OpenAI launched Workspace Agents inside ChatGPT on April 22, 2026. They're cloud-based team agents that replace Custom GPTs, and they keep working in the background whether you're signed in or not. For small businesses on ChatGPT Business, this is the first time an agent that actually does multi-step work lives inside ChatGPT itself. The catch: it's not on the Plus plan, and the pricing model flips on May 6.

I was reading through the OpenAI announcement this morning and kept thinking about what this changes for the small businesses we work with. The short version is that the agent conversation just moved from "which third-party tool do I buy?" to "is my ChatGPT seat already paying for it?" Here's the honest read on what it does, who it's for, and what to do if your business is already stacking AI tools.

What OpenAI Actually Launched This Week

Workspace Agents are shared AI agents that live inside ChatGPT (and now Slack), built for teams instead of individual users. The technical backbone is Codex, OpenAI's code-aware model, which lets the agents reason across data, trigger actions, and write or run code as part of a single job.

The announcement covered three shifts worth noting:

They run in the cloud. An agent can keep working after you close your laptop. Closing ChatGPT doesn't stop the job.
They can run on a schedule. Think "auto-generate the weekly revenue report every Friday at 5 p.m." instead of "ask me something."
They replace Custom GPTs. OpenAI is explicitly positioning Workspace Agents as the next step, and says conversion tools for existing Custom GPTs are on the way.

Integrations announced at launch include Slack (agents deploy directly into a channel), plus connections to CRMs, IT ticketing systems, and communication tools. OpenAI hasn't published a full supported-app list yet, so expect the roster to grow week by week.

This lands during OpenAI's broader launch week, alongside updates to team controls and enterprise permissions. It's the most aggressive agent move OpenAI has made to date, and the first one shipped directly inside the product most small businesses already pay for.

Who Actually Gets It (and the Pricing Surprise)

Workspace Agents are in research preview on the ChatGPT Business, Enterprise, Edu, and Teachers plans. They are not available on ChatGPT Plus, the Free tier, or the standalone Team plan that individual users often run.

For most small businesses, that means:

If you pay $25 per user per month for ChatGPT Business, you're in.
If you pay $20 per month for Plus, you're not.
If you have a handful of seats and never upgraded to a business plan, this is a real reason to consider it.

Pricing for the agents themselves is unusual. They're free until May 6, 2026, and then OpenAI switches to a credit-based model. Credits get consumed per agent run, and the exact rates haven't been published. Expect long-running or code-heavy jobs to eat more credits than a simple chat turn.

We'd be careful about architecting a critical workflow on free-tier pricing. The window between now and May 6 is for testing, not committing.

Three Real Things a 5-Person Business Could Ship This Week

Most of the coverage of the launch is "what is this product." That's fine, but not useful. The real question is: if I have five employees and a ChatGPT Business subscription, what can I actually set up this week?

Three examples that are realistic inside the research-preview window:

1. Weekly Revenue Snapshot (Runs Automatically Every Friday)

Give the agent access to your revenue source (Stripe, Shopify, HubSpot, a spreadsheet) and a template for what the weekly snapshot should look like. The agent pulls Monday through Friday numbers, writes the summary in your voice, and drops it in a Slack channel or an email every Friday at 5 p.m.

Before Workspace Agents, this took a Zapier workflow, a database query, and someone to write the prose. Now it's a single agent configuration. We'd start here because the output is easy to check and the failure mode is obvious. A wrong number stands out.

2. Slack-Based Lead Triage

Drop the agent into your sales Slack channel. When a new lead hits your CRM, the agent picks it up, checks it against your ideal customer profile, pulls any public company info it can find, and posts a qualification note with a recommended next step.

The useful part is context. The agent persists state across the week. It sees that Lead A went cold after two touches and that Lead B just hit your pricing page. A Zap doesn't know that.

3. Customer Support First-Response Agent

This one's worth staging carefully. The agent watches your support inbox (connected through a comms integration) and drafts a first response for every ticket. A human reviews and sends. Over time, the agent learns which tickets always get handed off to a person and which it can resolve.

The reason to stage this as a first-draft agent, not a fully autonomous one, is trust. Small businesses live and die on tone. A wrong reply from an AI is worse than a slower reply from a human.

All three setups are supposed to be possible inside Workspace Agents today, but the integration surface is still expanding. We'd test each one on a throwaway workspace before moving anything real over.

What This Replaces (and Why That Matters)

Workspace Agents officially replace Custom GPTs. That's the line most people glossed over when the announcement came out, but it's the biggest shift for small businesses.

If you built a Custom GPT for any part of your business, maybe an onboarding GPT, a brand-voice writer, a sales FAQ responder, that's the old way. OpenAI said conversion tooling is coming, but existing Custom GPTs are now on a migration track. If your team relies on one for daily work, start planning the move.

The reason this matters is that Custom GPTs were individual. You made one, you used it, maybe you shared a link. Workspace Agents are team objects with permissions, schedules, and shared state. You can't just share a chat link anymore. That's a different way of thinking about AI inside a business.

It also means the person on your team who built the best Custom GPT just became more valuable, not less. They already know how to define an agent's job. Now they get to build ones that scale across the team.

Where This Still Breaks for Small Businesses

A few things to be honest about:

Research preview means unstable. Expect the feature to change before it stabilizes. Don't build a process around behavior that might shift in two weeks.
Credit pricing is unknown. We can't tell you yet whether a daily triage agent will cost $5 or $50 per month after May 6.
Integrations are still rolling out. If the app your business actually depends on (QuickBooks, ClickUp, Airtable) isn't on the supported list yet, the workflow you want to build might not be possible this week.
Plus-only shops get nothing. If you're a one to three person team on the Plus plan, this launch doesn't include you. That's a hard pill.

None of these are reasons to ignore the launch. They're reasons to test small, document what works, and not bet the business on a research preview.

How Workspace Agents Stack Up Against the Field

The agent race is crowded right now. A quick read on where each major player sits:

| Product | Strength | Best for | |---|---|---| | OpenAI Workspace Agents | Inside ChatGPT, team permissions, scheduled runs | Small biz already on ChatGPT Business | | Anthropic Claude for Work | Deep reasoning, long-context agents | Teams with complex research workflows | | Microsoft Copilot | Native Office and Outlook integration | Microsoft-first shops | | Google Workspace AI | Docs, Sheets, Gmail native | Google-first shops | | Zapier + ChatGPT or Claude | Widest app-integration surface | Non-technical automation builders |

The practical read for small business owners is that the "right" platform is mostly decided by the rest of your stack. If your business runs on Google, go Google. If it runs in ChatGPT, this is the shortest path to a real agent.

We don't think any one of these wins across the board in 2026. Most of the small businesses we work with use two, not one.

What To Do This Week If You're on ChatGPT Business

If you're already paying for ChatGPT Business, here's the practical to-do list:

Turn on Workspace Agents in admin. The feature is off by default for most accounts.
Inventory your Custom GPTs. List every Custom GPT your team uses. That's your migration backlog.
Pick one low-risk workflow to test. Weekly report. Lead triage. Something where a bad output is obvious.
Read the credit pricing announcement when it lands. Plan your real budget from there, not from the free preview.
Revisit in four weeks. Integrations and capabilities will shift fast. The process you design today should expect change.

That's the whole plan for April. Most small businesses don't need to do more than that before the pricing model finalizes.

FAQ

What are ChatGPT Workspace Agents?

Workspace Agents are cloud-based AI agents that live inside ChatGPT for teams. OpenAI launched them on April 22, 2026 to replace Custom GPTs. They run in the background on a schedule or trigger, integrate with Slack and third-party apps, and can be shared across a team with permissions.

Are Workspace Agents available on ChatGPT Plus?

No. Workspace Agents are on ChatGPT Business, Enterprise, Edu, and Teachers plans only. Plus and Free users do not have access. For a small business that wants them, the relevant upgrade is usually ChatGPT Business at $25 per user per month.

How much do Workspace Agents cost?

They're free to use during the research preview, which runs through May 6, 2026. After that, OpenAI moves to a credit-based pricing model. Exact rates haven't been published. Expect long-running or code-heavy agent tasks to consume more credits than simple ones.

What's the difference between Workspace Agents and Custom GPTs?

Custom GPTs were individual AI assistants you built and used solo. Workspace Agents are shared team objects with permissions, scheduled execution, and cloud-based runs that continue when you're offline. Workspace Agents are meant to replace Custom GPTs, and OpenAI has said conversion tools are coming.

Can Workspace Agents work with Slack?

Yes. Slack integration was announced at launch. An agent can be deployed directly inside a Slack channel, where it responds to messages and handles triggered tasks. Other integrations include CRM systems, IT ticketing tools, and comms platforms, with more being rolled out.

Should small businesses switch to Workspace Agents right now?

Test, don't commit. The feature is in research preview, and post-May-6 pricing isn't public. A reasonable April plan is to try Workspace Agents on one low-risk workflow, inventory any Custom GPTs your team uses, and re-evaluate when the pricing model finalizes.

What happens to my existing Custom GPTs?

Nothing immediately. OpenAI said conversion tooling to migrate Custom GPTs to Workspace Agents is coming, but it hasn't shipped yet. Existing Custom GPTs keep working. Start listing which ones your team actually uses so migration is quick when the tool arrives.

OpenAI just put an agent inside the product most small businesses already pay for. That's the real story.

So: is this a reason to change how your team works this week, or a reason to watch and wait? We lean "test, don't commit" for April. But if you already built Custom GPTs that matter to your business, the migration clock just started.

ChatGPT Just Launched Workspace Agents. Here's What That Means for Small Businesses.