As usual, an eventful week in AI; lots of stories about corporate adoption of AI, and more offerings for AI agents. Here’s a quick rundown of what it means for business.

AI at Bank of America

Bank of America shared their success stories in deploying AI, one of the companies that was a fast-mover for adoption. They created an AI-powered IT Assistant to take care of IT requests, which is now used by 210,000 associates, saving time for everyone (especially the IT staff!). They also embraced AI coding tools, now in use by 17,000 programmers. These two uses are heavily leveraging generative AI, and serve as a role model for other companies.

AI at NovoNordisk

Another company with high AI adoption is NovoNordisk. This article in MIT Sloan Management Review(subscription required) describes their journey (mostly with Microsoft Copilot, for 20,000 employees). They found that, when deployed properly, it was truly transformative.

“…generative AI is uniquely disruptive, reshaping the nature of work itself in unprecedented ways.”
– How to Scale GenAI in the Workplace

Generative AI not only improved productivity, but employees reported higher satisfaction and felt their work quality improved. Their findings confirm that it’s never just about the technology:

• Adoption was nonlinear – don’t expect use to just keep growing; you’ll need an enablement plan
• Generative AI’s natural variability undermined employee’s trust, which was a big obstacle to adoption
• Don’t assume the younger generation will drive adoption; while they may be digital natives, more experienced employees were better able to spot opportunities for the greatest impact

Speaking of AI Adoption…

AI adoption is not primarily generational. And it’s not about the technology either. Conor Grennan has helped many companies wrap their arms around AI, and he always says it’s not about the technology; the key is changing human behavior. Here’s a great piece from his AI Mindset blog where he outlines three factors to change your org to be AI-focused:

• Make AI adoption mandatory, not optional (it’s a net plus, use it)
• Frame AI as amplification, not replacement (to mitigate fear)
• Create concrete incentives and expectations (make it cultural)

What’s Standing In the Way?

Artificial Analysis conducted a survey about the current state (and progress) of AI adoption; if you’re deploying it in your business you’ll want to take a look. Since it’s a survey the findings aren’t universally applicable, but this chart captured the challenges of adopting AI in the workplace well:

Intelligence and reliability were the top two concerns – which is all about the technology working. Intelligence is “can it get this right?” and reliability is “can I trust it to do it right?” Cost came in third.

AI Literacy = Faster Adoption, Right?

Lots of people are confused about generative AI. Some are afraid of it. Most don’t know what to do with it. So it’s natural to assume that those who are more familiar with generative AI are more likely to use it and adopt it, right?

“…lower AI literacy predicts greater receptivity to AI.”
– Why Understanding AI Doesn’t Necessarily Lead People to Embrace It

Not so fast. An interesting marketing study summarized in Harvard Business Review shows the opposite: employees who are less knowledgeable about generative AI are more likely to use it. People who don’t know much about AI feel like it’s magical, and can’t wait to try it. People who know how it works have demystified the magic, and either think it’s not a big deal, or are aware of the problems (hallucinations. etc.) and approach it with lower trust. This has interesting implications for adoption efforts, since giving people a greater understanding of the technology may take away the desire for them to use it.

AI Agents

Amazon’s Agent IDE: Kiro

If your business does a lot of coding, Amazon’s new agentic programming environment (called Kiro) may be of interest. It’s an agentic IDE (Integrated Development Environment) to assist programmers in developing applications. This is Amazon’s entry into the market of Cursor (who turned down an acquisition offer from OpenAI’s) and Windsurf (acquired by Cognition, right after Google acqui-hired Windsurf’s CEO, right after OpenAI expressed interest in acquiring them). While not particularly important for non-coders, the fact that Amazon is entering this space (powered by Antrhopic’s Claude) indicates that they think this market has a lot of opportunity and they want to be a part of it.

Kimi K2: Powerful Open Agentic Model

In the meantime, Moonshot AI released an open LLM designed to be used for AI agents, Kimi K2. It has state-of-the-art performance on many metrics, even compared to the closed models from the big players. This provides an alternative for companies that want to deploy AI agents without being subject to those big players. The agent wars are heating up!

And in non-business AI news…

OpenAI’s Agent

OpenAI released a general-purpose AI agent for subscribers that just might be able to do almost anything you ask it. Previous agents have had difficulty staying focused to complete the task, but OpenAI thinks they’ve made enough improvements that the tech is ready. It is impressive – it got a 41.6% on Humanity’s Last Exam (Grok 4 recently got the best score ever, 44.4% when using its multi-agent mode) so this is state-of-the-art performance. Of course it’s compute-intensive (those on the Plus subscription only get 40 queries per month).

While I put this in the non-business section, it’s very clear that this is a step towards OpenAI providing AI agents for business use too.

Google Will Call Businesses So You Don’t Have To

Google upgraded its “AI Mode” in search to use its latest model, Gemini 2.5 Pro (so now you get the best model when searching the web) and added a new feature: business calling. That means Google will (using an AI agent) call businesses for you to gather information that you need, so you don’t have to.

Unlike their first rollout a few years ago, this AI identifies itself as a robot when it calls the businesses. At first I thought businesses might rebel against a robot taking up their time, with an unknown payoff…but ultimately the robot is calling them because someone has asked it to, so maybe they need to respond just as if it’s a human, or risk losing business.

---

---

AI adoption is slower than the hype, but it is happening. As we’ve seen from the reports about NovoNordisk and Bank of America, companies that have adopted generative AI early are seeing benefits, and those benefits are significant. There is ample evidence for both quantitative gains (productivity, headcount) and qualitative gains (higher employee satisfaction and higher quality work) that there is no question that generative AI is a net positive with broad applicability. These companies are examples that others should follow, and if you’re not aggressively pursuing AI in your enterprise, it’s time.

These gains are coming (mostly) in the form of AI Assistants rather than Agents – they are examples of humans interacting with AI, not AI interacting with AI. Which means that as these companies continue their journey to agentic AI, the gains are likely to continue. The AI companies see this, as evidenced by Kimi’s agent-focused AI and OpenAI’s general-purpose agent.