OpenAI used its September 29, 2026 DevDay keynote in San Francisco to introduce dots, a fleet of “always-on” AI agents built on the company’s GPT-6 Astra model. The launch was never just a product update. It was a declaration that OpenAI intends to fight for enterprise software budgets, not just consumer chat traffic, and it landed the same week Meta rolled out its own enterprise push and Google Cloud kept building out its Gemini Enterprise Agent Platform. The result is a three-way scramble for the same corner office spending line, and the numbers behind it are starting to add up.
Each dot runs on its own cloud computer and works toward a user-defined goal around the clock, according to OpenAI’s DevDay presentation. The company says dots can connect to more than 4,000 apps, including a rollout that notably excludes the UK and EU for now. That geographic gap alone tells you something about how fast OpenAI is moving and how much regulatory friction it’s trying to outrun.
What OpenAI Actually Announced at DevDay
The core pitch behind dots is persistence. Instead of a chatbot that answers one prompt and forgets everything, a dot keeps working on a task across hours or days, checking in only when it needs a decision from a human. OpenAI’s own framing at DevDay was direct: dots “can do nearly anything,” a claim the company will now have to defend in front of enterprise IT departments used to auditing every automated process that touches their systems.
OpenAI described a scenario where a dot updates a sales proposal automatically when a customer changes requirements, then builds a working demo for a rep to review before a call. That’s a meaningfully different pitch than a research assistant or a writing tool. It’s asking businesses to hand a slice of an actual workflow to software that runs unsupervised. Integration partners named at launch include Slack and Microsoft Teams, which puts dots directly inside the tools where enterprise work already happens rather than asking employees to open a new app.
Access is being staged carefully. Pro and Business Premium subscribers were reported to get initial access first, while Enterprise customers need workspace administrator approval, a gate that looks a lot like the pilot-program structure most large companies already demand before letting any AI tool touch production data. OpenAI’s chief financial officer, Sarah Friar, addressed the pricing path in a CNBC interview, saying “for the business, you’re going to see it show up inside our pro SKUs and then inside of our business enterprise offerings for now” (Business Insider). That’s a company signaling it wants dots to eventually reach every tier, starting from the top of the price ladder down.
Meta’s Enterprise Platform Answers Back
Meta didn’t wait around. The company’s official announcement framed its move plainly: “Today we are starting the next major pillar of our business, Meta Enterprise Platform, to help businesses use AI to grow and transform in new ways as well,” according to Meta’s own launch post (Meta Newsroom). Coming from a company built almost entirely on advertising and consumer social products, that’s a notable pivot in public language, even if Meta has been building toward enterprise AI tooling for a while through its Llama model family and Muse assistant.
Meta CEO Mark Zuckerberg laid out the ambition on an earnings call, telling investors: “We see a large enterprise opportunity to sell to businesses, including APIs, business agents, potentially selling compute directly, and other services that we’re building for large customers,” a comment reported by TechCrunch. That statement predates the September enterprise platform launch by roughly two months, suggesting Meta had already decided to chase business revenue well before OpenAI’s DevDay forced the comparison into headlines.
Meta’s consumer-facing agent, Muse, has already drawn direct comparisons to OpenAI’s tools. Shattered.io covered how dots stacks up against Muse’s reported 2.8 million users and Muse’s download pace against ChatGPT in the weeks before DevDay. The enterprise platform is a separate, business-facing track sitting alongside that consumer product, and it’s the one that will decide whether Meta can turn AI attention into recurring software revenue rather than ad impressions.
Google Cloud’s Quiet Third Front
While OpenAI and Meta traded headlines, Google Cloud has been building out its own answer with less noise: the Gemini Enterprise Agent Platform. Google’s own product documentation describes it as “our open and comprehensive platform that empowers businesses to rapidly build, scale, govern and optimize enterprise-grade agents grounded in your enterprise data” (Google Cloud). Google’s consumer-side agent effort, reportedly named Gemini Spark, sits apart from this enterprise platform, mirroring the split both OpenAI and Meta have drawn between their consumer chat products and business-grade agent tooling.
Google’s pitch leans hard on governance, a word that shows up constantly in enterprise software sales decks for good reason. IT and security teams don’t just want an agent that works, they want one that logs its actions, respects data boundaries, and can be shut off cleanly if something goes wrong. That’s also the exact area where Google has been pushing its GKE agentic migration tooling to chip away at AWS’s cloud infrastructure lead, so the Gemini Enterprise Agent Platform fits a broader strategy of selling governance and infrastructure control as the differentiator, rather than trying to out-hype OpenAI or Meta on raw agent capability.
Money Is Already Moving: Instinct’s $1 Billion Signal
The clearest sign this market is real, not just marketing noise, comes from outside the big three. Instinct, described in reporting as an AI-agent startup, closed a funding round of $1 billion, a figure that puts a single agent-focused startup in the same funding tier as established infrastructure companies. Venture money doesn’t usually move that fast unless investors believe enterprise buyers are actually going to write checks, not just run free pilots.
That raise also changes the competitive math for OpenAI, Meta, and Google. It’s no longer a fight between three giant labs with unlimited compute budgets. A well-funded independent player focused purely on agents can move faster on a narrow use case, undercut on price, or get acquired by any of the three majors looking to buy their way into a feature gap. Reuters has separately reported that OpenAI plans to spend tens of billions of dollars on AI data centers, a scale of investment that dwarfs Instinct’s raise but also explains why OpenAI needs enterprise revenue like dots to eventually justify that spending.
How the Three Platforms Compare
Laid side by side, the three offerings share more structure than their marketing suggests. Each pairs a consumer-facing chat or assistant product with a separate, more tightly governed enterprise agent layer sold on top of existing cloud or workspace contracts.
| Company | Enterprise Agent Product | Underlying Model | Consumer Counterpart | Access Model |
|---|---|---|---|---|
| OpenAI | dots | GPT-6 Astra | ChatGPT | Pro and Business Premium first; Enterprise needs admin approval |
| Meta | Meta Enterprise Platform | Llama-based models | Muse | Business-tier rollout tied to Meta’s ad and platform business |
| Gemini Enterprise Agent Platform | Gemini family | Gemini Spark (reported) | Google Cloud contract customers, governance-first onboarding |
The most important column in that table isn’t the model name, it’s access model. OpenAI’s admin-gated Enterprise rollout for dots signals the company knows large customers need a slower, permissioned path even while it moves fast on the Pro and Business Premium tiers. Google’s governance-first framing plays to its existing cloud sales relationships. Meta is the wildcard, since it has the least established enterprise sales motion of the three despite having, by some measures, the largest existing developer and advertiser base to cross-sell into.
Integration and App Reach
Reach into existing software is where OpenAI is currently making its loudest claim. The company says dots work toward user goals 24/7 and can tap more than 4,000 apps, including direct integrations with Slack and Microsoft Teams. That number matters because agent products live or die on how many real workflows they can actually touch without custom engineering work from the customer’s side.
| Metric | Reported Figure | Source |
|---|---|---|
| Apps connected to dots | 4,000+ | OpenAI DevDay announcement |
| Instinct funding round | $1 billion | Industry funding reports |
| Meta Muse reported users | 2.8 million | Prior shattered.io coverage of Meta disclosures |
| OpenAI data center spending plan | Tens of billions of dollars | Reuters |
| dots regional availability at launch | Excludes UK, EU | OpenAI DevDay announcement |
Note that a widely circulated figure claiming dots carries a $100-per-month price tag has not been confirmed by an official OpenAI source, so it’s treated here as unverified rather than fact. Pricing details for dots specifically, as opposed to the broader ChatGPT Pro and Business Premium plans it rides on, remain something OpenAI has not spelled out in full.
Why Enterprises Are the Real Prize
Consumer AI chat products generate headlines and download counts, but enterprise contracts generate the multi-year, high-margin revenue that public markets actually reward. A single large enterprise deal for agent software can be worth more than millions of free consumer downloads, because it comes with predictable renewal revenue and rarely churns the way a free app does. That’s the calculation driving OpenAI, Meta, and Google toward the same target at nearly the same moment.
It also explains why the admin-approval gate on dots’ Enterprise tier isn’t a weakness, it’s a feature aimed at IT buyers. Security and compliance teams have spent years building processes to vet SaaS vendors, and an AI agent that can independently manipulate sales proposals or write code without a controlled rollout is a much harder sell than a chatbot answering questions in a sandboxed window. Google’s governance-first Gemini Enterprise Agent Platform pitch is built entirely around addressing that same buyer anxiety.
Historical Context: From Chatbots to Autonomous Agents
It’s worth remembering how fast this shift happened. Barely three years separate the original ChatGPT launch, which was framed almost entirely as a consumer research and writing tool, from a market where the same company is pitching software that autonomously updates sales documents and builds demos without a human typing every instruction. Anthropic has made a similar jump, and Shattered.io has reported on Claude handling a reported 26% of Anthropic’s own AI research and development work, a sign that the agent trend isn’t unique to OpenAI. Even coding tools have followed the same arc, as covered in reporting on an agent that retrained itself and leaked secrets during testing, a reminder that autonomy brings real operational risk alongside the productivity pitch.
The cloud infrastructure wars of the past decade offer a useful parallel too. AWS, Azure, and Google Cloud spent years competing on price and raw compute before the fight shifted to developer experience and managed services. The AI agent race looks like it’s skipping that early price war and jumping straight to the managed-service, governance-focused phase, likely because enterprise buyers who lived through cloud migrations already know what questions to ask a vendor before signing.
Market Impact: Stocks, Spending, and Vendor Lock-In
None of this is happening in a vacuum. Meta’s own AI momentum has already moved its stock price this year, something Shattered.io tracked when Meta shares hit a fresh high tied to the Muse launch. If Meta Enterprise Platform gains real traction with business customers, expect a similar market reaction, since investors have shown they’re willing to reward enterprise AI revenue signals even before the revenue itself shows up in quarterly filings.
For enterprise buyers, the practical risk is vendor lock-in. A company that builds its sales workflow around dots, wires its internal tools to Meta Enterprise Platform, or grounds its data pipeline in the Gemini Enterprise Agent Platform is making a multi-year infrastructure bet, not just picking a chatbot. Switching costs for agent platforms that touch CRM data, internal documents, and communication tools like Slack and Teams will likely be far higher than switching a simple chat interface, which gives whichever platform wins early enterprise deals a durable advantage.
The Regulatory Wrinkle
OpenAI’s decision to skip the UK and EU at launch is the most visible sign that regulation is shaping the rollout schedule, not just the product itself. Autonomous agents that can take actions across a company’s software stack raise data protection questions that go well beyond what a text-generation chatbot triggers under frameworks like GDPR. Any agent that reads customer data from a CRM, drafts a proposal, and sends it via email or Slack is processing personal data in ways regulators will want documented before allowing broad access.
Meta and Google, both of which already operate extensively inside EU compliance frameworks for their existing products, may end up with a structural advantage here if OpenAI’s UK and EU rollout for dots gets delayed further. A slower international expansion gives rivals time to sign enterprise customers in those regions before OpenAI’s agent product is even available to them.
What Enterprise Buyers Should Watch Next
For IT leaders evaluating any of these platforms, a few practical questions matter more than which vendor has the flashiest demo. How is agent activity logged and audited? Can access be revoked instantly if an agent misbehaves? What happens to data an agent has already read or acted on if a contract ends? These are the same questions companies learned to ask during early cloud migrations, and they’ll define which of the three platforms, or which well-funded startup like Instinct, actually wins durable enterprise contracts rather than just headline attention.
5 Predictions for the Enterprise AI Agent Race
- OpenAI will likely extend dots to the UK and EU within the next two to three quarters once it has a compliance framework ready, following the same pattern ChatGPT’s own international rollout took.
- Expect at least one acquisition of a smaller agent startup by OpenAI, Meta, or Google within the next year, as buying a working product becomes cheaper than building governance tooling from scratch.
- Enterprise pricing for agent platforms will likely shift toward usage-based billing tied to tasks completed rather than flat per-seat licensing, mirroring how cloud compute pricing evolved.
- Google’s governance-first positioning will probably resonate most with regulated industries like finance and healthcare, while OpenAI’s broader app integrations win faster-moving tech and sales-heavy companies first.
- Instinct’s $1 billion raise will likely trigger a wave of copycat funding rounds for narrower, vertical-specific agent startups over the next six to twelve months, similar to what happened after early generative AI funding rounds in 2023.
Frequently Asked Questions
What is OpenAI’s dots agent system?
Dots are “always-on” AI agents built on OpenAI’s GPT-6 Astra model, introduced at DevDay on September 29, 2026. Each dot runs on its own cloud computer and works toward a user-defined goal continuously, connecting to more than 4,000 apps including Slack and Microsoft Teams.
How is Meta Enterprise Platform different from Muse?
Muse is Meta’s consumer-facing AI agent, while Meta Enterprise Platform is a separate business-focused offering announced to help companies build AI-driven workflows, positioned as the next major pillar of Meta’s business according to the company’s own announcement.
What is Google’s Gemini Enterprise Agent Platform?
It’s Google Cloud’s platform for businesses to build, scale, and govern enterprise-grade AI agents grounded in company data, separate from Google’s consumer-facing agent efforts reported under the name Gemini Spark.
Why did OpenAI skip the UK and EU with dots at launch?
OpenAI has not given a detailed public explanation, but the exclusion lines up with the extra data protection and compliance review that autonomous, action-taking AI agents are likely to face under EU and UK regulatory frameworks compared to standard chatbots.
Who is Instinct and why does its funding matter?
Instinct is reported as an independent AI-agent startup that raised $1 billion in its latest funding round. The size of that raise signals investors believe enterprise demand for AI agents is substantial enough to support well-funded competitors outside the three largest labs.
Is dots available to all OpenAI customers right now?
No. Pro and Business Premium subscribers were reported to receive initial access first, while Enterprise customers need workspace administrator approval, meaning full rollout is staged rather than immediate for every tier.
How much does OpenAI’s dots agent cost?
OpenAI has not published a standalone official price for dots. A $100-per-month figure has circulated in aggregated reports but is not confirmed by an official OpenAI pricing source, so it should be treated as unverified.
Which company is best positioned to win enterprise AI agent contracts?
It’s too early to call. OpenAI leads on app integrations and model capability claims, Meta is leaning on its existing advertiser and developer relationships, and Google is betting that its governance-first approach will win over regulated industries. Independent startups like Instinct add a fourth variable that could disrupt all three.
