OpenAI Workspace Agents: Build-Your-Own AI for SMBs

OpenAI just turned ChatGPT into a team agent platform

On April 22, 2026, OpenAI launched Workspace Agents — a research preview that lets a team build a single AI agent and run it across Slack, Gmail, Google Drive, Calendar, Docs, Sheets, and Salesforce. It is the successor to custom GPTs, but with one big difference: these agents take action, and the whole team can use the same one.

For a small business that has been watching enterprise AI from the sidelines, the timing matters. Workspace Agents is free until May 6, 2026, then it switches to credit-based pricing. That gives roughly three days to test-drive what OpenAI is calling its answer to Anthropic’s Claude Managed Agents and Microsoft Copilot Studio.

The news in five facts

What launched: Workspace Agents, in research preview inside ChatGPT Business, Enterprise, Edu, and Teachers plans. Powered by Codex under the hood.
What they do: Take action — prepare reports, write code, respond to messages, and run scheduled jobs in the cloud. They keep working when you’re logged off.
Where they run: Direct integrations with Slack, Google Workspace (Gmail, Drive, Calendar, Docs, Sheets), and Salesforce, per VentureBeat’s launch coverage.
What it costs to access: ChatGPT Business is $30 per user per month. For a five-person team, that’s roughly $150/month before any agent usage credits.
The pricing cliff: Free until May 6, 2026. After that, OpenAI charges credit-based rates for token consumption, with subscription plans including a token budget and overage on top.

Why team-shareable agents matter for small businesses

The original Custom GPTs were a personal toy. You’d build one, paste in your brand voice, and use it yourself. If a coworker wanted the same thing, they cloned it, drifted it, and now you had two slightly-different brand voice bots floating around.

Workspace Agents is built differently. One agent, shared across the team, used inside the tools people already work in. That changes what you can actually trust an agent to do.

For a small contractor, that might mean an agent that watches the shared Gmail inbox, drafts replies to lead inquiries in your tone, files them in the right Drive folder, and posts a summary to a Slack channel where the office manager triages them. One person builds it, everyone uses it, and it improves as a single source of truth instead of forking into five inconsistent copies.

That is the same pattern we covered when Anthropic launched Managed Agents — the platforms have stopped pitching novelty and started shipping the team-coordination plumbing that makes agents survive contact with a real business.

How OpenAI’s pitch compares to Claude Managed Agents

These two products solve different problems for different buyers, even though the surface pitch is similar.

Claude Managed Agents is a developer platform. You write code against an API, define tools, and Anthropic hosts the runtime. Pricing is $0.08 per session-hour plus tokens. It is for anyone building a custom agent into a product or workflow that does not exist off-the-shelf.

OpenAI Workspace Agents is a no-code platform inside ChatGPT. You build agents in a UI, share them with the team, and OpenAI runs them across the apps you already use. There is no API surface to integrate against — the agent is the product, not a building block.

If you are a small business owner without an engineering team, that distinction matters. Workspace Agents is the first credible “I can build an agent on my lunch break” option that also runs in the cloud and survives across teammates. That is a real shift from Custom GPTs, which were essentially long prompts.

The trade-off is depth. Workspace Agents will be lighter on customization than a Managed Agents build. If your workflow is unusual, you’ll hit the ceiling fast.

Should a small business adopt now or wait?

Here is the honest read: try it during the free window, but do not bet your operations on it yet.

The bottom line: Workspace Agents is genuinely useful and priced reasonably for SMBs that already pay for ChatGPT Business. But “research preview” means the rules will change, and credit-based pricing has burned a lot of small business AI budgets already.

What we like

Team-shareable by default: This solves the “every person has their own slightly-broken GPT” problem that custom GPTs created.
Acts inside Slack and Gmail: Agents that send a message and disappear are a curiosity. Agents that work where you already work are real productivity.
The free window is generous: Three weeks to find out if a workflow even makes sense before any pricing kicks in.

What concerns us

The price floor is fuzzy. Credit-based pricing means you do not know your monthly bill until usage stabilizes. We’ve seen this pattern with 40 percent of agent projects getting cancelled when surprise costs land.
Research preview means breakage. Features will change. Integrations will get rate-limited. If you build a critical workflow on this in May and it shifts in June, that is on you.
It is locked to ChatGPT Business+. If you are on a free or Plus plan, you cannot try it. That gates a lot of the small businesses who’d benefit most from no-code agents.

What to do this week

If you already pay for ChatGPT Business, log in before May 6 and pick one workflow to prototype. Good candidates:

A lead-response agent that watches a shared inbox and drafts replies in your tone
A meeting prep agent that pulls notes from prior calls and Drive files into a brief before each appointment
A review-summary agent that scans recent customer reviews and posts a Slack digest every morning

If you are on ChatGPT Plus or Free, this isn’t your moment yet. The cheapest path in is a single Business seat at $30/month, which is worth it if you have a real workflow to test but not worth it just to “see what’s there.”

If your workflow is industry-specific — HVAC dispatch, restaurant intake, vacation rental guest messaging — a vertical agent purpose-built for that vertical will get you to production faster than a generic Workspace Agent. We package thirteen of those at Appalach.AI’s AI Employees, each tuned for one industry. Workspace Agents is best when your workflow is generic enough that a horizontal tool can handle it.

The shift here is real. The pricing cliff is real too. Use the free window to learn what’s possible, not to ship something fragile.

Need help figuring out which agent platform fits your business? Get in touch — we help small businesses pick AI tools that match their workflows and budget instead of the other way around.