• Everyday AI
  • Posts
  • Ep 771: ChatGPT Workspace Agents: How to Use OpenAI’s Most Overlooked New Feature

Ep 771: ChatGPT Workspace Agents: How to Use OpenAI’s Most Overlooked New Feature

Claude's new self-improving agents that dream, SpaceX and Anthropic team up for compute, OpenAI's ads get big boost and more

Author

Everyday AI
May 06, 2026

 

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Workspace agents might’ve been OpenAI’s most overlooked release this year, giving teams AI agents that can connect to apps, run workflows, and work directly inside Slack. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Google is testing major upgrades for its Antigravity IDE, Google Home just got smarter with Gemini 3.1, and Perplexity launched a real-time finance search API, and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Apple settled its iPhone AI lawsuit, OpenAI released new infrastructure tech for massive AI clusters, and Google DeepMind staff are pushing back on military AI work, and more. Read on for Byte Sized News.

💪 Leverage AI: How can your team use Workspace Agents? We walk you through it. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Big Tech is giving the U.S. early access to AI models, ChatGPT just upgraded to GPT-5.5 Instant, and Gemini Flash 3.2 is on the way, and more. Check it here!

Ep 771: ChatGPT Workspace Agents: How to Use OpenAI’s Most Overlooked New Feature


Did you ACTUALLY MISS OpenAI’s big Workspace Agents announcement? 🤔

You might have.

Sandwiched between huge announcements like GPT-5.5 and Images 2, OpenAI’s crazy powerful Workspace Agents didn’t get much media or human attention since release.

That’s one reason why we HAVE to go live and show you the basics of how they work, why you should be using them, and the best practices to start converting your manual workflows into Workspace Agents today.

Also on the pod today:

• Drag-and-drop agent builder demo 🧩 
• Audit every agent workflow 👀 
• Automate tasks across apps ⚡

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 –  Kanwas gives teams & agents one place to create, edit, share and compound product context, Inworld Realtime TTS-2 is The #1 ranked, most natural voice AI, Contrario is the AI Recruiting Platform powered by expert recruiters.

Google Antigravity — Google’s Antigravity IDE is testing screen sharing for agents and support for custom plugins, hinting at big upgrades to flexibility and visibility.

Meta Bone Analysis — Meta is now using AI to spot underage users by analyzing photos, videos, and profile clues.

AWS Agents — AWS is letting AI agents control its cloud PCs, but it could cost you big compared to using APIs.

Google Home Gemini Upgrades — Google Home just got smarter with Gemini 3.1, now handling multiple commands at once and better managing your smart devices.

Google Seperate I/O Event — Google is hosting a special Android Show right before I/O, hinting at a flood of new Gemini AI features for your phone.

Maket AI — Maket's AI lets anyone design floor plans in minutes, no architecture skills needed.

Perplexity Finance Search — Perplexity just launched Finance Search, letting you pull real-time, cited financial data straight into your tools with one API call.

AI Infra — OpenAI just rolled out MRC, a new networking protocol that lets over 100,000 GPUs train together faster and with fewer failures. Here’s how it works.

Microsoft Index Report — Most organizations using Microsoft 365 are falling behind their own employees when it comes to AI, while top “Frontier Firms” are setting the pace

AI Copyright — Meta is being sued by major publishers and author Scott Turow for allegedly using millions of books to train its AI.

1. Apple Settles iPhone AI Lawsuit, Cash Payouts Ahead 📱

Apple just agreed to a $250 million class action settlement after being accused of misleading U.S. customers about its hotly-hyped Apple Intelligence features on the iPhone 16 and iPhone 15 Pro.

If you bought one of those devices between June 2024 and March 2025, you could get up to $95 per phone, though most will see closer to $25, depending on how many people file claims. The lawsuit claimed Apple’s marketing set unrealistic expectations for its AI rollout, which arrived late and missing many promised features.

2. Anthropic Taps SpaceX Compute Ahead of IPO 🚀

Anthropic announced Wednesday that it will use all the compute capacity at SpaceX’s Colossus 1 data center, instantly adding access to more than 220,000 NVIDIA GPUs and expanding service for Claude’s paid users.

The timing is notable, with SpaceX just weeks from a highly anticipated IPO, giving the company a high-profile AI customer as it pitches itself as more than a rockets-and-satellites business. The deal strengthens Anthropic’s fast-growing web of compute partnerships as it races to scale globally and meet demand in regulated markets.

3. OpenAI Opens Up Ads Manager, Drops Big Spend Requirement 💸

OpenAI is shaking up the ad world by making its self-serve ads manager available to all U.S. advertisers and slashing its $50,000 minimum spend, finally welcoming smaller businesses into the ChatGPT advertising fold.

The company is moving fast, with cost-per-action bidding and third-party measurement tools in the pipeline, though details are still under wraps. OpenAI remains cautious about ad quality and brand safety, keeping tight control over ad delivery while slowly expanding eligible categories.

4. ServiceNow and NVIDIA Unveil New Era of Autonomous Enterprise AI Agents 🤝

At ServiceNow Knowledge 2026, tech titans ServiceNow and NVIDIA announced a major leap for enterprise AI, unveiling Project Arc, an autonomous desktop agent that brings advanced, secure automation to the workplace.

This partnership combines ServiceNow's workflow intelligence with NVIDIA's accelerated computing and open models, aiming to make AI agents smarter, safer, and easier to manage at scale. With innovations like NVIDIA OpenShell for secure agent execution and a focus on real-world performance, the companies are providing the infrastructure and guardrails enterprises need to trust AI with complex tasks.

5. Anthropic Adds Self-Improving “Dreaming” to Claude 😶‍🌫️

The research feature lets Claude review its own work between sessions, spot patterns, and update stored context like user preferences without human prompting. In simple terms, Anthropic is testing whether AI agents can get better at their jobs on their own over time, not just during active use.

6. Google DeepMind Staff Push to Unionize Over Military AI Concerns 🪖

In a bold move, nearly all unionized staff at Google DeepMind’s London headquarters have voted to unionize, demanding their AI work not be used for military purposes or to aid in international law violations.

Workers are calling on Google to formally recognize the union within ten days or face legal action, highlighting fresh tensions over tech’s role in global conflicts. The letter to management urges a ban on developing AI for weapons or surveillance and seeks the right to refuse work that clashes with personal ethics.

Most enterprise leaders still think ChatGPT is a chat box. OpenAI just quietly turned it into a deployable team employee that works overnight, connects to all your org's apps, and never gets distracted.

Your competitors are already building them.

OpenAI launched Workspace Agents two weeks ago. They're Codex-powered, cloud-running, and shared across your entire org. The free research preview just ended, and credit-based pricing starts today.

"That's an enterprise thing, not for my team." You know that thought.

Biiiig mistake.

We've got the full breakdown for every leader, technical or not.

What do Workspace Agents actually unlock?

They replace every workflow where you're manually carrying context between apps: briefings, lead research, weekly digests, Slack-triggered responses. All of it running in the cloud, every step auditable.

On today's episode of Everyday AI, we went hands-on building one live, showed how Workspace Agents compare to Codex and GPTs, and mapped out the three moves your team needs to make right now.

You've gotta window. Let's get after it.

1. Replace Your Team's Human Duct Tape 🔥

Your team has workflows where a human is the glue. An email lands, you open three apps, look something up, update a spreadsheet, fire a Slack message.

That whole chain, over and over. Workspace Agents can replace all of it, running in the cloud with zero local computer dependency.

What does that actually look like in practice?

In our example on today’s show, we had an agent connects to beehiiv via a custom MCP, pulls the five most-clicked stories from each of the past five newsletter issues, filters for actionable AI features, and emails a synthesized digest every Thursday at five PM. A process that used to take over an hour of clicking, reading, and getting pulled down rabbit holes.

Now it just runs. WITHOUT YOU. Think of that across a handful or a dozen of your most common workflows.

Try This

Pick one workflow your team repeats every week that crosses at least three apps. Write it out step by step like you're onboarding a new hire.

Then open the Agents tab in ChatGPT and describe it in plain English. Let the builder connect your apps, suggest a schedule, and draft the workflow in minutes.

Run it once. If the output looks close, you've prolly just bought back two-plus hours a week. That's your ROI case for the rest of your team.

2. Your Team's First Auditable AI Worker ⚡

GPTs never let you see what they actually did and couldn’t easily “do” the work or create artifacts. Workspace Agents log everything.

Every run logged. Every step traced. Every output reviewable.

Here's where it gets REAL for enterprise teams.

Role-based access controls let admins decide who builds agents, who uses them, and which tools each agent can touch. Build once, share across the org.

What one person builds this week becomes team infrastructure by next Monday. Persistent memory means the agent learns your processes over time, so next week's run knows what last week's run found.

What does this mean for your AI strategy?

You can move from individual AI productivity to SHARED organizational execution. That's the actual shift that's been missing.

Try This

After your first agent runs, don't just check the output. Click into the Activity log and trace every step it took to get there.

If the result missed the mark, you'll see exactly where the chain broke. Fix that one step. Rerun.

The audit trail isn't just governance. It's the tool that makes your agent sharper every week, and the fastest way to get skeptical leadership to trust what it's doing.

3. GPTs vs. Codex vs. Workspace Agents: Pick One 🚀

Today Workspace Agents pricing flips to credit-based. That free launch window just closed.

So which tool do you actually use?

GPTs still work fine for one-off in-chat tasks. But if you're on a team plan, stop building new ones.

OpenAI is building a conversion tool to move your existing GPTs into Workspace Agents. Don't invest more in a format that's likely to be phased out.

Codex is more powerful, but your machine has to stay on for automations to run. Workspace Agents run 24/7 in the cloud with ZERO machine dependency.

If your team runs on ChatGPT Business or Enterprise, Workspace Agents are your default automation layer right now. Not on a team plan yet? Two ChatGPT Business seats run $20 per person per month. Better rate limits than the $100 solo plan, for less money.

Try This

Write down the three most repetitive multi-step workflows your team does every week. For each one, ask: does it cross more than two apps? Does it need to run on a schedule? Does more than one person on your team do some version of it?

If yes to all three, that's a Workspace Agent job. Start with the one that wastes the most hours. You don't need a perfect workflow to start. The first draft gets you 80% of the way there.

Reply

or to participate.