B2B News

ChatGPT merges Chat and Work: what it means for B2B

OpenAI is merging Chat and Work in ChatGPT, making delegation to AI agents the default. What changes for B2B marketing and sales, and which processes to start with.
November 4, 2026
Krein News banner

On November 2, 2026, OpenAI emailed users that the Chat and Work tabs in ChatGPT will become a single experience, rolling out gradually from a small group of users. For B2B teams, delegating a task to an AI agent stops being a power-user feature and becomes ChatGPT's default behavior. Marketing and sales leaders need to decide now which processes to hand to agents.

📌 TAKEAWAY OpenAI claims 10 million weekly users across Codex and ChatGPT Work about 12 days after Work launched (July 9, 2026). The figure is self-reported, not verified by third parties, and counts both products together.

What did OpenAI announce on November 2?

OpenAI told users that Chat and Work become one experience inside ChatGPT. According to TechCrunch, which saw the email, the unified version goes first to a select group of users who will give feedback, and then to everyone else. Codex, OpenAI's software development agent, remains a separate experience in the ChatGPT desktop app.

How usage is measured changes too. From December, according to the OpenAI Help Center as reported by FourWeekMBA, ChatGPT and Codex draw on one shared allowance, valid for the Plus, Pro and Business plans. Instant chat does not consume it.

On October 29, in a separate announcement, OpenAI had already doubled the Pro plan's limits after criticism of a planned 50% cut (BleepingComputer). That announcement and the November 2 email are two distinct pieces of news.

How did Chat and Work get to a merger?

Work started as a mode of its own, and the merger arrives less than four months after its debut. MacRumors dates the launch of ChatGPT Work to July 9, 2026. It is an agent built on Codex and, at launch, on the GPT-5.6 model: it acts across apps and files, can stay on a project for hours and delivers spreadsheets, presentations, documents and small web apps. It connects to tools such as Slack, email, Drive, calendars and CRMs and, through scheduled tasks, keeps going while the user is away from the computer.

Anyone who wanted to use it had to know it existed and choose it. Adoption was fast anyway: on July 21, The Next Web and Implicator reported the 10 million weekly users that OpenAI declared for Codex and Work combined. Greg Brockman, OpenAI's president, had already indicated that Chat and Work would be merged by the end of the year (9to5Mac).

Anthropic took the same direction on September 16, 2026, 47 days before OpenAI's email, by merging Claude Cowork and Chat. Two of the main AI assistants now have a single front door, and behind it the system decides whether to answer or get to work.

Why does the merger matter for B2B marketing and sales?

Because it brings delegation to agentic systems into everyday use. Until now, asking (chat) and delegating (the agent) were two separate modes, and the second stayed with expert users. With one interface, whoever types a request leaves the system to choose between a reply and a multi-step job, across an audience of hundreds of millions of weekly users. [DATO DA VERIFICARE: official OpenAI figure, sources conflict between 900 million and 1.2 billion]

For anyone planning 2027 budgets, the question of how much to invest in AI, and where, becomes concrete.

📌 TAKEAWAY Gartner forecasts that by 2026, 40% of enterprise applications will include task-specific AI agents, up from less than 5% in 2025. It is a forecast published in August 2025, not a measurement.

A second data point shows where adoption stops. A SEON survey of 2,026 US adults, run between October 2 and 6, 2026, finds that 68% use AI agents for personal tasks and that 65% would blame the AI company if an agent committed fraud (SEON). The sample is American consumers, not B2B buyers: it is a hint about trust, not an estimate of the European market.

We read the shared allowance between ChatGPT and Codex as a new yardstick: value is measured more in work completed than in messages exchanged. That is our interpretation, and OpenAI does not state it.

What changes for B2B marketing and business development?

Five areas of work change. In each one, the agent takes the repeatable part and leaves judgment to people.

Content production. AI does not stop at a first draft and can deliver a package, from report to slides to landing page. Human value moves upstream, into the brief and the strategy, and downstream, into review and brand consistency.

Research and account intelligence. An agent connected to CRM, inbox and web can prepare a prospect dossier before a call or map a market's competitors. Among BDRs, AI use is already widespread: in 6sense's 2026 report, 99% say they use it. We have written about digital employees and AI agents in the enterprise and about the future of B2B sales with AI agents.

Reporting and analysis. Recurring campaign reports eat up hours of data gathering and layout. They are the natural candidate for an agent's scheduled tasks.

The buyer's side. Buyers will use agents too, to find suppliers, compare offers and build a shortlist. A B2B company therefore has to be understandable and quotable by an AI, with clear information, verifiable data and concrete case studies. We go deeper in the article on B2B buyers who ask AI first and in our B2B SEO and GEO service.

Process and governance. An agent that reads an inbox or edits a shared document needs rules: who decides what it can access, who checks the output before it reaches a client, how privacy is protected.

What risks come with handing processes to agents?

Four, and the first is the most insidious. An agent working for hours without supervision can produce a report that is flawless in form and wrong in substance. The more finished the output looks, the easier it is to trust it without checking.

Costs become less predictable. With an allowance that measures work done, long and complex tasks use more of it, so you need to know which tasks are worth delegating. Dependence on a single vendor hurts when prices, terms or quality change. And if everyone uses the same agents with similar requests, the results look alike, so tone of voice, point of view and sector expertise are what set a company apart.

For a concrete case of operational risk with agents, see the analysis of agentic AI risks in B2B.

📌 TAKEAWAY The shared ChatGPT and Codex allowance starts in December 2026 and covers 3 plans: Plus, Pro and Business. Instant chats do not consume it.

Which moves should you make now?

  1. Map your repetitive processes. Reports, preliminary research, first drafts, CRM updates: pick two or three pilot tasks and measure time saved and quality of the result.
  2. Get your data and sources in order. An agent is only as good as the information it can reach. An up-to-date CRM, tidy folders and a documented tone of voice separate usable output from output you have to redo.
  3. Set the rules. Decide which tools the agent can use, which data it must never touch, and what human review is required before any content leaves the company, in line with GDPR.
  4. Make your content readable by agents. Clear pages, structured information and concrete proof (numbers, references, case studies) are what an agent looks for when a prospect asks it to find the right supplier.

Krein helps B2B companies bring AI into marketing and sales processes: from workflow analysis to AI agents in B2B marketing and sales processes, and content built for AI search. Let's talk.

FAQ

Is the unified version of ChatGPT available to everyone yet?

No. According to TechCrunch it starts with a select group of users and widens after feedback. No dates have been announced for Europe. [DATO DA VERIFICARE: rollout timing in Europe]

Does Codex take part in the merger of Chat and Work?

No, it stays a separate experience in the ChatGPT desktop app. The link is usage: from December ChatGPT and Codex draw on the same allowance, so heavy Codex use can reduce what is left for ChatGPT.

Is the 10 million Work users figure reliable?

Read it with caution. OpenAI declared it, it counts weekly users of Codex and Work together, and third parties have not verified it. It does not measure users of Work alone, or how many use it for work.

Does the shared allowance cut the limits of paid plans?

The sources on the shared allowance describe a single allowance and do not report a cut to limits. The October 29 doubling of Pro limits is a separate announcement. To know how much work a plan allows, check the limits OpenAI lists when you activate it.

Are other AI assistants merging chat and agent too?

Anthropic did it on September 16, 2026, with Claude Cowork and Chat. For other vendors we have no verified sources.

Sources