By Yair Kuznitsov, Co-Founder & CEO, Anecdotes

Every week I talk to enterprise GRC teams who understand exactly what agentic AI can do for their profession. They’ve read the articles, seen the demos, and can articulate the difference between AI that makes a workflow go a little, or even a lot faster, and an agent that replaces it entirely.

Yet still, some remain reluctant to make the shift to agentic GRC.

When I ask why, the conversation moves away from technology pretty quickly. Most of them have the “AI budget” available, but something is holding them back from making the move and they can’t always name what it is.

The conversations all eventually lead to the same place, even if they can’t say it in so many words: they’re not sure who they are when the operations aren’t theirs anymore. It’s an identity and even value question above all else.

Most GRC practitioners carry an implicit belief about where their value comes from. That belief isn’t wrong, but it’s describing a role that’s being restructured, and those who make the transition the fastest will be the ones leading the industry in the coming years.

The Competence That Got Us Here

GRC professionals built their expertise around operational competence. Knowing how to gather the right evidence, managing audit cycles under pressure and keeping a complex compliance program running when it’s understaffed and under-resourced have been signs of a valuable GRC team member for years.

That competence took years to develop, and the people who have it are genuinely good at what they do and are rightfully valued by their business.

The problem with agentic GRC is that it doesn’t reward that competence the same way. Agents can gather evidence, open remediation tasks and can manage most of the audit cycle alone. Given that agents can handle those operations, the actual question is what a GRC professional is supposed to be doing instead, and most organizations haven’t asked it yet.

The Shift They’ve Been Waiting For

GRC wasn’t designed to be an operational function. It was designed to help organizations understand and manage risk. The evidence collection, the audit cycles, the status updates were always implementations of that purpose, not the purpose itself. The practitioners who got into this field weren’t drawn to it because of the “fun” of evidence collection.

They cared about whether the organization was actually protected, or just appearing to be, and wanted to provide that insight to the business.

What happened over time is that the tooling didn’t scale with the programs, and the operational burden consumed everything. The people who were supposed to be thinking about risk spent most of their time keeping the machine running, not because it was ever the point of the role, but because someone had to do it and there wasn’t another way.

What Agents Do, and What They Can’t

Agentic GRC doesn’t speed up workflows, it replaces them. Evidence no longer flows through a person; it’s pulled continuously from integrated systems. Controls aren’t checked periodically; they’re monitored in real time. Remediation isn’t tracked in spreadsheets; tickets are opened, assigned, followed up on, and closed automatically.

But agents don’t design themselves.The logic that drives them (what to collect, what constitutes a pass or fail, what triggers an escalation, what the auditor will accept as evidence) comes from a key combination: data context and human insight.

Someone has to define the risk appetite, decide what “remediated” actually means, know when the output looks right and when something is missing that the system can’t see.

Agentic GRC in Anecdotes is built around exactly this model. The agents handle the operations end to end, based on the robust data foundation we have spent years building, and the logic the GRC team defines. 

When agents can handle the evidence chains, control testing, and audit prep, the question of what GRC should actually be doing shifts. And for practitioners with real depth, that answer is what they’ve always known how to do. But that doesn’t make the shift easy.

Redefining a role is hard and comes with real fears. Many people are worried about their jobs because of AI, some more rightfully than others.

For GRC professionals specifically, this is less a threat than it is the opportunity they’ve been waiting for.

The practitioners who’ve made this shift describe it less like learning something new and more like getting permission to do what they were trained to do.

Their job became telling the agents what matters: setting the right risk appetite, deciding which controls are genuinely protecting something and which ones exist because they always have, knowing when an automated finding is a real problem and when it’s noise, and translating business context into compliance logic in ways no agent can replicate, because that translation requires judgment built from years of experience.

That judgment has been sitting in GRC teams all along, waiting for the operational load to lift.

The organizations that move first on this won’t win because their teams are better at AI. They’ll win because their GRC teams finally have the time and the mandate to do what compliance was supposed to do: think clearly about risk, act on what actually matters, and stop managing a program and start leading one.

Why Letting Go Feels Like Losing

The reluctance that comes up in these conversations makes more sense when you frame it this way.

Practitioners aren’t afraid of losing their value; they’re afraid of losing the operations that became their identity, even though those operations were never what they wanted. Letting that go feels like losing something, which makes it hard to see what’s waiting on the other side. And what is waiting is far more aligned with why they got into this work in the first place.

The shift, when it happens, is less a transformation than a return to what the role was always supposed to be.

Learn more about agentic GRC with Anecdotes at anecdotes.ai

Sponsored and written by Anecdotes.

The class of 2025 graduated into the worst entry-level job market in five years. Now a growing number of them are using AI tools during live job interviews, and a cottage industry of startups is rushing to sell them the means to do it. Whether that constitutes cheating or common sense depends on which side of the hiring table you sit on, but the numbers behind the trend are not in dispute.

Unemployment among recent college graduates aged 22 to 27 climbed to 5.7 per cent by the end of 2025, according to the Federal Reserve Bank of New York, well above the 4.2 per cent national rate. Underemployment, which measures graduates working in jobs that do not require a degree, hit 42.5 per cent, its highest level since 2020. The tech sector, once the default destination for ambitious graduates, shed roughly 245,000 jobs in 2025, according to tracking data from Layoffs.fyi and TrueUp. Another 59,000 have gone in the first three months of 2026.

The graduates who entered this market did so having watched an older cohort get hired, promoted, and then laid off at companies like Meta, Amazon, and Google in the space of 18 months. The lesson they drew was not subtle: competence and loyalty are insufficient protection. And so they arrived armed with a technology that their universities had spent four years telling them to learn.

The tools and the companies selling them

The phenomenon surfaced this week in a press release from LockedIn AI, a startup that sells a product called DUO: a service that combines real-time AI transcription of interview questions with a live human coach who can see the candidate’s screen and provide strategic guidance during the conversation. The press release, distributed via GlobeNewswire, was framed as a trend piece about generational resilience. It was, more precisely, a product advertisement.

LockedIn AI is not alone. Its founder, Kagehiro Mitsuyami, also co-founded Final Round AI, a similar product. Both companies have faced questions about the authenticity of their marketing: reviews on Trustpilot appear to be AI-generated, and independent reviewers have noted that the software can be visible to interviewers when candidates switch between windows. A Gartner survey of 3,000 job seekers found that six per cent admitted to interview fraud, including having someone else impersonate them. Fifty-nine per cent of hiring managers suspect candidates of using AI to misrepresent themselves.

The market for these tools is growing precisely because the conditions that created them are getting worse, not better. The National Association of Colleges and Employers found that 45 per cent of employers characterised the job market for the class of 2026 as “fair,” down from “good” the previous year. Hiring projections for new graduates are essentially flat, at 1.6 per cent growth. For candidates submitting dozens of applications and receiving interview invitations at rates below two per cent, the temptation to use every available advantage is considerable.

The hypocrisy argument

The most effective argument in favour of AI-assisted interviewing is not about fairness in the abstract. It is about a specific inconsistency in how technology companies treat AI.

Google’s chief executive, Sundar Pichai, disclosed during an April 2025 earnings call that more than 30 per cent of the company’s new code is now generated with AI assistance, up from 25 per cent six months earlier. Amazon, Microsoft, and Meta all encourage their engineers to use AI coding tools daily. Applicant tracking systems powered by AI screen and reject resumes before a human ever reads them. The hiring pipeline is automated from end to end, except on the candidate’s side.

For graduates who spent their university years being told that AI fluency would define their careers, being asked to pretend the technology does not exist during a 45-minute interview feels less like a test of competence and more like a test of compliance. The companies asking them to do so are, in many cases, the same ones that will expect them to use AI tools from their first day on the job.

This argument has real force, but it also has limits. There is a difference between using AI to write code more efficiently and using AI to answer questions about your own experience, judgment, and problem-solving ability. An interview is, at least in theory, a conversation designed to evaluate what a candidate knows and how they think. Outsourcing those answers to a language model, or to a human coach whispering through an earpiece, undermines the purpose of the exercise regardless of how unfair the exercise may be.

The employer response

Companies are already adapting. In-person interview rounds rose from 24 per cent in 2022 to 38 per cent in 2025, according to hiring industry data. Seventy-two per cent of recruiting leaders now conduct at least one in-person stage specifically to combat AI-assisted fraud. Some firms have moved to whiteboard exercises, pair programming sessions, and unstructured conversations that are harder to augment with real-time tools.

The deeper question is whether the interview itself is the right mechanism for evaluating candidates in an AI-saturated labour market. If the goal is to assess what a candidate can produce with the tools they will actually use on the job, then banning those tools during the evaluation makes little sense. If the goal is to assess raw cognitive ability and domain knowledge, then AI assistance defeats the purpose entirely. Most interviews attempt to do both, which is why the current system satisfies no one.

What is clear is that the class of 2025 did not create this problem. They inherited a job market reshaped by pandemic-era overhiring, aggressive cost-cutting, and an AI revolution that is simultaneously creating and destroying opportunity at a pace that neither employers nor candidates have fully absorbed. Their decision to use AI in interviews is not rebellion. It is the predictable behaviour of rational actors in a system that has told them, repeatedly and in every other context, that AI is not optional. The fact that the system now objects to them taking that message seriously is, at minimum, worth examining.

You can now let Claude take control of your computer to perform tasks like sending you a file you left on your hard drive, the AI’s developer Anthropic announced Monday. For the feature to work, you just need to be on a qualifying subscription plan.

In the wake of the viral explosion of the open-source OpenClaw framework earlier this year, Anthropic is the latest developer to deliver a tool that enables an AI model to act more independently. 

OpenClaw has spawned an entire ecosystem of “claws,” or AI tools that can take simple commands and perform them somewhat autonomously on your computer or with your tools or systems. Nvidia last week debuted NemoClaw, its framework for easily setting up and installing OpenClaw, with some security settings. 

Anthropic says that Claude will look for the right tools to complete the task at hand via connectors with apps like Google Calendar or Slack. If the tool or connector isn’t available, Claude can manually perform the task by typing or moving the cursor, as if it were using the keyboard and mouse. It can use programs like your web browser, dev tools and open files. 

When it’s performing these tasks, it can use a computer as you normally would — by scrolling and clicking around. The only difference is that Claude will always ask for permission beforehand. You can stop Claude from performing a task at any time.

Giving your chatbot the keys to your computer can be convenient for certain tasks, but it can leave your computer vulnerable to attacks. Experts told us one major worry with agentic AI is that it can take major, sometimes dramatic actions quickly and with little warning. Claws can also be hijacked by malicious actors, who can use your personal data and systems in ways you don’t want. 

Anthropic says it implemented safeguards to minimize risks like prompt injections. The system will automatically scan for this and more vulnerabilities as they are implemented.  

Despite some of its efforts to keep Claude’s computer use safe, Anthropic also provides a warning to users. The feature is new and may contain errors, and the company suggests not using apps that handle sensitive data —  so much so that some of these apps are disabled by default.

The research preview is available now for Claude Pro and Claude Max subscribers and limited to computers running MacOS.

Anthropic says the new computer-use feature works well with Dispatch, which allows you to assign tasks to Claude using your phone. Such tasks include checking your email every morning or opening up a Claude Cowork or Claude Code session. 

The combination of computer use for Claude and Dispatch enables you to do even more while you’re not even around. Anthropic says the feature combo can create a morning briefing or run tests, for example. 

Given that both features are new, some complex tasks might not work the first time. Anthropic said it’s releasing this research preview to gain early insight on where it needs the most attention to become an even more powerful tool. 

I also tested movies using the Mubi app because I’m a fuddy-duddy film fanatic, as well as a screener for Marty Supreme. Here’s where the upscaling, matte display, and AI truly shine: Marty Supreme is like an old ’70s flick, and The Frame Pro made me think I was in a movie theater. The matte display gave the Mubi films a more cinematic look.

On an Xbox Series X, I watched the entire Predator: Badlands film using the 4K Blu-ray version. It’s worth noting that though the colors look great, Samsung models don’t support Dolby Vision HDR, instead going for standard HDR10+.

For a more pure movie experience that disables any optimizations or AI enhancements, I tested the Filmmaker picture setting. In one scene, you can see the individual scales of the main character, which is awesome.

The Frame Pro has plenty of built-in cloud gaming features beyond any consoles you connect locally. I tested Cyberpunk 2077 using the Steam app and also fired up Senua’s Saga: Hellblade II on the Xbox app with no issues. One slight glitch is that, when I connected my Xbox controller to the TV it worked fine, but then I had to pair it again to the Xbox Series X later. It gave me a reason to buy an extra controller. When I tried a Sony Dualsense Wireless Controller with Cyberpunk 2077, it worked flawlessly. Forza Horizon 5 on my Xbox looked ultra-smooth, showing a Ford Bronco sliding around in the mud in a realistic way.

The Frame Pro supports up to a 144-Hz refresh rate from either the Connect One box or a Micro HDMI port, which makes it awesome for really smooth gaming. I tested Crimson Desert on an Acer Nitro 60 gaming desktop, and the colors, scenery, and overall clarity over the higher refresh rate looked stunning. If you have a stylish living room where you also want to game, this isn’t a bad choice.

During a chaotic battle, with snowy mountains off in the distance, I felt like I had jumped into one of the paintings. In the end, The Frame Pro is a fantastic TV for art—and just about everything else, as long as you don’t want the best backlighting around.