#ai-agents

#ai-agents

Every executive I speak with wants to deploy AI agents in their business. Yet most are making the same costly mistake: choosing the wrong tasks to automate. In my previous article, A Beginner's Guide To Building AI Agents, I explained how to get started with agentic AI. Now it's time to tackle the most critical step: finding the right jobs to use them for. Get this wrong, and you'll waste time and money. Get it right, and you'll transform how your business operates.

Amazon ( NASDAQ:AMZN) has been a quiet laggard in the Magnificent Seven this year after gaining just over 3% year to date. With just a few trading days left in the year, it's looking like the $2.45 trillion e-commerce juggernaut is about to disappoint yet again, despite all the encouraging AI projects going on behind the scenes, from "frontier" AI agents to those impressive Trainium3 AI chips, the rollout of Alexa+, and let's not forget about the warehouse robots.

In late 2025, Mark Karpelès, ex CEO of Mt. Gox, lives a quieter life in Japan, building a VPN and an AI automation platform. As Chief Protocol Officer at vp.net-a VPN that uses Intel's SGX technology to let users verify exactly what code runs on servers-he works alongside Roger Ver and Andrew Lee, the founder of Private Internet Access. "It's the only VPN that you can trust basically. You don't need to trust it, actually, you can verify".

David Cohen, Chief executive of the Interactive Advertising Bureau (IAB), has climbed onto the dangerous prediction limb, with his 2026 Predictions, oppulently titled Search Takes a Backseat as Brands Battle Inside AI's Black Box. AI seems to be top-of-mind, as four of the predictions (including the top three) focus on artificial intelligence. We will spill those top three below, but leaving out the editorial content which accompanies each of his forecasts:

Last week, Google DeepMind made the Interactions API available as a public beta. The new API represents a fundamental change in how developers work with AI models: from stateless to a stateful architecture with server-side context management. With this move, Google is following the path that OpenAI embarked on in March 2025 with its Responses API. Over the past two years, developers have been working with generative AI via a so-called 'completion' model.

Something new is now taking shape on the factory floors. AI agents, independent, context-aware and task-oriented, are functioning as a third layer of intelligence. Not a replacement for what came before, but a layer that complements and elevates it. These agents are not confined to a single screen or workflow. They move between systems, interpret context via semantic data, and solve problems across functional boundaries.

the one thing I have become most convinced of is that product teams are really in a unique position to unlock AI value for their organizations.

"Most of us are producers today," Nancy Xu, Vice President of AI and Agentforce at Salesforce, told the audience. "Most of what we do is we take some objective and we say, Okay, my goal is now to spend the next eight hours today to figure out how to chase after this customer, or increase my CSAT score, or to close this amount of revenue." With AI agents handling more tasks, Xu said that workers will shift "from producers to more directors." Instead of asking "how do I accomplish the goal?" they'll instead focus on "what are the goals that I want to accomplish, and then how do I delegate those goals to AI," she said.

Salesforce CEO Marc Benioff last week came closer to answering a multibillion-dollar question when he said seat-based pricing - with some caveats - was becoming the norm for its AI agents after flirting with pricing based on consumption and per-conversation payments. While there is much debate about how users will get money back from investing in AI agents, the stakes are also high when it comes to how they pay for them.

"The agent underneath is actually more universal than we thought." Instead of building new agents for every use case, companies should rely on a single general agent powered by a library of skills, Zhang said. Skills are "organized collections of files that package composable procedural knowledge for agents," Zhang said. They are simply folders that contain whatever an agent needs to complete a task consistently and efficiently.

Microsoft has begun rolling out a public preview of native support for the Model Context Protocol (MCP) in the latest Windows 11 Insider builds, edging its much-touted "agentic OS" vision closer to reality. The update is rolling out to Windows Insiders on the Dev and Beta channels as part of build 26220.7344 and provides insight into where Microsoft is going with the technology.

November was the month AI went full gladiator mode: three frontier labs released their best models within a week, Google reclaimed the throne with Gemini 3, and open source proved it can win Olympic gold medals in mathematics. Meanwhile, the agent revolution became official doctrine at Microsoft and Google, and China's AI ecosystem hit escape velocity with 10 million app downloads in seven days.

Salesforce CEO Mark Benioff crowed to investors on Wednesday that the company's Agentforce IT Service, which debuted in October, was already winning customers and saving them money. He singled out PenFed, one of the nation's largest credit unions, as one of the early users and claimed PenFed is projecting a 30 percent cut in operating costs, and $2 million in savings.

Agentic AI is all the buzz in the business world as people envision creating and using special AI agents to carry out both routine and complex tasks on their behalf. Now, Google is launching a new product for Workspace users who want to see if they can benefit from such agents.

C loud computing has now entered its mature adolescence i.e. it's still surprisingly developmental, changeable and occasionally irrational in some areas, but overall it's certainly old enough to know better and should really start behaving properly. With the debate between public and private cloud now long over and the hybrid norm now (mostly) a de facto standard for typical deployments, multi-cloud itself is still an oft misunderstood state of being, with FinOps constantly berating us for waste and inefficiency.

In late October, OpenAI launched ChatGPT Atlas, a web browser with ChatGPT at its core. In Agent Mode, it can perform actions on your behalf, such as pulling together an online order. Elyse Betters Picaro, Senior Contributing Editor at ZDNET, tried it that day, successfully using it to place a same-day delivery order from Walmart. The experience was so positive that she then used it to buy Disney on Ice Tickets, which she said saved her some money and hassle.

Conversations Inbox - Keeps customer communication organized across email, chat, and social channels. Small business owners can now respond faster, prioritize conversations and stay on top of every customer interaction - all in one place. Marketing Calendar and Social Posts Agents - Helps small businesses plan and launch marketing campaigns and create social posts, videos, and blog content for channels like Facebook, Instagram and more, driving engagement and saving countless hours each week.

AI agents connect the power of large language models with APIs, enabling them to take action and integrate seamlessly into employee workflows and customer experiences in a variety of domains: World-class IT organizations are adapting their strategies and practices to develop AI agents while mitigating the risks associated with rapid deployments. "Building a world-class IT team means leading the conversation on risk," says Rani Johnson, CIO of Workday.

In this comprehensive interview at SAP Connect in Las Vegas, Stephan de Barse, President of Global Business Suite at SAP, makes the case for why best-of-breed enterprise software is being replaced by best-of-suite approaches, driven by the commoditization of the application layer through AI. De Barse, who previously spent seven years at o9 Solutions, a best-of-breed supply chain vendor, brings a unique perspective to the debate.

After The New York Times story was published, Guss, Ozair, and about three dozen other people updated their LinkedIn profiles to list their affiliation with the Bezos venture. Several of those people also work at Foresite Labs. Details about Prometheus remain limited. Its founding date, formal name, and headquarters haven't been publicly identified. But the dinner Bajaj hosted in June provided other clues.

Buried in that tidal wave was news of something called Entra Agent ID, the main idea of which is to use Microsoft Entra to govern AI agents in the same way that Entra currently governs human users; that is, to give each agent a unique, managed identity and apply familiar Entra identity controls such as conditional access, identity governance, and identity protection. Entra is Microsoft's cloud-based identity access management (IAM) solution.