In Brief
OpenAI's DevDay centerpiece was Dots, AI assistants that can keep working between conversations.
The company also released a cheaper model, faster options, and new ways for teams to work in ChatGPT, but daily use of always-on agents remains an open question.
What Happened
At its San Francisco DevDay, OpenAI introduced Dots, AI assistants that can use their own cloud computers, connect to your apps, and continue a task after you leave the conversation. A Dot could read an interview transcript, propose video clips, and prepare show notes or social posts for approval. It can also look for useful work in connected apps without a fresh request.
Dots led more than 20 announcements.
-
GPT-6.1 Sol offers many of Astra's coding and computer-use abilities at one-fifth of its standard price for developers.
-
Astra gained an Ultra Fast option.
-
Codex, OpenAI's coding assistant, can now run in the cloud, while ChatGPT Space lets coworkers and AI agents collaborate around shared knowledge.
-
Developers gained tools for agents that use computers and work together.
The announcements also changed the value proposition for paid users. OpenAI launched a $500-a-month Pro tier with a higher usage allowance and lowered the allowance for its $200 plan. OpenAI says more efficient models should deliver more value per task.
SmarterX founder and CEO Paul Roetzer discussed the launch and its adoption challenge, on Episode 245 of The Artificial Intelligence Show.
The Key Numbers
20+ - Major DevDay announcements across OpenAI's products
One-fifth - GPT-6.1 Sol's standard API price relative to Astra
8x - Claimed response-speed increase in OpenAI's coding assistant
$200 - Monthly price of the Pro plan with a newly reduced usage allowance
$500 - Monthly price of the new higher-usage Pro tier
Why Always-On Agents Still Face an Adoption Test
The launch did not produce a breakout moment. Roetzer says “the overall reaction was pretty meh.” Early reactions to Dots were mixed. The product can improve, but novelty alone will not bring people back every week.
Using them regularly or not. Roetzer thinks most people will experiment with an agent and then wonder what to do next. “People are going to struggle to know what to do with these things,” he says. Booking a trip is an easy demo. Daily assistance takes knowledge of someone's goals, routines, and unfinished tasks.
A question of giving access to sensitive information. A Dot needs visibility into where work lives: email, calendars, documents, and other accounts. Users can choose app permissions and approval rules, but they still have to share sensitive context. A business faces that choice across more people and systems.
“For these agents to identify use cases, they have to be connected to everything, which means you have to trust implicitly the platform that you're connecting them to.”
— Paul Roetzer, founder and CEO of SmarterX on Episode 245 of The Artificial Intelligence Show
Product launches are outrunning training. Four years after ChatGPT arrived, “most organizations are still struggling to drive adoption for the most basic capabilities of standard AI systems like ChatGPT and Claude,” Roetzer says. The flood of releases reinforced his push for continuing, platform-specific education when each update can make a static course obsolete.
SmarterX Take
Businesses should start with a recurring task and test whether a Dot can handle part of it under clear permissions and review rules. A weekly summary from approved sources is more revealing than a one-time experiment. Measure how often people keep, improve, or act on the result.
Pricing also deserves a practical test. A cheaper model or a higher usage tier matters only if it improves the cost and quality of a completed task. Teams should compare actual outputs across plans before assuming a more expensive model creates better value over a cheaper one.
What to Watch
Regular use will be the real verdict on Dots. A strong first week can reflect curiosity. Sustained use will show whether the work justifies account access and human supervision.
Apple could have an opening in personal agents. Roetzer thinks Apple could arrive late with agents and still win because people already trust it with their devices and sensitive data. A useful agent needs that access, and people have to be willing to grant it.
One in Four Organizations Has Scaled AI
Only 25% of organizations have reached the Scaling phase of AI adoption, meaning they use AI widely and consistently improve performance, according to the 2026 State of AI for Business Report. That gap helps explain why an always-on agent can be available but a company is uncertain if and where it fits in daily work.
The full report draws on more than 2,100 responses from professionals across roles, functions, and industries. It shows where organizations stand on adoption, training, and governance, the foundations that will determine whether tools like Dots become useful at scale. Read the full report →
Mike Kaput
Mike Kaput is the Chief Content Officer at SmarterX and a leading voice on the application of AI in business. He is the co-author of Marketing Artificial Intelligence and co-host of The Artificial Intelligence Show podcast.