Enterprise AI Barriers
Enterprise AI Adoption Stalls As Companies Face Data And Deployment Barriers
Anonymous
Key Points
1
Comp AI’s founders previously built LeapAI, a workflow platform that surpassed one million users before being shut down because it lacked a sufficiently “sticky” use case. Their experience scaling that company—and spending months handling SOC 2 compliance manually—helped motivate Comp AI’s new focus.
2
Teradata’s Agentic AI Maturity Index finds that only 7% of organizations have reached the “operationalizing” stage. Forty percent are still “developing,” 28% are experimenting, and 25% are building, indicating that most companies have not yet established the data and governance foundations needed for scaled autonomous systems.
3
OpenAI co-founder Greg Brockman described his preferred direction as a single, unified AI that reduces the need for people to “wrap[] yourself around the computer,” arguing that computer-use capabilities may be simpler than rebuilding every software product for agents.
4
Security experts criticized the emphasis on third-party AI audits as potentially diverting attention from basic internal controls. Luta Security CEO Kate Moussouris called the proposal an example of “outsourcing,” while researcher Sayash Kapoor argued that marginal investments in control may be more effective than additional alignment work.
Enterprise AI is moving from early experiments toward real, working systems that can handle security and compliance tasks on their own. But fragmented data, tangled software pipelines, and weak oversight are blocking most companies from actually deploying these agents at scale. TechCrunch reported that Comp AI, a new security startup, just raised $34 million in Series A funding to build autonomous agents that create policies, collect audit evidence, and monitor controls continuously — a move that reflects growing market demand for AI systems that work without constant human intervention.
According to Teradata's research, only 7% of enterprises have reached the stage where they can truly operate agentic AI systems. Forty percent are still developing their approach, 28% are experimenting, and 25% are building initial prototypes. That means 93% of companies lack the data foundations and governance tools they need to scale autonomous AI — a massive gap between the hype and what companies can actually do today.
From LeapAI Frustration to Security Automation
Comp AI's three founders — Lewis Carhart, Claudio Fuentes, and Mariano Fuentes — learned their lesson from LeapAI, a workflow platform they built that hit one million users before shutting down. The problem was simple: the product lacked a strong enough business case to survive. During those months of rapid growth, the team spent enormous time on compliance work, manually handling SOC 2 audits instead of building features. That painful experience inspired Comp AI's focus on automating security and compliance tasks that large organizations struggle with every day.
Carhart stated that TechCrunch security and compliance are directly tied to revenue at most software companies. He emphasized a key reality: "An agent might draft a policy, for example, but a person still reviews and approves it." In other words, Comp AI's agents are not meant to replace humans — they reduce tedious, repetitive work and let people focus on decisions that matter.
Data Fragmentation and Model Fallbacks Stall Deployment
Even when companies want to deploy agentic AI, they run into hard technical barriers. Practitioners cite problems like unreliable multi-provider fallbacks (systems that should switch to a backup AI model when the primary one fails, but often do not), inconsistent model limits across different AI services, and weak observability — meaning they cannot see what the agent is actually doing or why it failed. On top of that, human-review processes are inefficient, turning what should be a quick sign-off into a slow bottleneck.
OpenAI co-founder Greg Brockman prefers a simpler approach: build one unified AI system that can interact directly with computers, rather than creating separate connectors for every single application. Other companies are developing Model Context Protocol (MCP) servers, APIs, and command-line integrations to bridge the gap. But the real challenge is not technical plumbing — it is governance. Organizations deploying AI across hybrid environments must weigh data gravity, latency, cost, performance, and control while keeping regulatory compliance in mind.
Only 7% of Enterprises Ready to Operationalize Autonomous AI
Teradata's 2026 maturity research found stark numbers: despite 90% of tech leaders increasing AI investment, 63% report minimal or no measurable ROI. Healthcare, manufacturing, retail, and financial services remain largely unprepared to run agentic systems at scale. The core issue is not money or technology — it is that most companies lack unified data foundations and the governance structures needed to let AI systems operate without constant human supervision.
Safety experts argue that organizations need stronger permissions controls, continuous logging, and better incident-response procedures alongside external audits. Kate Moussouris, CEO of Luta Security, criticized the push for third-party AI audits as "outsourcing" core risk management. Sayash Kapoor, an AI researcher, went further, arguing that basic improvements to internal software controls often deliver far more safety value than additional theoretical alignment research.
Tying AI ROI to Real Business Outcomes Remains Elusive
The biggest challenge facing enterprises is not building AI agents — it is connecting them to measurable returns. Companies deploying hybrid AI workloads must consider where data lives, how fast it needs to move, and what it costs to run agents in different places. ESG applications, supply-chain optimization, and compliance automation all promise value, but only if organizations actually finish what they start.
The gap between pilot projects and production systems is widening. Companies need to stop treating AI as an isolated experiment and start treating it as infrastructure that serves a clear business goal. Until enterprises build proper data pipelines, governance layers, and clear ROI metrics, most agentic AI deployments will stay trapped in the "developing" and "experimenting" stages — far short of the autonomous, production-grade systems they promised their boards they would deliver.
Sentiment
Mostly NegativeNEGATIVENEUTRALPOSITIVE
Publishers
48
Articles
16
Reach
64
A
AnonymousAuthor
