ROI Reality Check
A recent study shows only 4% of businesses see ROI from AI, with many executives blaming users for lack of training. Research by Economist Impact finds 38% have AI budgets, and just 16% offer internal training. Nearly all senior executives claim to promote AI skills.
A survey by Slingshot and Infragistics highlights a gap between executives and employees on AI’s workplace role. 86% of executives see AI as essential, but only 49% of middle managers agree, per HR Dive. About 40% of employers view AI as a team member, but just 20% of employees do, preferring to see it as a tool. Only 31% of employees rely on data often, compared to 70% of executives.
Governance Gap Widens
The reason this gap keeps widening is that AI is being funded around adoption speed while training, oversight, and accountability are treated as downstream problems. Leadership buys speed, operators inherit uncontrolled tools, and the bill arrives as security incidents, audits, and cleanup work.
A survey by Ernst & Young finds over half of department-level AI initiatives lack formal oversight, raising security and governance concerns. While 85% of tech leaders prioritize speed, 78% feel AI adoption outpaces their risk management.
Grammarly is facing criticism for using prominent experts’ identities without consent in its new “expert review” feature, and says it will continue unless individuals opt out. “Grammarly’s posture—‘we’ll use identities unless you opt out’—is a micro version of a bigger shift: institutions are normalizing risk transfer as the default. And once risk is deferred that way, the costs don’t disappear—they show up somewhere else.
Cleanup Economy Rises
Once governance is deferred, the costs reappear downstream as legal exposure, remediation spend, and public backlash.
A lawsuit against Eightfold AI raises concerns over AI in hiring. Two applicants filed a class action, claiming Eightfold functions like a consumer reporting agency but lacks the rights such designation entails, raising transparency and compliance issues. Reports show 88% of AI vendors limit liability, and only 17% ensure regulatory compliance, risking legal exposure.
Employment is where AI turns into lawsuits fast because you’re making decisions about people with systems you may not be able to explain. Eightfold is a reminder that the question is not whether you used AI—it’s whether you can prove oversight, notice, and accountability.
President Trump announced a new cyber strategy that expands the operational role of private firms in offensive cyber activity and national cyber posture. That sounds like a security headline, but structurally it is a liability headline: as private firms take on more operational responsibility, they also inherit more contractual, insurance, and legal exposure.
Cohesity, Rubrik, ServiceNow, and Datadog are building tools to detect and recover from AI-driven corruption and workflow failure. That is not adjacent innovation. It is a market forming around cleanup.
And the pattern does not stop inside software. Backlash against new AI datacenters shows the same cost shift in physical form: power demand, emissions, and local disruption are pushed outward into communities and regulators while the upside stays concentrated elsewhere.
Why Do We Care?
Leadership buys the tool, declares the investment, and moves on. The operator inherits an uncontrolled system with no training, no governance, and no clear accountability when something breaks. And then—this is the part that should make every MSP uncomfortable—the remediation spend shows up somewhere else entirely. Not in the AI subscription line. In incident response. In legal defense. In the Cohesity and Rubrik recovery tools that three major vendors are now racing to build because AI-generated damage is happening at scale.
Grammarly using identities without consent until you opt out isn’t a PR problem. It’s a template. That’s the default posture of the industry: deploy fast, disclaim liability, distribute the cost. The EY data showing 50%-plus of department-level AI initiatives lack formal oversight isn’t surprising—it’s the expected outcome of a market where 85% of tech leaders prioritize speed and 88% of vendors cap their liability.
The Eightfold lawsuit is where this gets concrete for MSPs. Employment decisions made by AI systems that can’t be explained or audited are now class-action territory. If you have clients making hiring decisions with AI tools and you haven’t had that conversation, you are already behind.
And it’s not just vendors doing this. Policy is moving the same way—expanding the private sector’s operational role without removing the liability that comes with it.
What to Consider
- Audit all current service agreements for AI tool coverage. Any agreement that doesn’t explicitly define what AI tools are in scope, what constitutes an AI-related incident, and who bears remediation cost is a liability gap.
- Identify clients using AI in hiring, HR, or employment decisions. That is a billable compliance engagement now, not a future problem.
- Build an AI incident response tier into your service catalog. Cohesity and Rubrik are signaling that AI environment recovery is a distinct workload. Price it separately. Do not let it collapse into standard backup/recovery SLAs—the scope is different and the labor cost is higher.
- Reframe AI governance as a managed risk service, not a feature. The 86% of executives who see AI as essential are the same population whose organizations have 16% training rates and 50%+ governance gaps. That is a consulting and audit opportunity, not a technology sale.
If this trend continues, MSPs will be forced to sell an “AI Governance & Recoverability” bundle the same way they sell security stacks—because customers won’t be able to renew cyber insurance or pass procurement reviews without audit logs, acceptable-use enforcement, training attestations, and a tested recovery plan for AI-corrupted data and workflows.