Fixing Enterprise AI’s ROI Problem: Lessons From the People Side

Enterprises are pouring money into AI pilots, platforms and partnerships, yet many still struggle to see clear, sustained returns. Technology alone is not the problem; the people and processes around it are. Looking at AI through the lens of culture, skills and change management reveals why ROI stalls — and what leaders can do to turn scattered experiments into measurable value. This article unpacks practical, people-centric strategies to close the gap between AI ambition and actual impact.

Why Enterprise AI Struggles to Prove Its ROI

Across large organisations, AI is now part of almost every strategy deck and budget cycle. Yet when boards ask, “What did we actually get for this investment?”, answers are often vague: a handful of pilots, some productivity anecdotes, a showcase demo. The problem is rarely a lack of algorithms or infrastructure. It is a lack of alignment between AI initiatives and how people really work, decide and create value.

Leaders like Atlassian’s Chief People Officer increasingly frame AI not as a pure technology project, but as a human systems challenge. If teams, incentives and culture do not evolve alongside tools, AI becomes a cost line, not a value engine. Fixing enterprise AI’s ROI problem therefore starts with rethinking skills, ownership, governance and everyday workflows.

From Shiny Pilots to Business Outcomes

Many enterprises get stuck in what might be called the “AI theater” phase: high-profile proofs of concept that seldom move into the operational core of the business. This happens for several reasons:

• No clear value hypothesis: Projects begin with technology curiosity rather than a sharp business problem.
• Disconnected sponsors: Pilots are run by innovation labs without strong line-of-business ownership.
• Lack of integration: Even successful prototypes do not plug into existing systems or workflows.
• Unmeasured impact: Teams track activity (models built, prompts tested) instead of impact (time saved, revenue protected, risk reduced).

To move beyond this stage, organisations must shift from measuring how much AI they have to how much friction they remove and value they create.

Make the People Function a Co-Owner of AI

One of the clearest shifts in organisations that do see returns is the role of the people function (HR, talent, learning). Instead of treating AI as solely a CIO- or CTO-led domain, they position People teams as strategic co-owners. Why?

• Skills and capability: AI value depends on what employees can actually do with it, not just what the tools can do in theory.
• Change readiness: People leaders understand morale, trust and resistance patterns across the company.
• Work design: As tasks and roles shift, job architecture, performance expectations and career paths need redesign.

When People leaders sit alongside technology and business leaders in AI steering groups, decisions naturally account for adoption, skills and ethics – not only infrastructure and budgets.

Anchor AI Projects in Specific Workflows

AI ROI becomes tangible when it is embedded into repeatable, measurable workflows. Rather than launching broad, abstract programs, high-performing enterprises pinpoint a narrow set of use cases where AI can relieve immediate pain or unlock new capacity.

Choosing the Right First Use Cases

• Start where processes are well-understood and routinely performed.
• Prioritise areas with clear cost or time baselines (e.g., average time to complete a support ticket).
• Aim for low-risk, high-friction processes (internal operations before customer-critical decisions).
• Ensure a direct line of sight to a financial or strategic metric.

Examples include automating internal knowledge search, drafting routine documentation, or assisting support teams with suggested replies – all of which can be timed, tracked and iterated.

Measure What Actually Changes in the Work

To claim ROI credibly, organisations need to measure how AI changes real work, not just how often it is used. That means tracking three levels of impact:

1. Task-level efficiency: Minutes saved per task, error rates, rework frequency.
2. Team-level throughput: Volume handled, cycle time improvements, backlog reduction.
3. Business outcomes: Revenue lift, churn reduction, faster time-to-market, lower operational risk.

Connecting these levels requires intentional design before pilots begin: define baselines, select leading and lagging indicators, and decide how you will attribute improvements to AI versus other changes.

Invest in AI Literacy, Not Just Specialists

Another barrier to ROI is the assumption that AI value will radiate outward from a small group of experts. In reality, the majority of gains come when frontline employees and managers feel confident experimenting with and questioning AI outputs.

Core Elements of Enterprise AI Literacy

• Conceptual understanding: What AI can and cannot do; limitations and failure modes.
• Responsible use: Data privacy, security, bias, and when to escalate or override AI decisions.
• Prompting and configuration skills: How to express tasks clearly, iterate and evaluate outputs.
• Critical thinking: Comparing AI suggestions to domain knowledge and organisational policies.

People leaders can embed this literacy into onboarding, leadership programs and regular learning cycles, treating AI competency as core to modern work rather than a niche skill.

Redesign Roles and Performance Expectations

If job descriptions, KPIs and incentive structures remain unchanged while AI is introduced, employees naturally treat AI as an optional add-on. To unlock meaningful ROI, organisations must deliberately redesign work around human–AI collaboration.

Practical Adjustments to Make

• Clarify ownership: Specify which tasks AI assists with and where humans remain decision-makers.
• Update KPIs: Incorporate how effectively employees leverage AI tools into performance discussions.
• Protect time for experimentation: Allocate a small percentage of time for teams to improve workflows with AI.
• Recognise and reward: Highlight teams that demonstrate measured, responsible gains from AI use.

When performance frameworks reinforce AI-enabled ways of working, adoption stops depending on individual enthusiasm and becomes part of how the organisation operates.

Build Guardrails: Governance That Enables, Not Blocks

Governance often shows up as a brake on AI initiatives, introduced late in response to risk concerns. A better pattern is to design enabling guardrails early, giving teams clear boundaries within which they can move fast.

Key Elements of Practical AI Governance

• Use case tiers: Classify use cases by risk (e.g., internal productivity vs. customer-facing decisions).
• Approval flows: Lightweight review for low-risk experiments, deeper oversight for high-impact deployments.
• Data policies: Clear rules on which data can be used for training, prompts and outputs.
• Explainability expectations: When and how AI decisions must be documented or explainable to humans.

Cross-functional councils that include people, legal, security and business leaders can keep governance aligned with both risk appetite and innovation needs.

Comparing Approaches to Scaling AI in the Enterprise

The emerging best practice blends these approaches, using a joint ownership model where People teams help ensure that AI efforts are usable, trusted and embedded in real work.

A 7-Step Playbook to Turn AI Spend into Measurable ROI

Enterprises looking to course-correct can follow a simple, repeatable sequence that keeps both technology and people in view.

1. Clarify outcomes: Pick 2–3 business metrics you aim to influence (e.g., time-to-resolution, sales cycle length).
2. Map workflows: Identify where work currently slows down or is error-prone within those metrics.
3. Select use cases: Choose 1–2 low-risk, high-friction workflows to augment or automate with AI.
4. Co-design with teams: Involve frontline employees, managers and People partners in designing how AI fits into daily work.
5. Define guardrails and baselines: Set policies, ethics guidelines and starting metrics before launching.
6. Launch, learn, iterate: Run a time-boxed rollout, gather feedback, adjust prompts, processes and governance.
7. Scale and codify: Once impact is proven, update training, performance frameworks and tooling so the new way becomes the standard.

Final Thoughts

Enterprise AI’s ROI problem is not a deficit of algorithms, but a misalignment between tools, people and value. Organisations that treat AI as a joint responsibility across technology, business and people functions are far more likely to see sustained impact. By choosing grounded use cases, measuring real work changes, investing in AI literacy and redesigning roles and governance, leaders can move beyond AI theater and build systems that reliably turn investment into outcomes.

Editorial note: This article was inspired by commentary around enterprise AI and people strategy, including perspectives from Atlassian’s leadership. For further reading, visit the original source at raconteur.net.