What did I say? What did I mean? Do you understand me? We ask each other those three questions constantly without noticing. Now we ask them of machines — and this episode is about what happens in the gap.

Host Bobby Brill pulls together conversations with the ServiceNow linguists, research engineers, and AI governance leads working on voice, language, and chat, including a real recording of a voice agent failing in real time. Same words. Different meaning. And a machine in the middle trying to work out which one you intended.

In this episode:
Why a single English word carries no meaning without context, and what that does to translation
How a name that isn't in the training data breaks the first model in the chain
The three-model cascade behind every voice agent, and what gets lost at each handoff
A voice agent that hears a confirmation code correctly three times and rejects it three times
Why formality — tu or vous — is a fluency problem, not a grammar problem
Who decides what's acceptable for an AI to say when the same word is fine in London and not in Chicago.

Check out our past episodes on voice and voice AI:
https://youtu.be/cGXCnABXKow
https://youtu.be/yxoHmZj5gOk
https://youtu.be/x7Ks932T18o

Guests:
Lyena Solomon, Director of Globalization and Accessibility
Midam Kim, Machine Learning Engineer
Tara Bogavelli, Research Engineer
Katrina Stankovic, Staff Machine Learning Engineer
Gabrielle Gauthier-Melançon, Staff Applied Research Scientist
Louis-Philippe Morin, AI Governance Product Manager

CHAPTERS

0:00 "Half ten" — same words, an hour apart
2:04 What did I say: the word "order"
3:19 The name a machine can't hear
5:23 Three models handing meaning to each other
8:19 Naming conventions 10:04 Listen: a voice agent fails
14:47 Tu or vous: talking to you correctly
16:39 Who decides what's acceptable
18:49 Teaching it the way you'd teach a kid
20:15 Why any of this matters

For more information, see:

ServiceNow Training and Certification: http://www.servicenow.com/services/training-and-certification.html

ServiceNow Community: https://community.servicenow.com/community

ServiceNow Insights Podcast: https://www.youtube.com/playlist?list=PLCOmiTb5WX3qvGq7Cp3o2KkCiplJyqQOK

For general information about ServiceNow, visit: http://www.servicenow.com/

#ServiceNow #VoiceAI #AI #ConversationalAI #Localization #AIGovernance

See omnystudio.com/listener for privacy information.

Why multilingual content is a business risk, a compliance question, and an AI-readiness problem - not just a translation checkbox - with ServiceNow's Lyena Solomon.

#ServiceNow #Localization #Globalization #AI

Chapters

00:00 Cold open: "half ten" and the meaning problem
00:29 Welcome + episode topic
00:53 Meet Lyena Solomon
01:12 Why this isn't just translation
03:49 The word "order" - context matters
05:05 What is language governance?
06:04 Translating "pizza"
07:03 The regulatory reality (Quebec, EU AI Act)
09:30 Self-localization: Maori and Inuktitut
13:23 The real business risk of inconsistency
15:43 AI readiness and language risk
16:28 A support ticket in three languages
20:32 It's about trust, not just translation
21:11 Closing thought: the joy of understanding
22:17 Wrap-up + subscribe

For more about ServiceNow - https://www.youtube.com/@ServiceNowDocs
To watch these episodes on YouTube - https://www.youtube.com/watch?v=yxoHmZj5gOk&list=PLCOmiTb5WX3qvGq7Cp3o2KkCiplJyqQOK

See omnystudio.com/listener for privacy information.

Voice AI sounds simple until you try to deploy it at scale — across airports, accents, languages, and thousands of employees at once.

In this episode of ServiceNow Insights, host Bobby Brill sits down with Midam Kim, an ML engineer and linguist at ServiceNow, to unpack what it actually takes to build voice AI as an enterprise product. From the out-of-vocabulary problem (why AI still struggles with names) to why turn-taking in conversation is a linguistic skill most people never think about, Midam breaks down the human science behind the technology.

In this episode:

- Why voice AI is replacing typing as the default way to interact with enterprise systems

- The difference between building for employees (B2B) vs. building for their customers (B2B2C)

- Why an airport is one of the hardest possible environments for voice AI — and what ServiceNow does about it

- The "out-of-vocabulary" problem: why AI still struggles with names, accents, and rare expressions

- Why ServiceNow's secret sauce is hiring linguists, not just engineers

- The linguistic framework behind every voice interaction: sounds, words, and turn-taking

- Why voice AI is like teaching a kid to speak for the first time

Chapters
00:00 — Welcome to ServiceNow Insights
00:22 — Meet Midam Kim, ML Engineer & Linguist
00:35 — Why voice is replacing typing
01:48 — What voice AI actually does for employees
03:40 — B2B vs. B2B2C: who's really using this?
05:11 — Desk employee vs. airport traveler: two different problems
06:42 — Building for an ever-changing environment
08:53 — Why airports are the hardest use case
09:54 — Accents, fluency, and the diversity problem
11:20 — "My Name Is. My Name Is. My Name Is." — the OOV problem
12:50 — The coffee shop name story
13:20 — How ServiceNow trains its models
14:59 — The 3 linguistic layers: sounds, words, interaction
16:38 — Midam's turn-taking story from Korea
18:33 — Why voice agents can't be "that person you avoid"
20:16 — "We can make it great"

Subscribe for more ServiceNow Insights episodes on AI, voice technology, and enterprise innovation.

Related episode: Voice AI Agent Evaluation — how ServiceNow measures whether voice AI meets human expectations. https://youtu.be/x7Ks932T18o

For more about voice in AI from Midam Kim - https://youtu.be/3NUf6W_FMWs?is=wGc7BfyhiDp8JlOW

#VoiceAI #EnterpriseAI #ServiceNow #ArtificialIntelligence #Linguistics #ConversationalAI #AIProduct #Podcast

See omnystudio.com/listener for privacy information.