Platform Land Grab
In February, the U.S. economy experienced an unexpected decline, with employers shedding 92,000 jobs and the unemployment rate rising to 4.4 percent. This downturn, which affected nearly all major sectors, reflects a broader weakness in the labor market. Economists noted that the report indicates a sluggish job market, with revisions suggesting that job growth has essentially stalled over the past three months.
Tech jobs in the U.S. are declining sharply, similar to the dot-com bust and 2008 recession.. Economist Joseph Politano states tech job losses are now surpassing previous downturns, with the sector shedding jobs rapidly, partly due to AI’s rise, which may have contributed to losses in computer system design.
Productivity growth has surged to 2.5%-3% since late 2023, up from 1%-2%, driven mainly by manufacturing and data center construction, not just AI.
OpenAI’s latest model, GPT-5.3 Instant, has successfully reduced hallucinations by up to 26.8% compared to its predecessor. The company conducted internal evaluations in critical areas such as medicine and finance, revealing a 19.7% increase in reliability when using internal knowledge and a 22.5% decrease in hallucinations during web searches.
Google Workspace has introduced a new Command Line Interface (CLI) that allows users to access Gmail, Docs, Sheets, and other apps directly, streamlining automation for developers and AI agents.
Anthropic has launched the Claude Marketplace, enabling enterprises to utilize Claude-powered tools from various partners like GitLab and Replit.
Microsoft revealed new AI features for SharePoint’s 25th anniversary, enhancing building and governance tools. Key updates include a natural language experience for creating plans for sites, pages, lists, and libraries, and custom AI skills to guide AI behavior.
Hexnode introduced an upgraded conversational interface for its UEM Console via Hexnode Genie AI, enabling IT admins to interact naturally for real-time device data, commands, and troubleshooting. Building on previous functions like script writing and documentation access.
This is a structural shift in liability allocation, not a technology story. The Claude Marketplace and Google Workspace CLI announcements are platform land-grab moves, not product launches. Anthropic is building a distribution moat through enterprise partner integrations before commoditization compresses model margins. Google is embedding AI agents into productivity infrastructure to make switching costs prohibitive. And this is all happening while the labor market is soft.
Who Owns Failure
The technology is the mechanism; the real movement is who owns the consequences when agentic AI acts incorrectly at scale. So let’s talk about the movement.
According to a study by Omdia, 88% of organizations are now using AI-driven remediation, with 44% implementing it for most exposure types. This shift comes as companies face expanding cyber threats and the need for faster responses; however, concerns remain regarding trust in AI decision-making, with 49% of security teams expressing skepticism.
A recent report by LinkedIn reveals that AI engineering is among the most sought-after skills in today’s job market, with job postings requiring AI literacy increasing by over 70% year-on-year. Despite two-thirds of executives expecting employees to actively pursue AI skills within six months, less than half of U.S. professionals feel adequately supported in their training efforts.
A recent study from Boston Consulting Group and the University of California, Riverside reveals that excessive use of artificial intelligence tools is leading to a new form of mental fatigue known as “AI brain fry” among workers. More than 14 percent reported experiencing cognitive exhaustion characterized by feelings of mental fog and difficulty focusing. Those suffering from AI brain fry are 39 percent more likely to make significant mistakes at work and show a higher intent to quit their jobs compared to their peers who do not experience this condition.
Anthropic developed an early-warning system for potential AI-related job losses in white-collar sectors. Their index shows limited evidence of increased joblessness but aims to track AI-driven disruptions. Tasks automatable by AI are already being implemented, especially in programming and customer service. Despite risks, unemployment hasn’t risen significantly in vulnerable jobs. Economists Massenkoff and McCrory introduce “observed exposure,” indicating AI hasn’t fully disrupted the labor market yet.
Organizations are increasingly recognizing the need to manage AI incidents as a distinct category. A recent survey by IBM indicated that 63% of organizations lack formal governance policies for AI, exposing them to significant operational risks. This year, businesses will need to implement specific remediation processes for AI incidents, emphasizing responsible AI adoption and proactive communication strategies.
The 88% AI remediation adoption figure, if taken at face value, suggests the market has moved. It hasn’t — it’s moved on paper. Deployment without governance is not adoption; it’s exposure accumulation. The 49% skepticism among security teams means the humans closest to the blast radius don’t trust the automation they’re nominally operating. That’s a liability configuration, not a security posture.
The “AI brain fry” finding from BCG/UC Riverside is easy to dismiss as soft research, but the operational consequence is concrete: workers experiencing cognitive exhaustion from AI tool overuse are 39% more likely to make significant errors. For MSPs whose technicians are being pushed to use AI copilots across ticketing, documentation, scripting, and remediation simultaneously, this is a compounding risk — automation-induced errors layered on top of human errors made by fatigued operators supervising that automation.
Skills Over Titles
A recent report from shows AI transforming skill needs in the IT sector, with executives claiming “every job is becoming an AI-enabled job.” Insights from Wharton and Accenture reveal a shift where job titles are less relevant and skills more important. The report notes a “signaling gap,” stressing skills like judgment, technical expertise, and real-world application. Companies like Accenture see rising demand in data science and AI infrastructure, marking a move toward execution skills. AI’s integration across work emphasizes adaptability and linking technology with business goals.
According to research from the Federal Reserve Bank of Dallas, younger workers in the tech and finance sectors are facing significant job losses due to AI, while older, more experienced workers are seeing their employment grow. Between 2023 and 2025, employment for younger workers dropped by 20%, while it increased by 12% for those aged 30 to 59.
In Channel Dive, Duane Barnes of RapidScale highlights how the software industry is shifting from traditional to usage- and token-based pricing to stay competitive. He notes current billing methods are rigid and burdensome, advocating for usage-based charges instead of fixed fees. Although some AI startups adopt outcome-based pricing, legacy vendors are slow to adapt, creating challenges and opportunities for channel partners in this evolving landscape.
Legacy per-seat licensing was predictable. Token-based consumption is not — and MSPs who pass through consumption costs without margin controls are building a business on variable-cost exposure with fixed-price contracts.
Why Do We Care?
Forty-nine percent of security teams don’t trust the AI remediation systems their organizations have already deployed. That’s from the same Omdia study claiming 88% adoption. You have a majority of organizations running automated remediation that their own security practitioners are skeptical of — and IBM confirms that 63% of those same organizations have no formal governance policy for AI incidents. That’s not a confidence gap you close with training. That’s a liability configuration that’s waiting for its first significant event.
Now layer in what Hexnode just shipped — natural language commands to UEM consoles, real-time device actions, troubleshooting execution. That’s the same category of agentic action. An IT admin types a natural language instruction, the system interprets it, and endpoints respond. The blast radius of a misinterpreted command at scale is not theoretical.
The BCG research on AI brain fry is the piece that connects these threads operationally: technicians who are cognitively fatigued from AI tool overuse are 39% more likely to make significant errors. So you have automated systems executing actions that humans don’t fully trust, governed by policies that mostly don’t exist, supervised by technicians who are increasingly fatigued by the volume of AI-mediated decisions they’re making. That’s a compounding failure stack, not a productivity story.
And the pricing shift that Duane Barnes flagged at RapidScale — token-based consumption replacing per-seat licensing — means the cost structure of that entire stack is becoming variable at exactly the moment the operational risk is increasing. MSPs with flat-fee contracts are absorbing both sides of that equation.
The bad decision is treating AI remediation as a feature to sell before you’ve solved it as a governance problem. The MSP that deploys first without the contract language and the incident playbook doesn’t win the market — they absorb the first high-profile automated failure and fund the case study that everyone else learns from.
What to Consider
- Audit all AI tools that can act autonomously. If you can’t identify who authorized actions or the rollback process, you’re without governance.
- Managed services agreements not excluding AI remediation liability need renegotiation before deploying agentic tools.
- AI platform costs must be passed with markup or capped; absorbing these costs under fixed fees harms margins as AI use grows.
- Address AI tool fatigue operationally by rotating technicians and measuring error rates, not as a wellness initiative but as quality control.
- Develop an AI incident response playbook now; 63% of organizations lack formal governance—your prospects.
If this trend continues, MSP contracts will split into two tiers within 24 months: “human-in-the-loop managed services” and “automation-first managed services,” and the automation-first tier will require audit logs, rollback guarantees, and explicit liability caps the same way cyber insurance forced baseline controls.