OpenAI DevDay 2026: What Changes for Business Automation?

Plugins can expose service information and supported actions to customers inside ChatGPT. For 13labs, that opens a concrete build opportunity: connect a useful customer task to the sales and delivery systems behind it. Start with service fit, current information or an authorised request status, and assess enquiry eligibility before including a conversion flow. A public listing and selective enhanced distribution are different routes. Plan acquisition through your existing channels and measure useful enquiries rather than assuming publication guarantees traffic. Read ChatGPT plugins for business services, then explore our plugin development offer.

Dots: consider ongoing research or monitoring. A sales owner might want a reviewable brief on upcoming prospects. OpenAI's Dots documentation explains access requirements. Our Dots business guide turns that into a bounded pilot. GPT-6.1 Sol and Ultrafast: compare the cost and waiting time of an AI workflow. Read our automation cost guide before treating lower token prices or faster generation as a business return. Space and Pages: consider a shared project brief needing current context, clear decisions and a responsible owner. The Space handover guide explains how to keep it useful. Codex Cloud and Security Cloud: consider building and maintaining your internal tool. Our Codex guide covers a reporting backlog and review process. Agents API computer use: consider a browser-based step where your usual integration is unavailable. The computer-use guide explains when it deserves a pilot. Plugin Extensions, Sites, MCP Events and Sign in with ChatGPT: consider reusable access to company tools, an internal interface, an event that starts work or eligible account-based access to an app. Our connected workflow guide separates those decisions. Slack, Teams and Team Tasks: consider turning discussions and recurring updates into owned work. The team workflow guide starts with a sales-to-delivery handover.

An enquiry can pass through several tools before anyone has a useful conversation with the buyer. A form collects a message, an inbox receives it, someone researches the company, and the CRM waits for an update. The loss often happens between those steps. Buying an agent does not establish who owns the enquiry or what a qualified lead means. Map the current path using a recent enquiry. Record where the message arrived, when it was assigned, what information was missing and how the next action was decided. Separate fixed rules, such as routing by location, from judgement, such as assessing whether a vague request fits your offer. Fixed rules usually deserve a direct automation. Judgement may deserve an AI step with a reviewable explanation. A proposed pilot could gather public company context and prepare a draft qualification brief for the assigned salesperson. Keep the contact record and ownership in the CRM. Let the salesperson approve any external reply. Track time from enquiry to owner, time to approved reply and whether the brief was useful. A convincing demo is less informative than those measures over actual enquiries. This is the connection to 13labs revenue automation: the aim is a useful sales conversation with fewer missed steps, measured across the whole workflow.

The next gap often appears when a deal closes. Delivery needs the customer's goals, agreed scope, outstanding questions and sales commitments. Those details may sit across a proposal, a CRM record, messages and a meeting note. Asking an agent to summarise everything can produce a polished document that still leaves the delivery owner guessing. Define a handover with required fields: customer owner, approved scope, commitments, dependencies, first milestone, unresolved questions and the source for each claim. A shared Page could hold the reviewable brief. A Team Task could help prepare it on a supported schedule or event. A plugin could expose the approved CRM record. Which combination works depends on the connections and permissions available in your account. Treat missing information as a visible exception. An agent should mark a start date as unconfirmed when the proposal and CRM disagree. The delivery owner resolves the conflict before scheduling work. That design is more useful than asking the model to choose whichever date sounds plausible. The operational outcome is a complete brief that delivery accepts. 13labs operations automation covers onboarding, reporting and handover work. The document is one part of the process; ownership and the next action matter just as much.

Use a small, explicit worked example. Suppose a team prepares 30 enquiry briefs a week and spends 10 minutes on each. That is 5 hours of current effort. If a pilot still needs 4 minutes of review per brief, it leaves 2 hours of review and suggests a potential reduction of 3 hours. This is an illustrative calculation, not a measured customer result. Include time spent collecting missing information, correcting mistakes, maintaining connections and checking failed runs. If those jobs take another 2 hours, the apparent gain falls to 1 hour. That might still be useful if response quality improves, but the original estimate would have overstated the return. Use the same logic for reporting and software changes. Count the whole job until the next person can use the output. A cheaper model is valuable when it lowers cost per accepted result. A faster tier is valuable when it reduces waiting that blocks someone. A cloud tool is valuable when useful work can proceed without the operator present. Record the assumptions beside the calculation. Volume, review time and exception rate are often more sensitive than the headline subscription price. Return to the estimate after the pilot and replace assumptions with observed results.

Choose a workflow with a stable input, a useful output and a person who can judge it. A weekly internal pipeline brief is easier to assess than an agent responsible for the entire sales operation. A handover checklist is easier to verify than an instruction to improve customer success. Write the pilot brief before configuring tools. Include source systems, permitted actions, sample records, reviewer, success measure and what happens when the agent cannot finish. Check account availability against the relevant OpenAI documentation. A launch announcement does not establish that a particular plan, connector or workspace supports your design. Collect ordinary and difficult inputs. Include an enquiry with no company name, a conflicting date, an existing customer mistaken for a prospect and a record the connected account cannot access. A pilot should show how the workflow handles those cases. Agree on a review date and a named owner. Keep the pilot small enough that stopping it leaves the business able to operate normally. Extend a successful workflow one step at a time. If it fails, the evidence should tell you whether the problem was the tool, source data or process.

Should we replace existing automations after DevDay? Review a specific failure or cost before replacing a working system. A fixed routing or data-transfer step may already be well served by your current tool. AI is useful where interpretation, research or drafting adds value. Which announcement should a small business try first? Start with a job that has clear inputs and an easy review. That could be a draft sales brief, a project handover or a recurring report. The best first release depends on the gap and available connections. Can 13labs help decide between these options? Start with the workflow audit, then discuss a scoped build or work alongside your team through buildAutomation. Identify the first worthwhile workflow, its integrations and who will own it.

Product facts were checked on 1 October 2026 against the official OpenAI DevDay documentation. Follow the topic guides for release-specific sources and limitations. Availability can change during rollout. Run the workflow audit with the repeated jobs your team handles. Bring that ranked list to an automation discussion. It gives the conversation a concrete starting point: a process, an owner and a measurable outcome.