OpenAI’s New Agent Delivers a Reality Check for the Enterprise

Good morning. Amid a mounting wave of stories about AI agents running wild, and just a day after shelving GPT-6.1 Astra over safety concerns, OpenAI unveiled an always-on agent called Dot, built on the earlier GPT-6 Astra model.

If Meta’s Muse is the friendly assistant for the masses, Dots—as they are collectively known—currently appear more built for work. For example, developers can use a Dot to run tasks in Codex. Access requires OpenAI’s $200-a-month Pro plan or some of its business plans, the WSJ reported.

Furthering the enterprise ties, OpenAI says users can reach a Dot through ChatGPT as well as through workplace tools such as Salesforce’s Slack and Microsoft’s Teams.

As my colleagues have reported, AI agents have been the “stuff of enterprise tech” for several years. Some organizations have even stood up human-resources teams around these so-called digital workers. Agents like Dots, which keep working in the cloud after users log off and can be reached through workplace tools, can seem like new employees, but ones with a track record the industry is still learning to read.

OpenAI says Dots come with built-in safety and privacy protections to keep them within the scope of users’ instructions. The company also cautions that Dots can still make mistakes. Given the recent run of AI agents going rogue, it’s one more reminder for enterprises that governance needs to be in place before agents are let loose.

The stakes are rising. Agentic AI accounts for 22% of the value companies get from AI this year, up from 17% in 2025, according to a Boston Consulting Group report on AI agentic value released today. BCG projects that share will reach 39% by 2030.

Scale brings its own risks. The more agents a company puts into production, the greater the danger of what BCG calls agent sprawl, with duplicate agents, conflicting permissions and unclear ownership among other issues. And because agents can combine data, tools, memory and other agents across workflows, the report warns, risks can “compound at machine speed.”

Most companies aren’t ready. BCG found that 42% of companies expect AI agents to have real decision-making authority by 2030, but only 5% have enterprisewide controls to do that safely today. “The technology is ready to act but the guardrails aren’t,” said Vlad Lukic, BCG’s global lead for AI at Scale. “Companies’ ambitions for truly autonomous agents are outrunning governance infrastructure, signaling that leaders need to act now to meaningfully benefit from this trend down the line.”

In other words, before companies unleash agents like Dots, they may need to connect a few of their own.

Three Questions for Toshiba America’s Chief Information Officer

Andrea Schulze Dias, vice president and group chief information officer of Toshiba America, is a nearly two-decade Toshiba veteran who oversees a variety of businesses—from industrial operations to hardware and energy. With such a wide remit, there are unique challenges: Dias needs to support a different technology strategy for each business, she recently told me.

And as a manufacturing-focused company, another challenge for Toshiba continues to be the transition from manual, even paper-based processes, to digital ones and automation. Or, in other words, “making sure that we interconnect, so we have access to the data that then will be the precursor for future AI innovations,” Dias said.

I asked Dias about her role as CIO, and how she sees the current AI landscape in areas including token cost and change management. Her responses have been lightly edited for clarity.

WSJ Leadership Institute: You say that there is still some foundational work remaining before you can fully leverage AI, but how have you used AI so far?Schulze Dias: Initially, AI was solely seen as a cost optimizer to reduce head count or reduce head cost. But we’re seeing AI very differently now that we’ve experimented with it and have a bit more notches on our belts. It is an accelerator.

We’ve improved, in our field services operations, the ability to auto-assign certain cases to individual field service workers. We’ve seen our service level significantly improve there, which results in higher case capacities. That means higher revenue.

We’ve seen improvements in quote automations and in using AI to identify errors in quoting and in requoting, rework, and the speed in which the quote can be produced to the customer. And we’ve seen some revenue growth there as well.

WSJLI: How do you select the AI tools and models you want to invest in, all while managing AI token cost?

Schulze Dias: Now that we’re in this tokenomics world, we’re being even more conscious of what models we are using for what tasks, which was not so much in the forefront for us before.

There are some concerns that you are investing into an AI solution and it becomes obsolete in a year. I’ve seen competitors and other companies within industries, they’re really only signing contracts for a single year instead of going multiyear because technology shifts so quickly. Deciding on what platform to use becomes even more challenging.

We were really hit hard with the GitHub situation: We are a heavy Microsoft shop, and when Microsoft announced in June that they are going consumption-based on GitHub, that hit us pretty hard, especially in our software development side.

Microsoft does give some visibility on the back end to administrators, so we had to revamp our allocation model entirely. You can set limitations and thresholds, so we can say, ‘I don’t want you to go over $3,500 in spend this month,’ and then it actually cuts it off. But there is this challenge where you don’t want to cut it off and hinder productivity.

WSJLI: How have employees responded to your rollout of AI tools internally, and how do you think AI adoption will ultimately play out?

Schulze Dias: A few years ago, we really pioneered in our organization here in North America an AI Champions program. We really wanted to widely make AI education available.

Now we are in this next phase where everybody can be an AI champion, but we’re seeing individuals who are still not embracing AI for various reasons: some being ethical and thinking that AI will take over the world, and they don’t want to feed into the machine. Then we have others that are worried their job is going to be replaced with AI, and then we have the ones that are really on board, embracing that change.

I don’t believe everybody will become an AI champion, or be that AI person. I think there are the high-performers that are already adopting the technology, and they will accelerate even further.

The WSJ Tech Council brings together CIOs, CTOs and CISOs advancing innovation and shaping the future. Join this trusted community where tech executives connect with peers to explore emerging trends and gain the perspective they need to stay ahead of disruption.

Request Information

Follow Isabelle Bousquette on LinkedIn, Instagram, X, and TikTok for more behind the scenes on her tech and AI coverage, and lately, her contributions to the WSJ Leadership Institute’s new Executive Resilience series, where she’s profiling America’s top execs about their fitness and wellness habits.

Follow Belle Lin on LinkedIn and X for her latest reporting on enterprise technology and AI.

Steven Rosenbush is chief of the enterprise technology bureau at the WSJ Leadership Institute. He also has a column. You can follow him on LinkedIn.

Tom Loftus is the editor of The Morning Download. He suggests following Isabelle, Belle and Steve on their various social channels. But if you insist, here’s his LinkedIn.