What Must Change for AI to Pay Off?

Companies love buying technology tools first and figuring out the work later. During the pandemic, organizations rushed to deploy collaboration software without proper training. Workers struggled and productivity tanked. 

Now we're repeating this pattern with AI, but the stakes are exponentially higher. The result isn’t just missed efficiency. It can lead to disengaged employees, squandered capability and millions in sunk costs for tools that sit underutilized because workflows never changed. 

A 2025 McKinsey & Co. study reported only 21% of companies that use generative artificial intelligence say they have redesigned workflows. In fact, most use cases are off-the-shelf tools with limited customization, offering narrow automation gains. That means 4 out of 5 companies are spending budgets on AI automation that doesn’t move core business metrics. To me, that looks like theater rather than real transformation.

No Handoffs = Handcuffs

The pattern is familiar: Executives buy the story about transformation, IT deploys the tools, human resources rolls out generic modules, managers become bystanders and everyone wonders why productivity barely budges. AI sits idle because nobody designed the handoffs between human and machine.

What companies get instead is automated dysfunction wrapped in sophisticated technology. AI systems create as many problems as they solve: claims misrouted by algorithms, schedules mangled by automation, customers stuck with “system errors” that reps cannot fix. Even when AI adds value, companies track deployment instead of outcomes.

Missing the Point

The deeper issue is cultural. Too many teams treat AI like software installation instead of redesigning how people and machines collaborate. Features get all the attention while the real question goes unanswered: How do we get work done better?

However industry leaders believe that AI can enhance operations, there is a smart way to do it that optimizes chances for success. This shift demands rethinking how you define work, measure performance and develop capability. AI literacy and orchestration are becoming core leadership competencies, not optional skills.

1. Study the work before buying: Start with people and tasks, not vendor demos. Map what is rule-based versus what needs judgment. Identify where mistakes cause the most pain, such as compliance gaps, safety risks or customer churn. Inventory your workforce’s capabilities as data so you can align tasks, skills and tools.

2. Redesign workflows with AI included: Draw how people and machines will work together. Define handoffs, triggers for human judgment, guardrails and escalation paths. If you cannot write a clear playbook for the human-machine process, you are not ready to go live.

3. Learn while doing real work: End one-size-fits-all training. Build micro lessons tied to upcoming projects so skills apply immediately. On-the-job upskilling beats chasing scarce talent and also strengthens retention: 76% of employees are more likely to stay with continuous training, and 86% of HR managers see it as critical for retention, according to an SHRM Business report released this year.

4. Customize training by role: A developer, reviewer and service rep need different lessons even on the same platform. Tailor modules to each role’s actual workflow. Teach reviewers to spot hallucinations and show service reps when to override automation. Relevance keeps people engaged and training effective. Executives give training programs a B-, while employees give them a C. Closing that gap requires contextual learning tied to real work.

5. Measure what matters: Regardless of industry, track where AI contributes to real work, the errors it prevents or creates and how much rework disappears. Gauge employee trust, because if people do not believe in the AI, they will route around it. Dashboards should reflect hybrid team performance, not isolated human or machine outputs.

The added benefit of adopting these practices is that managers no longer sit as passengers in technology rollouts; they become orchestrators of hybrid teams responsible for coordinating how people and AI agents work together. This shift reshapes what great management looks like.

First, performance management evolves. Counting human inputs is obsolete when digital agents handle routine tasks. In a contact center, for example, AI may triage calls while humans step in for the complex escalation. What matters is resolution quality and cycle time, not the number of calls logged by each rep. These outcomes capture how well human and digital teammates perform together.

Second, managers must take responsibility for quality, accuracy and ethical compliance of AI outputs. That means establishing automatic sampling of AI output, setting clear escalation protocols and auditing closely for drift or bias. Digital workers need to be treated like junior colleagues—capable and fast, but requiring supervision, feedback and periodic recalibration.

Finally, managers need to master orchestration. They cannot simply wait for IT to flip the switch on new tools. Instead, they must understand what agents can do, where they fall short and how to build AI-ready workflows so that people apply judgment where it matters most. This manager-led orchestration often delivers results faster and more effectively than months of top-down process mapping that never reaches the front line.

For leaders, this evolution means performance management, learning and leadership development all need to reflect these new hybrid dynamics. The future manager isn’t a process enforcer; they’re the conductor of a human + digital workforce.

The Real Shift

Deploying a tool does not count as victory. Competitiveness now depends on continuous learning tied to real work. Think of this as operating infrastructure.

Treat AI like a new hire. Onboard it, define its role, teach the team how to work with it and measure what the team produces. Organizations that adopt this mindset—tools as teammates, managers as orchestrators, workflows as products—will likely shape the next productivity cycle. Technology doesn’t create advantage, but teams do. The next productivity surge will belong to leaders who treat AI as part of the workforce, not apart from it.