Why AI Operationalization Is the Real Enterprise Challenge

/ AI, Application Development, Enterprise IT Strategy

What’s Happening

ECI Research principal analyst Paul Nashawaty sat down with Molly Moore, President and COO of Liveops, to examine why AI adoption is outpacing AI operationalization across the enterprise. The conversation centers on a pattern now visible across industries: organizations can stand up proofs of concept quickly but consistently fail to embed AI into production workflows with the governance, accountability, and operational design required to deliver real customer outcomes. Moore’s position is direct: execution has become the competitive differentiator, and most organizations are optimizing for the wrong metrics entirely.

The Bigger Picture

Adoption Without Governance Is a Liability

The surface-level AI adoption story looks impressive. ECI Research data shows that 92% of organizations report AI capabilities integrated into at least one stage of their software delivery lifecycle, a sharp increase from 71% in early 2024. But that headline obscures a serious structural problem. According to ECI Research, 50.7% of organizations rely on public AI tools such as ChatGPT and Copilot, while only 20.2% report enterprise-wide AI deployments built on a governed framework.

That gap is not a technology failure. It’s an accountability failure. When customer interactions, business decisions, and operational workflows run through ungoverned AI systems, there is no clear ownership when something goes wrong. Moore put it plainly: organizations applying AI on top of broken processes don’t fix those processes, they accelerate them. Fragmented workflows become more fragmented. Unclear escalation paths become faster and more confusing. The dysfunction simply runs at machine speed.

For ITDMs, this is a budget and risk question as much as a strategy question. Nashawaty noted that a quarter of 2025 IT budgets were allocated to AI projects without clear ROI frameworks. Measuring success by token consumption rather than resolution quality or customer effort is a governance failure masquerading as a metrics problem.

What Mature AI Operationalization Actually Looks Like

Moore’s framework for AI maturity is useful precisely because it leads with organizational design, not technology selection. The four pillars she describes (i.e., governance, workflow readiness, organizational alignment, and execution capability) place accountability structures before any technical deployment decision. Organizations that skip the first three and jump straight to implementation are the ones accumulating pilot debt: a growing inventory of proofs of concept that never reach production at scale.

The orchestration model Liveops advocates for deserves a spotlight. Rather than asking where AI can be deployed, high-performing organizations ask what outcome they’re trying to achieve, redesign the workflow around that outcome, and then determine where AI and human expertise each belong. This is a fundamentally different starting point than the tool-first approach that dominates most enterprise AI programs.

The healthcare example Moore cited is instructive. AI can gather patient information, verify eligibility, and explain a process. But when a member is frustrated or confused, human judgment and empathy become operationally necessary, not just nice to have. The handoff between AI and human agents, with full context preserved and no requirement to repeat information, is where customer trust is won or lost. That handoff is an engineering and governance problem, not just a UX consideration.

What This Means for Developers

Developers building AI-integrated applications need to treat the handoff as a first-class technical requirement. Context persistence across interaction channels, clean escalation APIs, and auditability of AI decisions are not post-launch improvements. They’re production requirements from day one. The organizations seeing the worst outcomes are those where AI systems make decisions that nobody can explain or attribute, because accountability was never designed into the architecture.

ECI Research data points to a meaningful perception gap that developers and engineering managers should recognize: 45% of AI practitioners cite operational complexity as their primary challenge, compared to only 31% of managers, who more commonly prioritize reliability outcomes. Practitioners experience friction daily that leadership doesn’t see in aggregate reporting. That gap in perception is itself a governance problem. When the people closest to the system aren’t surfacing operational signals that reach decision-makers, the organization is flying partially blind.

The Measurement Problem

The metrics most organizations use to evaluate AI, speed, containment rates, cost reduction, are operational efficiency metrics. They measure how the system ran, not whether the customer succeeded. Moore’s AI maturity benchmark draws a sharp distinction: mature organizations evaluate AI on resolution quality, customer effort, operational consistency, and business outcomes. Efficiency improvements that come at the cost of customer outcomes are not wins. They’re optimization in the wrong direction. This has direct implications for how ITDMs structure AI investment governance. Procurement decisions and vendor evaluations need to align on outcome-based metrics before deployment, not after the fact when course correction is expensive.

For a deeper look at this shift toward outcome-based measurement, Liveops’ 2026 Resolution Gap Report examines where organizations are still falling short in turning AI-driven customer interactions into effective resolutions. The report offers additional insight into the gap between automation, operational execution, and the customer outcomes that ultimately determine whether AI investments are delivering value.

What’s Next

From Experimentation to Execution: The 2026 Priority

The market is moving from a question of whether to adopt AI to a question of how to operationalize it at scale without accumulating governance debt. Liveops occupies a specific and defensible position in this transition: the combination of operational design, a managed workforce, and technology integration addresses the full stack of what organizations actually need, not just the software layer.

The companies that will lead in AI-driven customer experience over the next 18–24 months are not those with the most advanced models. They’re those that have done the harder organizational work: clear ownership structures, optimized workflows before automation, and measurement frameworks tied to customer outcomes. That’s a leadership problem, as Nashawaty noted, before it’s a technology problem.

Agentic AI Raises the Stakes on Governance

As agentic AI moves from pilot to production, the governance requirements become more acute, not less. ECI Research’s 2025 AI Builder Summit survey found that two-thirds of enterprise AI leaders have already implemented multi-agent collaboration in live or pilot workflows. When agents can coordinate and delegate tasks autonomously, ambiguous ownership and ungoverned handoffs carry higher consequence. Organizations that haven’t built accountability into their current AI deployments will find the problem compounded when agents begin making chains of decisions without a human in the loop.

The crawl-walk-run framework Moore described offers a practical sequencing tool: governance and workflow readiness first, alignment second, and only then execution at scale. Vendors and advisors who help organizations honestly assess where they currently stand on that continuum, before selling them on the next capability, will earn the durable relationships that matter in this market.

Author