October 5, 2026
At DevDay on September 29, 2026, OpenAI introduced Dots: always-on agents built on GPT-6 Astra, each one given its own cloud computer and its own browser, according to TechCrunch. A Dot isn't a chat session you babysit. It runs in the background, reaches into more than 4,000 apps and services, and shows up wherever you already work -- ChatGPT, text messages, email, Slack. 9to5Google's reporting confirms the same cross-channel footprint and the 4,000+ app figure. Dots can handle recurring work, follow up on open tasks without being asked, and delegate pieces of a job to sub-agents. You can set rules requiring confirmation before certain actions, and the first Dot is free with Pro and Business Premium plans in eligible markets.
The coverage is framing Dots as a leap in agent autonomy, and it is. But look at the number that matters most: 4,000+ apps. OpenAI didn't build Dots to replace your CRM, your helpdesk, your project tracker, or your invoicing tool. It built Dots to sit on top of all of them at once. The entire value proposition depends on you continuing to pay for Salesforce, or HubSpot, or Asana, or whatever stitched-together pile of SaaS your business runs on today -- because a Dot's job is to log into those tools, click around in them, and coordinate between them on your behalf. Take away the 4,000 subscriptions and Dots has nothing to orchestrate.
That's not a side effect. That's the business model. OpenAI makes more money the more fragmented your stack stays, because fragmentation is exactly what creates demand for a paid coordinator sitting above it. Dots is a land grab for the orchestration layer -- the connective tissue between tools -- not an attempt to shrink the number of tools. It's the same instinct behind workflow automation products generally: wire together what already exists rather than question why it needs eleven separate logins in the first place.
Think about what a Dot actually costs a mid-size business once it scales past the free seat. You're still paying full freight for every app in that 4,000-app graph -- your CRM license, your helpdesk seats, your project management per-user fees, your accounting software. Then you add OpenAI's agent fee on top, priced for the privilege of having something remember to follow up, re-key data between systems, and nag your team on your behalf. You haven't reduced your vendor count. You've added one more vendor whose entire job is to paper over the gaps between the other dozens you still owe money to every month.
This is the exact math we've been pointing at in pieces like GitHub's Copilot billing change and Rippling's modular pricing: per-seat and per-agent fees compound against a business precisely because the underlying tool count never goes down. An agent that babysits your sprawl is still a recurring cost layered on recurring costs. It makes the sprawl more tolerable to live with, which is a different thing entirely from making it smaller.
Dots encroaches directly on workflow-automation territory -- the same territory Zapier, Make, and no-code app builders have fought over for years. By making itself the layer that follows up on tasks, routes work between tools, and spins up sub-agents to handle pieces of a job, OpenAI is positioning ChatGPT as the default coordination surface for every business process, regardless of which underlying apps are involved. That's a smart land grab. It's also a bet that businesses will keep tolerating 4,000-app sprawl forever as long as something smooths over the seams.
We think that bet is wrong for a meaningful share of the businesses evaluating it. The connective-tissue layer Dots wants to own -- the rules about who gets notified, what happens after a deal closes, how a support ticket becomes a project task -- is exactly the layer a business should own outright, not rent indefinitely from whichever foundation model is winning the news cycle this quarter. We've written about what that handoff actually looks like when you own it: see automating the handoff from closed-won to client portal or building an expense approval workflow without per-seat fees.
ViibeStack's bet runs the opposite direction from OpenAI's. Instead of paying an agent to log into your CRM, your helpdesk, and your project tracker forever, build a CRM, helpdesk, and project management layer you actually own, on a single platform, with one data model underneath it. There's no coordination tax because there's nothing left to coordinate across -- the data already lives in one place. We've mapped this out tool by tool for the systems businesses most commonly outgrow, from replacing Salesforce to replacing Asana to replacing Zendesk, and the buy vs. build vs. ViibeStack comparison lays out the math plainly.
None of this means agents are useless -- plenty of recurring, cross-app busywork is genuinely better handled by something that never sleeps. But an agent that makes 4,000-app sprawl more livable is solving the wrong problem. The businesses that come out ahead over the next five years won't be the ones with the most elegant AI coordinator sitting on top of their software graveyard. They'll be the ones who used this moment to actually shrink the graveyard. That's a software-ownership decision, not a subscription decision, and it's worth making before you add yet another recurring line item to babysit the ones you already have.
Sources