- Everyday AI
- Posts
- Ep 825: New: OpenAI Presence. Has The AI Customer Service Takeover Finally Arrived?
Ep 825: New: OpenAI Presence. Has The AI Customer Service Takeover Finally Arrived?
OpenAI brings 'Jarvis' voice control, Amazon lays off AGI staff, Claude improves voice mode, ChatGPT Health releases to all and more
Everyday AI
July 23, 2026
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Outsmart The Future
Today in Everyday AI
8 minute read
🎙 Daily Podcast Episode: OpenAI's new Presence platform helps businesses deploy AI customer service agents. Here's how it works. Give today’s show a watch/read/listen.
🕵️♂️ Fresh Finds: Microsoft releases its best AI image model, Claude brings more security to Claude Code, Cursor's new router claims 60% savings and more. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: AMD landed a major Anthropic deal, Amazon cut jobs in its AGI division, and Alphabet is boosting AI spending. And more. Read on for Byte Sized News.
💪 Leverage AI: Is OpenAI’s Presence the AI voice wave that we were promised years ago but never happened? Keep reading for that!
↩️ Don’t miss out: Miss our last newsletter? We covered: The White House accused Moonshot AI of copying Claude, OpenAI says its AI hacked Hugging Face, and OpenAI launched Presence. And more. Check it here!
Ep 825: New: OpenAI Presence. Has The AI Customer Service Takeover Finally Arrived?
OpenAI launched Presence, and it seems like no one really noticed.
They should have.
Presence gives companies a way to build and manage real-time AI voice and chat agents across customer service, sales, HR, and IT, which could make this one of OpenAI’s most important enterprise launches yet.
The technology finally looks fast, natural, and capable enough to disrupt customer service at scale.
That could mean faster answers and fewer terrible phone trees.
Or an endlessly patient corporate gatekeeper.
We’re breaking down what OpenAI actually launched, what the early proof leaves out, and whether the AI customer service takeover has finally arrived.
Also on the pod today:
• OpenAI Presence launch fallout 🤖
• Real-time AI voice agents 🗣️
• Silent response to major release 🤫
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – Teable Turns your data into AI workflows and custom apps that 100% fit your business, Podcastor Creates Video Podcast with your AI Hosts, PromptQL It's a multiplayer AI agent with shared threads and a shared brain.
Qwen-Image-3.0 Launch — Qwen-Image-3.0 is being pitched as a big step up in image generation, with stronger handling of dense layouts, tiny text, and realistic details.
AI Smart Glasses — Samsung’s new AI smart glasses mix fashion partnerships with everyday functionality, hinting at a bigger push beyond just phone-linked tech.
NVIDIA Super Computer — NVIDIA just brought a DGX GB300 online at the Naval Postgraduate School, giving students and faculty serious AI compute for everything from weather prediction to cybersecurity.
Anthropic Economic Connector — Anthropic just launched a Claude connector for its Economic Index, so anyone can ask what AI is actually being used for across jobs, regions, and tasks.
Cursor Model Router — Cursor just launched a model router that sends each request to the cheapest model that can still do the job, claiming frontier-quality results at 60% lower cost.
AI Images — Microsoft released its newest AI image model, MAI-Image-2.5-Pro
Claude Security Plugin — Claude just dropped a beta security plugin for Claude Code that scans changes for vulnerabilities before you commit.
Meta AI Content Settings — Meta says it’ll now flag teen Meta AI chats about suicide or self-harm and notify parents, with a stricter content setting also rolling out.
Google AI Investments — Alphabet is betting bigger on AI, even if it trims margins.
Big Tech AI Pledge — Big utilities and data center companies are signing Trump’s pledge to keep AI costs off regular electricity bills.
1. AMD lands huge Anthropic AI chip deal 💻
Reuters reports AMD will supply Anthropic with tens of billions of dollars’ worth of AI servers and invest up to $5 billion in the Claude maker, a fresh sign that the chip wars are getting even more expensive.
The deal is aimed at up to two gigawatts of AMD’s upcoming Instinct MI450 chips, with deployment set to start in the first half of 2027 and AMD’s investment tied to those rollout milestones.
2. OpenAI backs DOE science push 🧑🔬
OpenAI says it is deepening its work with the U.S. Department of Energy’s Genesis Mission, a timely move that puts frontier AI directly into the country’s scientific pipeline.
The company plans $4 million in Codex access for about 2,000 researchers, $3 million for two major science campaigns, and up to $10 million in API usage for $2.5 million spent, alongside special access for select lab teams in biology and cybersecurity.
3. Amazon trims AGI jobs amid AI push 🫸
Amazon confirmed Wednesday that it cut an unspecified number of roles in its artificial general intelligence organization, a move first reported by Reuters, even as it pours money into helping customers deploy AI more effectively.
The company said it is narrowing its focus to the work that matters most for customers, signaling that the AI race is shifting from flashy model-building to making those systems useful in the real world.
4. ChatGPT desktop gets voice control for work 🗣️
OpenAI rolled out a major ChatGPT desktop update today, adding GPT-Live-powered Voice mode to the Mac and Windows app so users can talk through tasks and steer work across Chat, Work, and Codex in real time.
On Mac, the new Appshots feature adds extra context by letting ChatGPT see the active window, which makes the voice experience more aware of what is happening on screen. The update also expands support for local files, plugins, and OpenAI’s Computer Use feature, while voice access is now available on Plus, Pro, Business, Edu, and Enterprise plans.
5. Claude Voice Gets Smarter with Upgrade 🎙️
Anthropic is rolling out a major beta update to Claude’s voice mode across desktop, mobile, and web, and the timing matters because it directly tackles one of the feature’s biggest weak spots. Voice chats can now tap Sonnet or Opus instead of being stuck on Haiku, which means tougher questions should get stronger answers, and paid users will default to the last model they used in text chat.
Anthropic is also adding support for connected apps like Gmail and Slack, so Claude can pull in more relevant context when users allow it. The catch is that it still works in a turn-based way, not true live back-and-forth speech, and free users remain limited to Haiku with a single connection.
6. ChatGPT Health rolls out to U.S. adults as OpenAI widens medical AI access 🏥
OpenAI expanded ChatGPT Health to all logged-in U.S. users 18 and older on free, Go, Plus, and Pro plans, marking a fast-moving push to put health-focused AI tools into mainstream use.
The feature, which began as a dedicated hub earlier this year, now lets users pull in connected health data from services like Apple Health, MyFitnessPal, Epic, Oracle Health, One Medical, and Function Health across regular chats.
Many companies kinda gave up on AI customer service agents after the 2024 hype cycle flopped. OpenAI just quietly proved that call dead wrong.
Why now?
It's called OpenAI Presence, the company’s newest platform for voice and chat agents. It landed 24 hours ago with almost no buzz.
"Another AI customer service bot? Cool, more hold music."
Wayyy different this time y’all.
The ‘AI voice’ agents of 2024? Based on older and slower models that just transcribed your voice to text and vice versa. Today’s models? Bi-directional models that can think, reason, call tools and respond and interrupt mid-sentence.
Don't worry, you don't need to code anything to get why this matters.
On today's Everyday AI, we broke down what Presence is, how OpenAI keeps agents from going rogue, and why the voice race got way more crowded.
Back tomorrow with more. Let's go, y'all.
1. OpenAI Didn't Ship a New Voice Model. It Shipped Trust 🎯
Most companies chase whichever model gets loudest headlines. Presence isn't that.
Nope.
It's the layer of guardrails, approvals, and escalation rules that sits above OpenAI's voice tech and makes it safe enough for a real customer.
Every agent starts with one narrow job, like a billing issue or a password reset, and only gets the system access that job needs. Most companies rushing into AI agents do the opposite: one bot that tries to handle everything, which is backwards.
Presence forces narrow scope and human escalation from day one instead of bolting it on after something breaks. That's the real unlock: trust gets engineered in, not hoped for.
Try This
Pick one narrow, painful workflow, like billing disputes or password resets, before you touch a company-wide rollout. Map exactly what data that single job needs and nothing more.
Set the escalation bar low at first so your team can watch where the model breaks. That's how you build trust before you scale it, prolly the biggest lesson buried in this launch.
2. How OpenAI Keeps Its Agents From Going Rogue ⚡
Here's the part every leader glosses over: a model that nails the demo can still go sideways once a frustrated customer gets on the line.
So Presence tests agents against common and high-risk scenarios before a single customer talks to one, and after launch, Codex keeps watching, reviewing conversations and proposing fixes your team has to test and approve.
Nothing here ships on AUTOPILOT. The real innovation isn't the voice, it's the governance wrapped around it, and that part barely made a headline despite being what actually matters.
Skip it, and even a great model turns into a liability the moment it's wrong in front of a customer.
Try This
Before deploying any AI agent, build your own benchmark for what counts as a win, a loss, and a guardrail violation. Public leaderboards are a fine start, but they won't save you alone.
Run your riskiest scenarios through simulation first, the angry customer, the ambiguous yes that's actually a no, and catch the failure in a sandbox, not on a live call with your best customer.
3. The Real-Time Voice Race Just Got Way More Crowded 🚀
Here's why this works now and didn't two years ago: early voice agents sounded robotic and paused mid-sentence, which was death on a live call.
The fix is a second model working quietly in the background: searching, pulling data, and calling tools while the first keeps talking.
No more dead air. No more guessing.
OpenAI's been quietly DOMINATING this space. Google's pushing Gemini Live hard too, though its bigger Pro model has been delayed for months, and Anthropic's rumored to be building its own voice mode.
More competition here means faster progress, and every major lab is racing toward the same thing: an assistant you can just talk to that gets things done.
Try This
Go test GPT-Live or a competing voice agent this week, not a scripted demo, an actual conversation. Notice where it pauses, stumbles, or starts to feel human.
That gap between an impressive demo and handling real customers is where most companies get burned. Know that gap before you sign a contract, and you'll ask sharper questions in every vendor pitch next quarter.
Reply