Businesses across industries have fully embraced AI, looking to 10x productivity and supercharge profits. Most companies—78%, according to McKinsey—use AI in at least one business function. But a recent survey by IBM found that only 1 in 4 AI pilots brought about the ROI leadership expected. Even fewer (16%) had been scaled across organizations.

The gap is real. Many AI efforts remain stuck in pilot mode or isolated at the edges of businesses. When adopting new technologies, organizations typically invest in training, update their processes, and plan for change. With AI, many teams are rushing in without doing that foundational work, leading to stalled progress and limited returns.

At PagerDuty, we’ve seen that the companies that succeed with AI start from a different place: a deep understanding of how their people work and where technology can truly make a difference. That means identifying your biggest operational challenges and finding the right AI approach—whether GenAI, embedded intelligence, or autonomous agents—to solve them. 

Why AI initiatives stall and fail

AI’s rapid development has made it difficult for executives to chart a smart path to growth without falling behind competitors. Leaders feel intense pressure to move fast, which puts them in a tough spot when new tech comes on the scene. 

Cisco’s 2024 AI Readiness Index found that 85% of executives believed a failure to deploy AI within 18 months of the survey date would cause them to fall behind competitors. At the same time, in IBM’s 2025 C-Suite Study, only 37% of CEOs believed “fast and wrong” was better than “slow, but right” when it came to adopting new tech. 

Many organizations have launched AI programs without an overarching plan. The result: AI initiatives start well but fail to scale or hit ROI targets, leaving companies wondering what their investment was for. 

Leaders who push forward in this situation typically end up with fragmented AI implementations. Teams are left to choose their own AI strategy and vendors rather than evaluating options based on company priorities or operational needs. At best, their projects don’t integrate; at worst, teams end up pushing for different approaches and creating more work for each other.

On the other side of the equation are leaders who are too cautious. They delay decisions, wait for the perfect strategy, or hold back until every question is answered. Employees, seeing competitors pulling ahead or simply recognizing the potential in AI, jump in on their own. Or jump to another organization, leaving a talent and skill gap in their place.

When formal adoption lags, “shadow AI” proliferates, exposing companies to risk and making it harder to formalize a cohesive strategy down the line. Employee-driven adoption also comes with risks. CybSafe and the National Cybersecurity Alliance found that 38% of employees have shared confidential or otherwise private information with these systems without their employers’ knowledge. 

Leaders who hope to avoid these pitfalls must be strategic yet agile, adapting to new use cases and pivoting internal initiatives when they don’t pan out as expected. More importantly, they must identify areas of potential for AI and resist the temptation to go big before seeing the results they’re after.

How to move forward: Align AI with how your team actually works

AI’s superpower in the workplace isn’t replacing humans. It’s augmenting your best workers’ capacity. That’s why all successful AI programs start with a grounded understanding of how your organization works and where the greatest potential for transformation lies. 

To find those areas of potential, ask:

• What repetitive work do teams do? 
• What workflows slow teams down or burn employees out?
• What processes are partially automated but still require human intervention?
• What decisions that employees make are routine or predictable enough that they can be delegated to a well-programmed tool?

For example, engineers frequently have to respond to alerts that may signal an impending incident. However, some of these alerts are simple issues that can be remediated with pre-existing workflows or automations. These menial fixes frequently interrupt higher-level work, but they can’t safely be ignored without putting your company at risk of a serious incident. If AI could step in to intercept low-level alerts and auto-remediate problems, employees would have more time for the type of novel work AI can’t manage. 

Your first projects should be small, focusing on repetitive, low-level tasks. Seek early wins that will show the potential of AI, building confidence in your efforts. AI that makes flow-chart decisions or connects existing automations into end-to-end workflows will free up your employees from toil.

Your employees are the best source of information on where AI can be most useful. Involve them at the strategy stage to foster trust and ownership over your company’s AI plans and policies, thus boosting each pilot’s chance of success. Treat engineers and operators as the primary architects of AI-driven workflows. They’re the ones who will be supervising these tools and their work. 

Once your project has launched, make time to hear and act on feedback. Your front-line employees are best positioned to tell you whether a pilot is working as planned and suggest changes that might improve a flailing initiative or redeem a dead-end one. 

It’s time to greenlight AI progress

The right vendor will enable you to build resilience while reducing manual toil with AI. Look for platforms built to address real problems in the workplace rather than systems that add AI without considering how it will benefit end users. Avoid fragmentation by making sure the technologies you select work with the rest of your tech stack. Finally, prevent unexpected privacy or security risks by choosing enterprise-grade tools with security certifications and built-in compliance. 

PagerDuty’s approach to AI-first operations is grounded in the principles outlined above. With more than 15 years of experience supporting mission-critical operations, we know what it takes to make AI work in complex environments. While others experiment, PagerDuty delivers reliable, AI-powered operations grounded in data, resilience, and real-world context. Learn more in The PagerDuty Vision for AI-First Operations.