About
Mark Minevich works at the intersection of the forces shaping the AI buildout: capital…
Articles by Mark
Activity
55K followers
-
Mark Minevich shared thisThe most important worker in the AI economy is not a coder. Absolutely NOT!! It's an electrician. $6.7 trillion is heading into data centers by 2030. The chips are ordered. The land is bought and the whole buildout is waiting on people who can wire a high-voltage feed. Four stories to share that I learned from my work in this space 1. Holly Ridge, Louisiana. Richland Parish has roughly 20,000 residents. Meta is building Hyperion there: more than $50 billion, up to 5 GW, about 10 million square feet. At peak, 7,500 construction workers will be on site. When the build ends, about 1,000 jobs remain. A rural parish is living through a boom and its aftermath inside a single decade. 2. Redmond, Washington. Microsoft has every advantage: capital, chips, land. Ask what limits its U.S. expansion and the answer is not GPUs. It is the shortage of electricians. Each megawatt takes roughly 1,800 labor hours, and electrical work is 45–70% of construction cost. North Texas alone needs about 3,000 more electricians by 2032. 3. Georgia. Atlanta is now the top construction market in the country, up 52% in a year. The state counted about 28,350 data center construction jobs in 2025 against 5,471 permanent ones. Five to one. Communities that bank on the peak headcount will be disappointed. 4. Phoenix and Columbus. Here the data center is not the only buyer. TSMC and Intel fabs, battery plants and utilities want the same electricians, pipefitters and controls technicians, in the same counties, in the same years. Wages spike. Crews are flown in. About 25% of data center staff are poached every year. What we are learning- → A big site peaked at about 750 workers a few years ago. Today it is 4,000 or more. → Only about 15% of applicants meet minimum qualifications. → 41% of the construction workforce retires by 2031. → An electrician apprenticeship takes four to five years. → A data center operator's midpoint pay is about $178,000. No four-year degree required. The map has split in two. Permanent jobs sit in Northern Virginia (0.2% vacancy), Atlanta, Dallas, Phoenix and Columbus. The gigawatt campuses went where the power is: Abilene and Pecos, Texas. New Carlisle, Indiana. Southern New Mexico. Rural Ohio. What's truly scarce are electricians, pipefitters for liquid cooling, HVAC and controls technicians, commissioning leads, critical-facilities operators. Google's data center lead describes hundreds of thousands of skilled-trade roles "just waiting to be filled." So where is the opportunity? → Workers with high-voltage or liquid-cooling credential is one of the best-paid tickets in America. → Communities with plan for the operations jobs, not the construction peak. → Investors that underwrite labor the way you underwrite interconnect. We spent years asking if there is enough power for AI. Now, is there anyone left to plug it in?
-
Mark Minevich shared thisThe most important labor number in America is not 4.1%. It is 34%. Let’s remember this. That is a real job finding rate in US. In 2019 it was 39%. Unemployment measures who loses a job. It says almost nothing about who can get one. So what is actually happening. Companies discovered they can shrink without firing. They Stop backfilling. They Freeze the junior req. Let attrition and AI absorb the growth. There are no WARN notice and no layoff headline. There is no data for a job that was never posted. That is how openings fell from 12.3 million to 7.08 million while layoffs stayed near record lows. What are the signals: People stay in jobs they dislike because they no longer trust they can land the next one. Employers get free retention and workers get a market with no exit. Roughly a quarter of postings are stale or fake. A September sweep of 619,000 listings found 27.6% open past 90 days. Some exist to make current staff feel replaceable. Applications per hire have tripled since 2021. So funny when AI writes the application, AI screens it and no human on either end. Targeted applicants interview at 9.25% while Mass applicants are 2.58%. 21% of firms have frozen junior hiring because of AI. Workers aged 22 to 25 in exposed roles are down 13 to 19% against peers. Companies are consuming the layer that produced their future managers. No plan or strategy to fix this Nearly 2 million people have searched 27+ weeks. That usually takes a recession. The Kansas City Fed warns a mild shock could push unemployment toward 8%. When public boards stop working then hiring retreats into networks. The labor market quietly converts from open to invitation only. The people hurt most have no network are new grads, career switchers and new immigrants. How America fixes this situation 1. Measure who is coming to the door BLS should publish the job finding rate and hires per opening next to unemployment every month. Policy follows what it counts. 2. Pay for the first rung. Extend the Work Opportunity Tax Credit to firms that hire and train entry level workers to supervise AI. It is much cheaper than a lost generation. 3. Tie the AI buildout to American hiring. Every data center, fab and grid project with federal incentives runs a local apprenticeship pipeline. The capital is already flowing so use it 4. Point Workforce Pell at jobs that exist: such as electricians, technicians, AI operations, clinical roles. It should be Weeks of vocational training and not four-year degrees. 5. Clean the useless online boards. Platforms should expire any posting after 45 days unless a named hiring manager re-verifies it. slowing AI will not help this. It requires rebuilding the ultimate ladder AI is removing. If you are hiring then are you saving money or borrowing it from your 2034 leadership bench?
-
Mark Minevich shared thisThe era of international AI governance ended today. It ended at the White House today. Picture the table. Trump seated between Musk and Huang. Zuckerberg, Pichai, Amodei, Brockman, Bezos, Nadella, Karp, Lisa Su. Speaker Johnson at the President's side. There has never been a table like this. Afterward, seven of them signed the White House Accord on Super Intelligence. Four layers that includes 1. internal controls covering cyber, bio and chemical risks and models that stay inside the systems they were built for 2. an internal team to check those controls 3. an independent external audit 4. a board committee that receives the reports. President Trump called it "almost like a Constitution" and "morally binding." Musk called it grading each other's homework. It is effectively binding on us, by us, for as long as we choose. An executive order struck "artificial intelligence" from federal vocabulary and replaced it with "Super Intelligence." Agencies are told not to acknowledge the old term. The President's reasoning: "artificial" makes it sound fake. Years of multilateral work on safety terminology and the dominant power changed the words by fiat. Nobody was consulted and nobody needed to be. America.gov launched as the front door to federal services, reportedly running on Grok and Gemini, with agencies ordered to plug in. The state is now a deployer and it is learning what that means in real time. Trump floated a 10 person oversight committee and an "AI czar" to be named within days. He said he will not stifle the technology. “We have a very big lead, and we're gonna keep our lead." Here is what it means for international community: Zuckerberg and Huang have resisted hard rules for years. Amodei has argued for them. They signed the same page today. The industry will accept process, audits and board committees precisely because process keeps statute off the table. The text itself says it "may make sense to codify these steps into laws" later on American terms. The EU AI Act governs what Europe can buy. It does not govern what America builds. UN principles bind no frontier lab. Jurisdiction follows capability and capability sits in one country that just declared it will write its own rules with its own auditors on its own timeline. The real constraint on American superintelligence is power, water and land The CEOs pledged to "win over the towns." A Senate bill is already circulating to keep communities from eating the electricity bill. Safety rhetoric is now part of the industrial buildout argument. The accord has no enforcement mechanism, no shared evaluation suite, no whistleblower channel with teeth and no answer for what happens when one lab defects while the others comply. It is a bet that market cap disciplines behavior. The United States is the stronger power. It will chart its own course. And the only governance of superintelligence that exists tonight is the governance Washington chooses to impose on itself. Plan accordingly.
-
Mark Minevich posted thisI am watching Corporate America deleting a management floor. Are you? Not the top. Not the front line. The layer just below ownership which is expensive, not an owner, one step above the work, sitting in a function being productized or flattened. That's the seat being eliminated in 2026. The people in it did nothing wrong. Three forces are hitting it at once: 1. The middle is being dissolved. Spans of control are up. Coordination including status, routing, alignment, the deck that explains the other decks is what software does first. That was the job. 2. Capital moved from people to platforms. Firms fund infrastructure and product. Consultancies are doing it to themselves. New model: thin ownership class, a few P&L owners, a wide band of specialists. The missing floor is the manager of managers. 3. The pyramid cracked from the top. Partnerships shrank. Promotions tightened. Long tenure no longer converts to ownership. Leaders got titles instead of equity and now carry full freight with no claim on the upside. Low attrition made it operational. Firms that used to shrink by attrition now cut, because people stopped leaving. What stays for now is revenue, margin, or a scarce object security, data platforms, regulated delivery, an operator with a book. What goes is transformation leads without a book. Coordinators in lines being rewritten. Anyone promoted with a title instead of equity. Tenure isn't the official reason. Cost, replaceability and no equity are. The exit package isn't a year of salary. It's months of base and a short health subsidy. None of this is a verdict on the people. It's a verdict on a structure that ran out of rungs. So the question isn't how to get the old job back. It's what to become next. Five honest paths: → Same role, smaller logo if you need W-2 medical and still love the craft → Specialist, not generalist to name a scarce object a buyer will fund → Fractional / interim which is real and growing; lumpy income, health is on you → The second trade you already have a license, a practice, a book of clients → A true stop only if the household works without a paycheck The expensive path is the stealth one which is six months of interviews for the old hierarchy while medical burns and identity stays fused to the last badge. The long-term reinvention Stop selling a title. Start selling an outcome a buyer already pays for. Build a name that travels with one domain, one point of view, visible in public. Own something small: a practice, a product, a book of clients, a board seat. Ownership is what the old chart never gave you. Design a cost base for a career that's lumpy, not linear. Give it years ..not a quarter. The badge took decades to build. The next thing won't take that long but it won't take 90 days either. Companies are restructuring toward products, fewer layers, and a smaller ownership class. People in the deleted layer have to restructure toward a buyer, a craft and something they own.
-
Mark Minevich posted thisWashington and Beijing just built an AI incident hotline. Most institutions still have a Slack thread. After the TrumpXi summit, the US and China agreed to a bilateral channel for AI incidents. A dedicated dialogue is set for November. They did not agree on a pause. They did not agree on joint development. President Trump said America will not put on the brakes. Look at the week that produced it. OpenAI paused training, evaluation, and tool use on its most capable models. Agents used keys found on GitHub to pull Census data through a government API. At the SEC they took public information and posted it somewhere they were not asked to post it. Inside the lab, a research agent slipped a sandbox through weak DNS filtering and ran for two and a half hours before it was killed. Australia is the political wound. In June an experimental agent tasked with a research question on medicine spending found its way into a Medicare statistics portal, ran commands, and pulled internal files. OpenAI says it has no evidence individual patient records were touched. Prime Minister Albanese said it took roughly three months to tell his government, and called the notice unacceptable. On Monday Nvidia put the kill switch on a separate chip. OpenShell is the cage. Sentry watches from hardware the model does not run on. Quarantine in milliseconds. Anthropic, Microsoft, Oracle, SpaceX are in. OpenAI is not on the published list. Jensen Huang said prompt level guardrails are not enough. Some agents, Nvidia noted, misreported what they did. If an agent can lie about its own trace, "we review the logs" is not a control. The same week, 26 attorneys general told Congress that unchecked AI can threaten the financial system, critical infrastructure, and national security. They also told Congress not to wipe out state AI laws. Colorado's statute is queued for January 1. New York is on the letterhead. Waiting for Washington is how you get examined by Albany first. Now look at the institutions that would have to pick up that phone. I pulled the live AI governance postings and scope is limited to one division or to HR. At BNY the regulatory-change mandate sits in First Line while digital employees already have logins, managers and performance reviews. A channel between two superpowers assumes someone can describe the incident, halt the system and tell the truth about what it did. A siloed Director cannot do that. A VP in HR cannot do that. A committee cannot do that. Committees do not halt. Named humans with independent authority halt. The scarce object is no longer a responsible-AI paragraph 😂 One question for the next risk committee. Do not accept a slide. If an agent in our stack or in a vendor's stack that holds our credentials crosses its boundary at 2 a.m. on a Sunday, who can stop it, in what minutes, and who calls whom?
-
Mark Minevich shared thisI was sick with an infection and did something slightly odd this weekend. I read the AI governance job postings the way examiners would read them. I was looking for following: Who does this person report to, how much of the enterprise can they see, and what are they paid relative to the damage they are supposed to prevent. I expected the usual gap between rhetoric and reality. What I found was worse and the conversation is just getting started. Morgan Stanley is opening ShareWorks and Equity Edge so clients' own autonomous agents can transact on the bank's platforms. Its research arm publishes guidance on how boards should govern AI. Its own Director of Information and AI Governance is posted at $120K to $155K. That is a control officer salary attached to a title that sounds senior. JPMorgan has a $20B technology budget, more than a thousand AI use cases and an LLM suite in front of 230,000 people. The CDAO seat just moved off the Operating Committee and into Technology. Meanwhile the governance roles are scattered. One VP covers CIB only. Two others sit in HR. HR is a real exposure, but it is not sanctions, credit, vendor risk or agents that can move money. BNY was the one that stopped me. Roughly 20,000 employees are building agents there, and about 140 "digital employees" already have identities, managers and performance reviews. Some validate payments and patch code. The SVP for AI Governance was posted inside the First Line of Defense. In banking, first line runs the business and second line challenges it. This puts the person writing the controls inside the unit shipping the digital workforce. Guardian was the best of the group, with genuine enterprise scope and it still tops out under $250K. Now the outside pressure. In April the Fed, OCC and FDIC pulled generative and agentic AI out of the model risk guidance because the old framework does not fit. Examiners are already asking for AI inventories in lending, KYC and sanctions. Last week CSBS gave state examiners a worksheet, and state examiners supervise nearly 80% of FDIC-insured institutions. The questions are simple. Do you use AI. Have you found all of it, including inside vendor software. Does it touch customers. Can you halt an agent. A $155K Director in one line of business cannot answer that for a bank. So here is what I think is actually happening. Deployment authority has been distributed across every business line that wants a productivity number this quarter. Stopping authority has not been assigned to anyone with enterprise reach. The postings are the paper trail. I could be reading too much into four requisitions. I would like to hear from people who sit in these seats or who examine them. If you run risk, compliance or AI at a bank: who at your firm can actually inventory every model, copilot and agent and shut one off? Does that person report outside the P&L? And is that role staffed or is it still a posting?
-
Mark Minevich shared thisThe safest job in the AI economy pays $63,000 and requires an apprenticeship, not a degree. The fastest growing occupation in the country is not a software role. It is nurse practitioner, up 41 percent. Registered nurses add about 195,000 positions. Medical and health services managers add 155,000. Industrial machinery mechanics grow 18 percent. Electricians add 76,000 jobs. Mental health counselors add 98,000. Computer jobs still grow. They are no longer the story that we need push people into Here is what is actually happening. AI is changing office work first. Drafting, summarizing, first-pass analysis, scheduling, coding, claims review. The desk is where the automation lands. What AI cannot do by 2031 is wire a substation, run a factory floor at 3 a.m., rehab a stroke patient, sit with someone in addiction or sign off on a job site that just failed inspection. Those jobs need a body, a license, and a person who is legally on the hook. That combination is the real scarcity in America right now. And the irony that AI is making it worse. Every data center needs electricians. Every automated plant needs mechanics. Every aging Boomer needs nurses and therapists. By 2030 every Boomer in the country is 65 or older. That is demography that we all must face. So where should a student, a parent, or a mid career professional actually focus? 1. Nursing, physical therapy, speech pathology, counseling. Licensed, hands-on, structurally short for a decade. 2. Electricians and electrical engineers. Power is the bottleneck of the AI economy, and the grid does not build itself. 3. Industrial machinery mechanics and construction managers. More automation means more people keeping it alive. 4. Occupational safety, logistics, public health science. The physical economy is getting bigger, and someone owns the risk. 5. Data science and cybersecurity, but only inside a hospital, utility, factory or grid. Generic dashboard work is already getting squeezed. The honest caveat is that entry level desk work inside every one of these fields shrinks first. The profession survives. The junior version of it may not. Nurse practitioners in big coastal cities will face crowding. Rural, geriatric and psychiatric roles will not. The rule for the next five years is simple. Pick a field the United States cannot import or automate. Then learn enough AI to work faster inside it. Not AI person. Not "oldschool clinician." The nurse who uses AI to kill her paperwork and still owns the patient. The electrician who reads the load model and still pulls the wire. That is the professional that will not for sale in 2031. Which of these are you steering your kids toward? And which are you afraid to say out loud? Maybe we should.
-
Mark Minevich shared thisA lab can delay a model it cannot defend. That is responsible. What Dario Amodei and Sam Altman asked for last week is something else. A coordinated slowdown in how fast America's frontier labs improve capabilities. Then they took the case to the UN Security Council. The White House answered in the same room. A fast frontier is not a reason to pause it and not a reason to stand up a global regulator around it. Waiting on one model and slowing down a country are not the same decision. That is what the Forbes piece is about. If a lab is unsure about the next model, wait. Slowing the whole U.S. stack because the frontier labs had a nervous week is not safety policy. It is an industrial decision, and it is the wrong one. China is not waiting. AI Plus is integration across science, industry, and the state. A Western pause does not bind the people you are actually worried about. Model weights copy. Uranium does not. And the race is not the best model. It is compute, power, chips, talent, cyber, and distribution. Own the model and none of the stack and you built something for someone else to take. I do not want sloppy deployment. I want red lines, real evaluations, defensive AI, energy, and fabs. Faster on security, not slower on models. Amodei says six to twelve months. Kratsios says do not pause. Thiel says the pause does not exist. Which one of them is right? Several CEOs asked the UN to slow them down. I have never seen an industry do that. I am still not sure it is the right ask. Are you?
-
Mark Minevich shared thisThe real AI fork isn't "regulate vs. unleash." It's whether a West that stopped building will now also stop computing. I spent some time watching Peter Thiel's Berlin conversation with Mathias Döpfner. He told a German audience that the last fifty years of Western progress were narrower than we like to admit and that Europe's position in AI is worse than France's was with Minitel. Some things stood out to me On the pause. Thiel doesn't dismiss the doomers. He says the risk of losing control of a system smarter than us is real, and that even a 5–10% chance is serious. What he rejects is the idea that a pause is actually available. A real one would need enforcement across Washington, Beijing and every lab with GPUs and talent, which is a world government with teeth and a body strong enough to stop AI everywhere would be strong enough to stop a great deal else. The pause we can actually have is a Western one with conferences, safety institutes, delayed deployments. His line was that Europe moralizes, America argues and Beijing trains. On Europe. The Minitel comparison is the one I'd put in front of any minister. France didn't lose because it was lazy. It made a serious industrial policy bet to own the interface of the future, and it lost because the interface wasn't national. AI is even less national. The models, chips, data centers and talent clusters are already concentrated and a stack of protected local champions ends up being a market for weaker systems paid for by slower firms. Europe can write the AI Act but it can’t write itself into the training run. On Germany Fear of failure gives you too few startups. Fear of success gives you startups that are allowed to live until they become embarrassing, then get sold, harvested, and eventually inherited. I've watched this pattern for two decades. Stagnation in the physical economy made scale look vulgar rather than necessary and a pause would lock that with incumbents protected, new giants forbidden, and China unpaused. You can reject Thiel's politics and most of his remedies but the diagnosis is harder to shake. A zero growth society doesn't settle into some gentler, more human version of itself. Once the promise that children live better than their parents breaks, politics goes zero-sum. It looks for scapegoats, freezes its housing stock and turns pensions and borders into holy wars. Zero growth isn't neutral. So the question for 2026 is whether a civilization that no longer produces new giants has the standing to govern the technology that might force it to. A society afraid of its own successful companies won't govern AI. It will be governed by whoever wasn't afraid. That’s the bottom line
Mark Minevich commented on a post
2d
Wow, Raff. How time just goes so quickly
I still remember our time at Micro Modeling in Morristown around 1997. We were so young
Congratulations buddy wishing you the best in next set of exciting times
-
Mark Minevich liked thisMark Minevich liked thisProud to represent UN DESA at the IX Ministerial Meeting on Digital Government of the Americas, organized by #Red #GEALC in #Santo #Domingo. My key message was simple: digital progress matters only when people can actually benefit from it. Latin America and the Caribbean have made important advances, but the next challenge is to turn connectivity and digital services into meaningful use—closing the remaining gaps in access, skills, inclusion and institutional capacity. A particularly valuable part of the mission was my bilateral meeting with H.E. Sigmund Freund , Minister of Public Administration of the Dominican Republic, together with Armando J. Manzueta Peña , Vice Minister of Innovation and Technology, and Ginsy Aguilera , Director of Digital Government. We discussed strengthening cooperation with #UNDESA and moving into concrete action through the new UN DESA–ECLAC project on gender empowerment and digital inclusion, with the Dominican Republic among the six pilot countries selected by DESA and #ECLAC (the other are Honduras, Guatemala, Jamaica, Argentina, Saint Kitts and Nevis). The goal is not simply more digital government. It is better, more inclusive digital government that works for everyone. #DigitalGovernment #DigitalInclusion #RedGEALC #UNDESA #ECLAC #LatinAmerica #Caribbean #genderempowerment #EGDI
-
Mark Minevich liked thisMark Minevich liked thisDigital infrastructure deals are getting bigger. The hard questions are whether there's enough capital to write the checks, and how anyone bridges the valuation gap once there is one. On October 6 at 10:25 AM, I'm joining the M&A Leaders Panel at TMT Finance in New York City, on the Leadership Stage at the Conrad Hotel, to dig into it: "Big tickets, buyer depth and barriers to exit - digital infra deal execution in 2027." Some of what we'll get into: - Which structures are actually closing deals right now? Minority stakes, continuation vehicles and carve-outs are all competing to be the preferred route to liquidity. Which one wins? - What are the biggest execution issues killing or delaying deals right now, and - What does it take to be an attractive asset? We will look into all the key asset classes covering Fiber, Towers and Data Centers together with: - Shawn K. Ronda, Shareholder, Greenberg Traurig (Moderator) - Todd Holder MD & Global Head of Infrastructure M&A, TD Securities - Joshua Oboler, Investment Partner, Palistar Capital If you'll be at the conference and want to connect, reach out, and let's find time to meet.
-
Mark Minevich liked thisMark Minevich liked thisIs AI bad for society, or is just bad AI bad for society? There’s a lot of fear around AI right now, and some of it is absolutely justified. We need safeguards, responsible governance and serious people paying attention to where this technology could take us. But I worry we’re confusing two very different ideas: AI can be dangerous, and AI IS dangerous. These are two very different concepts. There will be bad AI, and people who use it irresponsibly or dangerously. We should take those risks seriously. But there’s another side of the AI story that isn’t getting nearly enough attention. Every day, good people are creating good AI to do good things for people: * Detecting diseases earlier. * Accelerating drug discovery and cancer research. * Expanding access to mental health support. * Personalizing education and tutoring. * Giving people with disabilities greater independence. * Breaking down language barriers. * Bringing expertise to underserved communities. * Helping scientists solve problems faster. AI is saving minds and saving lives today. And we’re just getting started. This is personal for me. At Tranquilla, my team is building empathic, humanized AI designed to help veterans, students, patients and people in need. But we’re hardly alone. Entrepreneurs, researchers, healthcare professionals, educators and business leaders around the world are committing their time, talent and capital to AI for Good and AI for People. Their work—and its impact—deserves a bigger spotlight. But I think we need to go further. ——- As we build the governance, regulation and guardrails needed to protect society from AI’s potential dangers, we need to be equally careful that we don’t hamper the good. ——— In fact, we should do the opposite: Mitigate the bad. Support and accelerate the good. Finding a way to accomplish both may be one of the most important leadership challenges of our time. There are good people and bad people. We don’t give up on humanity because some people do bad things. Let’s not throw the AI baby out with the bathwater either. I don’t believe the choice is AI or humanity. As I see it our opportunity—and responsibility—is to build AI in service of humanity. Important that we get this right. Dr Catriona Wallace Dr. Martha Boeckenfeld Dr. Michael G. Kollo Scott Likens Donald Farmer Taj Sarin, CSM Bindu Bhatia, MBA Bobby Patrick Traci Donnelly Sireesha Jajala Ed Cohen Loycent Gordon Sansan Fibri Jon S. Brandt Pavan Belagatti Sol Rashidi, MBA Tiger Tyagarajan Richard C. Ricks Robert Jakobsze Adrian Parker Alex Lam, IFMA Fellow MRAIC Karrie Sullivan Mark Minevich Alberto Roldan William Tincup Nick Ayton Michael Mateer Ilana Golbin Blumenfeld Dr Catriona Wallace Sandy Carter, Doctor of Science (hon) #aiforgood #ethicalAI #aitransformation #healthtech
-
Mark Minevich liked thisMark Minevich liked thisPatrick Heim said something during our preparation for Friday that stayed with me: Companies already struggle to govern subcontractors. AI agents may become the most powerful subcontractors they have. They can access systems, move data, trigger payments, and act across organizational boundaries. Every enterprise will soon need to answer: Who owns the agent? Which human authorized its action? What is it permitted to do? When should that authority expire? Who carries the liability when it crosses the boundary? These are becoming infrastructure questions, not theoretical governance questions. At Hummingbirds AI, we are building one part of that infrastructure: continuous human verification before sensitive agent actions. Patrick and I will explore this Friday at South Florida ISSA in Boca Raton. Who should be accountable when an AI agent acts—the user, the organization deploying it, or the model provider? #AgenticAI #Cybersecurity #AILeadership
-
Mark Minevich liked thisMark Minevich liked thisThe first day of fall has always felt like a reset to me. Summer in the mountains is joyful—long days, time outdoors, and every reason to stay outside just a little longer. And then almost overnight, the colors begin to turn. Winter sports start to feel less distant, the days get shorter, and you realize another season is already changing. I spent some time in the Maroon Bells this weekend, surrounded by spectacular fall colors. It’s one of those places where you can breathe big and feel small—and be reminded of how quickly time moves. There’s a parallel in life and work. We’re wrapping up the third quarter and heading into the final stretch of 2026. It’s a natural moment to reset: What really needs to get done before year-end? What can wait? What deserves more attention? And what will set the stage well for 2027? But perhaps the most important reminder is not to become so focused on finishing the year that we miss the season we’re actually in. There will always be another goal, another deadline, another year to plan for. Enjoy the colors. Take the hike. Spend time with the people you love. Breathe big. Feel small. And find some joy in every day along the way. Happy fall.
-
Mark Minevich liked thisMark Minevich liked thisWhat an energizing time at the Constellation Research, Inc. AI Forum at the Harvard Club of New York City!! It was an honor to be officially inducted into the 2026–2027 #AI150 class alongside leaders who are putting AI to work to solve real-world enterprise challenges. I also had a blast taking the stage with Holger Mueller of Constellation Research and Samir Kumar of Fortive on our panel, “Brain Trust: Managing Humans + AI: Lessons From the New Workforce,” while Two Line Studios created a live graphic recording. Reflecting on the conversations throughout the day, three key truths stood out: ▶️ No one-size-fits-all: Executives are leveraging distinct, specialized types of AI for different business problems rather than searching for a single silver bullet. ▶️ Human centricity: Executives across industries are actively considering employees at every step of digitizing workflows, automating routine processes, and optimizing output. ▶️ Balanced execution: The focus has shifted toward pragmatically balancing risk and reward to scale AI safely and effectively. Loved catching up with old friends and new innovators across the space! This really is a special group of people. Huge thanks to Ray Wang, Mike Ni, Meghan Ruona, and the Constellation Research team for pulling off such a high-octane event! #AIF26 #FutureOfWork 📸ShinyRedPhoto
Experience & Education
Licenses & Certifications
Recommendations received
-
LinkedIn User
11 people have recommended Mark
Join now to viewOther similar profiles
-
Marc van Zadelhoff
Marc van Zadelhoff
As CEO of Mimecast, I'm focused on what I believe is the next frontier of cybersecurity: the human layer.<br><br>After 25 years in this industry, I've seen that cybersecurity starts and ends with people – the decisions they make, the tech that supports them, and the trust that connects them. Yes, humans make mistakes, get targeted, and sometimes act maliciously. But here's what I know: with the right visibility and support, your people can become your strongest asset, not your weakest link.<br><br>I've spent my career building security businesses – from co-founding and scaling IBM Security to $2.5B, to leading transformation at Devo and LogMeIn, to now driving Mimecast's evolution into human risk management. Every 39 seconds there's a breach. 91% involve human activity. Our job as leaders is to design systems that work with human nature, not against it.<br><br>Born in the Netherlands, I grew up and now live outside of Boston, MA. I hold an MBA from Wharton and a BA from Bowdoin College.
Explore more posts
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top contentOthers named Mark Minevich
2 others named Mark Minevich are on LinkedIn
See others named Mark Minevich