door is already open.

Fix #3 — Rewrite the Instruction Prompt

If your AI receptionist runs on a prompt — and most do — the name lives in that prompt. That's where you fix it.

The instruction prompt is doing most of the heavy lifting in how your AI behaves. If the business name is spelled correctly in that prompt but sounds wrong out loud, the fix is simple: write it phonetically inside the prompt.

How to do it:

Open your AI's instruction or "playbook" setup.

Find every instance of your business name.

Replace with the phonetic version — or add a parenthetical note directly in the prompt, like: "Always say the business name as 'Braz-ee-el Group' — not 'Brazil Group'".

Some platforms let you add a specific pronunciation rule as a playbook instruction. Use it.

One more thing to check: If your AI is still reverting after you've updated the prompt, it may be pulling the name from a separate extension or database field that overrides your custom instruction. Check every field where the name appears — profile settings, extension labels, greeting templates — not just the prompt.

Fix #4 — SSML Tags (for Developers and Custom Builds)

If you have access to the underlying voice configuration — or your platform supports Speech Synthesis Markup Language — SSML tags give you surgical control.

SSML is the language that tells text-to-speech engines exactly how to speak. It's not just for developers. Some AI receptionist builders expose it in advanced settings.

The tags you need:

xml

<phoneme alphabet="ipa" ph="ˈbræz.i.əl">Brazziel</phoneme>

Or, if IPA feels like too much, use the simpler alphabet="x-sampa" version or just the say-as tag to control how the engine interprets the string:

xml

<say-as interpret-as="spell-out">BMW</say-as>

When to use this:

Custom-built AI voice flows (Twilio, Google CCAI, HoundCatOs, Amazon Lex)

Platforms that expose raw SSML in their prompt or TTS settings

When phonetic spelling alone isn't giving you the exact sound

If you're not sure whether your platform supports SSML, search the help docs for "SSML" or "speech markup." If it's there, it's usually buried in advanced voice settings.

Fix #5 — Escalate to Support (With the Right Information)

If you've tried the above and the AI is still mispronouncing your name, escalate — but escalate smart.

Support teams move faster when you hand them everything upfront. Don't send a vague "it's saying my name wrong" ticket.

What to include in your support request:

The exact business name as currently entered in every field

A phonetic spelling of the correct pronunciation (write it like you'd explain it to someone who's never heard it)

A recording of the AI saying it wrong — screen record a test call

The specific fields you've already tried to update

Whether the issue happens on every call or only on transfers/specific triggers

That last point matters. If the mispronunciation only happens on call transfers, the name is being pulled from an extension label — not the greeting prompt. That's a different fix than what most support reps will assume.

Platform-Specific Notes

| Platform | Where to Fix It | SSML Support? |

|---|---|---|

| RingCentral ACE | Dedicated pronunciation field in AI settings | Check advanced voice config |

| Jobber AI Receptionist | Company profile + greeting prompt | Not publicly documented |

| Custom builds (Twilio, HoundCatOs, Google CCAI) | SSML in TTS config | Yes — full support |

| Generic prompt-based builders | Instruction/playbook prompt | Varies by builder |

Always test with a live call after every change — not just the preview.

The Move You Can Make Today

Open your AI receptionist settings right now. Find every field where your business name appears. Write it phonetically in each one. Make a test call.

If it's still wrong after that, escalate with the five-point checklist above.

Your AI is only as good as what you put in. It's not going to figure your name out on its own — but once you tell it, it won't forget.

FAQ

Q: Why does my AI say my business name correctly in the preview but wrong on actual calls?

A: The preview and the live call may be pulling from different fields. The preview uses what's in the greeting or prompt; live calls — especially transfers — may pull from an extension label or database field. Check both.

Q: What's the easiest phonetic spelling method if I don't know IPA?

A: Write it the way you'd text it to a friend who's never heard your name. Break it into syllables with hyphens, capitalize the stressed syllable, and use common English letter combinations for the sounds. Test it, adjust, repeat.

Q: Does phonetic spelling in the prompt affect anything else the AI says?

A: Only if the phonetic spelling creates a word the AI tries to interpret in other contexts. Keep the phonetic version isolated to the greeting or name field when possible — don't replace every instance of your name in a complex prompt without testing.

Q: My platform doesn't have a pronunciation field. What do I do?

A: Rewrite the instruction prompt with phonetic spelling first. If that doesn't work, submit a feature request (RingCentral ACE has an active ideas board for exactly this) and escalate to support with a recording. Document everything.

Q: How do I know if my platform supports SSML?

A: Search your platform's help docs for "SSML" or "speech synthesis markup." If you're on a custom build using Twilio, HoundCatOs, Google CCAI, or Amazon Lex, SSML support is built in. Most out-of-the-box AI receptionist tools have limited or no SSML access for end users.

FAQ

Why does my AI say my business name correctly in the preview but wrong on actual calls?

The preview and the live call may be pulling from different fields. The preview uses what's in the greeting or prompt; live calls — especially transfers — may pull from an extension label or database field. Check both.

What's the easiest phonetic spelling method if I don't know IPA?

Write it the way you'd text it to a friend who's never heard your name. Break it into syllables with hyphens, capitalize the stressed syllable, and use common English letter combinations for the sounds. Test it, adjust, repeat.

Does phonetic spelling in the prompt affect anything else the AI says?

Only if the phonetic spelling creates a word the AI tries to interpret in other contexts. Keep the phonetic version isolated to the greeting or name field when possible — don't replace every instance of your name in a complex prompt without testing.

My platform doesn't have a pronunciation field. What do I do?

Rewrite the instruction prompt with phonetic spelling first. If that doesn't work, submit a feature request and escalate to support with a recording. RingCentral ACE has an active ideas board for exactly this type of issue. Document everything.

How do I know if my platform supports SSML?

Search your platform's help docs for 'SSML' or 'speech synthesis markup.' If you're on a custom build using Twilio, Google CCAI, or Amazon Lex, SSML support is built in. Most out-of-the-box AI receptionist tools have limited or no SSML access for end users.