Good morning. As the perception around the long-term impact of the Trump Administration’s tariffs changes and softens, stocks have been all over the place.

The Dow soared as high as 1,100 points Wednesday, but it closed up only 400.

Big Tech, meanwhile, had a great day, led by Tesla, which closed up 5% despite posting a remarkably weak earnings report.

We’ll see if it sticks, or if we’re looking at yet another dead cat bounce.

— Ian Krietzberg, Editor-in-Chief, The Deep View

In today’s newsletter:

• 🌊 AI for Good: Underwater robots

• 🔏 Anthropic details how Claude is being misused

• 💻 How Microsoft is justifying its new AI-centric 365 evolution

AI for Good: Underwater rescue bots

A team of researchers from Germany’s Fraunhofer IOSB's Institute for Applied Systems Technology (AST) is designing an underwater autonomous life-saving robot. 

The details: The idea — when dealing with drowning victims in a pool — is relatively straightforward: surveillance cameras pick up patterns of movement associated with a drowning victim, then send those coordinates to the robot. 

• Once sent, the robot leaves its docking station at the bottom of the pool, travels to the coordinates, locates the victim using onboard cameras and transports them to the surface. 

• When dealing with outdoor environments, such as lakes, drones and balloons can take the place of the surveillance cameras. 

In an early open-water test at a lake in Germany, the robot was able to find, secure and ‘rescue’ a dummy weighing 175 pounds. The robot brought the dummy to the surface — from a depth of three meters (9.8 feet) in around one second, then identified the shortest route to shore, where it delivered the dummy to a waiting team of first responders, who the robot alerted to the emergency the moment it received a signal from the overhead drones. 

The whole operation took around two minutes, enough time to resuscitate victims while avoiding permanent damage.

Why it matters: The country — and the world — has been dealing with a growing lifeguard shortage for years, now. More than 4,000 people in the U.S. die from drowning every year. 

7 Reasons Agentic AI Leads the Way in CX Automation

Agentic AI represents a major breakthrough in the evolution of AI — ushering in a new generation of autonomous AI Agents capable of complex decision-making and goal-driven behavior.

While the potential applications span industries and user groups—from enterprises to end consumers—its impact on customer service automation is especially transformative. Purpose-built implementations tailored for contact centers enable faster resolution, fewer handoffs, and more natural, humanlike experiences at scale.

Read this article to discover 7 ways Cognigy empowers contact centers to unlock the full potential of Agentic AI.

Anthropic details how Claude is being misused

When we talked the other day about AI as a normal technology, the core point — when it comes to AI-related risks — had nothing to do with superintelligence. It had to do, instead, with flawed humanity. 

“That is because these risks arise from people and organizations using AI to advance their own interests, with AI merely serving as an amplifier of existing instabilities in our society,” the paper reads.

In a truly great example of this in action, Anthropic shared a somewhat detailed report on the ways in which bad actors are misusing its Claude chatbot. If generative AI is the fighter, social media is the battlefield. 

• The “most novel” application here involved what Anthropic calls an “influence-as-a-service” network, where Claude was used “to orchestrate over a hundred social media bot accounts for the purpose of pushing their clients’ political narratives … the operation utilized Claude to make tactical engagement decisions such as determining whether social media bot accounts should like, share, comment on or ignore specific posts created by other accounts based on political objectives aligned with their clients' interests.”

• Across Facebook and X, the operation included more than 100 bots. Each was given a political persona, and each engaged with thousands of real accounts. 

Anthropic said the operation is “consistent with what we expect from state-affiliated campaigns,” though added that it wasn’t able to confirm whether this was, in fact, the case. 

Other examples included attempts to use Claude to forcibly gain access to security cameras, as well as the operation of a sizeable scam employment ring in Eastern Europe intended to steal personal information from unsuspecting victims. One user leveraged Claude to develop better malware than they were capable of writing themselves. 

Anthropic said it identified and banned each of these users, adding, however, that it’s unsure whether the scams were successful. 

Abuse has been tied to generative AI almost from the moment ChatGPT went live. Deepfake abuse, accelerated fraud and scams, social engineering attacks, election-related fraud, general harassment … the list goes on. We’re talking about a tool that is supercharging cybercriminals. 

And the numbers, each time they come in, are staggering. 

The FBI’s Internet Crime Complaint Center reported $16.6 billion in total losses to online fraud last year, a 33% increase over 2023’s numbers. The average loss was nearly $20,000. 

Spring Cleaning, But Make It Personal. 😎

Apply that fresh start energy to your face - it's time to upgrade your grooming routine.

For a limited time, get 25% off The Chairman™ Pro Package - the complete face shaving kit from MANSCAPED®. This simple three-step routine is all you need to spruce yourself up and look your best.

✨ The Chairman™ Pro electric foil face shaver
🤩 Power Shave Gel - hydrating shaving gel
😃 Face Shave Soother - aftershave serum
😁 SkinSafe® Technology helps reduce nicks, cuts, irritation and razor burn
🌿 Shaves up to five days of hair growth down to the skin
💦 Waterproof for easy clean-up
✈️ Airplane and travel-friendly
🤑 Try risk-free with a 30-day money back guarantee

Hurry, this offer won't last long.

Order today and save 25% at checkout + free shipping!

• Oscar doesn’t care: The Academy Awards posted an update to its rules this week, saying that the use of generative AI tools would “neither help nor harm the chances of achieving a nomination.”

• Policy Push: Texas’ House of Representatives has passed HB 149, the Texas Responsible Artificial Intelligence Governance Act. The bill would enshrine certain civil liberty protections pertaining to the use of GenAI, requiring, for instance, high levels of transparency around the use of the technology. But now it’s on to the Senate.

• Musk’s Neuralink eyes fundraise at $8.5 billion valuation (Bloomberg).

• Kenya’s solution to teacher shortage: Embrace AI (Rest of World).

• Law experts, former employees push to stop OpenAI’s nonprofit structure change (Semafor).

• Bad news for China: Rare earth elements aren’t that rare (Wired).

• Meta makes ads on Threads available to all eligible advertisers (CNBC).

How Microsoft is justifying its new AI-centric 365 evolution 

Microsoft is going all-in on generative AI “agents,” that fuzzy term that Microsoft defines as an application of generative AI that goes a step further, actually doing work on your behalf. 

On Wednesday, Microsoft announced its Microsoft 365 Copilot Wave 2 spring release, centered around an updated Copilot app that’s designed to serve as “your window into the world of agents.” 

The details: The update includes an AI-powered search function and a new “agent store,” where customers can shop for and select predesigned agents. 

• The cheapest Microsoft 365 plan for businesses starts at $6 per month per user, and includes access to Copilot chat. But if people want to snag any agents from Microsoft’s agent store, they’ll have to subscribe to Azure, where they can pay for the agent on a metered basis. 

• Microsoft said in March that it will charge customers $.12 “per interaction” for the use of its SharePoint agents. 

The announcement coincides with Microsoft’s publication of its 2025 Work Trends Index report, a report that is serving as the cornerstone of the product update. The report this year is the result of survey data from 31,000 workers across 31 countries, LinkedIn trends and “trillions of Microsoft 365 productivity signals.” 

The focus of the report is something Microsoft is terming a “frontier firm,” a type of company with the following traits: “org-wide AI deployment, advanced AI maturity, current agent use, projected agent use and a belief that agents are key to realizing ROI on AI.”

Of the 31,000 people sampled, 844 employees work at companies that meet that qualification. That’s roughly 2.7%. And it’s not clear if there’s any employer crossover among those 844 employees. 

The key findings: Though these “frontier firms” are a bare blip in the context of every other company, Microsoft expects “every organization” to be on a journey to becoming one within the next two to five years. 

• 82% of “leaders” polled are confident that they’ll use “digital labor” to expand their workforce capacity over the next 18 months; 45% called that a top near-term priority. 

• 78% of these leaders are also considering hiring AI-specific roles, such as “AI trainers, data specialists, security specialists, AI agent specialists, ROI analysts, and AI strategists in marketing, finance, customer support and consulting.” 

38% of leaders expect their teams to soon be redesigning their business processes with AI. 

Microsoft added that employee mindset “will also be important. Digital colleagues aren’t just tools — they’re teammates.” Still, 52% of respondents see the tech as a tool, rather than a “thought partner.”

Microsoft said the “thought partner mindset” will be vital going forward. 

Bits and pieces of Microsoft’s report align with other reports on the state of enterprise AI adoption, but it paints a rosier picture than we’ve seen from other reports that have identified how few AI-related experiments make it out of the pilot phase and into actual production. 

• After pursuing around 900 different AI use-cases, Johnson and Johnson is opting for a more targeted, narrow deployment of specific AI-related applications. 

• After replacing human call center workers with AI agents, Klarna reversed course, opting to be the “best at offering a human to speak to.” 

Technological limitations and concerns, despite all the benchmark gains, still persist. 

When I asked Alexia Cambon, Microsoft’s Senior Research Director, about this, she said that she thinks “this is the worst AI will ever be. And if you think about that, it's kind of galling. These tools will only continue to get better, and they'll hallucinate less.”

Still, Microsoft says it's not interested in replacing human workers. 

Aparna Chennapragada, Microsoft’s Chief Product Officer of experiences and devices, told me that the Work Trends data shows a growing interest in upskilling. “57% of the leaders really want to get more out of the employees, more out of the organization,” she said. “And 60% of the employees say they’re maxed out. When you look at that stat, it screams augmentation to me. It screams upskilling them … and augmenting them. That’s the biggest opportunity that we have at this point in time.”

Cambon likewise said that we’re “not going to run out of knowledge work.” She just thinks that, with generative AI, it’s going to get more efficient. “Today, it takes work to do work … I spend probably half my day triaging email rather than actually replying to email. I wouldn't be surprised if in a couple of years, I'm not in my inbox ever because an agent is just tapping me on the shoulder, telling me what I need to know about what's in my inbox.”

In the face, however, of uneven adoption, of customers who just want what I’d call an analog, non-AI version of 365, Chennapragada told me that Microsoft has been thinking hard about how to “build a bridge” from the old 365 to the AI-invigorated 365, without sacrificing trust. 

She said that the base of this bridge begins with the free accessibility of Copilot chat to all accounts, something to “wet (the) appetite” of reluctant adopters. 

Less than 3% of respondents in a given survey doesn’t often indicate a trend. 

Perhaps the beginning of one, but without context and details, it doesn’t seem indicative of much.

But that number is at the core of the report, and it’s at the core of the 365 updates. The Frontier Firms. 

I think this is a good time to remember that companies are in the business of selling shit to customers. This alone is a good enough reason for a general layer of scrutiny and skepticism, especially when AI-related claims get fantastical. 

Everything in this report could well be true. And in a few years, we might see these early hints break out into full-on trends. 

But it’s not lost on me that agent use is charged on a meter, and greater use of AI in 365 requires more expensive subscriptions. And even with free Copilot access, Microsoft, according to its own privacy policy, “may use your data to develop and train our AI models.” 

This is the agentic push. This is the thing that justifies $80 billion in capital expenditures. 

For it to all pay off, people need to actually adopt it. 

🤔 Your thought process:

Selected Image 1 (Left):

• “There is a Grok watermark!”

You got me there, completely missed that.

… the one time I use Grok …

💭 A poll before you go

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.

P.S. Enjoyed reading? Take The Deep View with you on the go! We’ve got exclusive, in-depth interviews for you on The Deep View: Conversations podcast every Tuesday morning. Subscribe here!

If you want to get in front of an audience of 450,000+ developers, business leaders and tech enthusiasts, get in touch with us here.