Business leaders believe artificial intelligence (AI) must understand business processes to deliver effective results highlighting a growing concern that enterprises investing in AI without proper process intelligence (PI) risk inefficiency and poor outcomes.
A study from Celonis revealed 58% of executives worry process inefficiencies will limit AI’s effectiveness, while 81% of companies plan to use AI to improve business processes within the next year.
The report showed more than half of business leaders believe process inefficiencies will undermine AI’s effectiveness.
“Traditional methods of process discovery are no longer sufficient,” said Kerry Brown, transformation evangelist at Celonis.
She explained that process intelligence serves as the foundation for AI, ensuring it operates effectively within complex business environments.
“Process intelligence is the critical enabler for AI—just as maps are essential for GPS, PI provides AI with the necessary business context to navigate operations effectively,” Brown said.
Without this context, AI initiatives risk automating broken processes, leading to poor decision-making and inefficient workflows.
Naresh Duddu, AVP and global head of Infosys’s modernization practice, said organizations should leverage process mining and digital twin technologies to map workflows, identify inefficiencies, and provide AI with contextualized data.
“Embedding AI within business processes requires continuous monitoring and feedback loops to refine decision-making dynamically, ensuring seamless integration between AI and enterprise,” he explained.
Another key challenge in AI implementation is data fragmentation and quality. The report revealed 75% of IT leaders feel they aren’t maximizing their tech investments, and 87% need better visibility into how their systems are actually used.
“A ‘facts over feelings’ approach is critical,” Brown said. “AI success depends on data-driven decision-making over subjective assumptions, requiring strong governance, process standardization, and real-time monitoring.”
Without structured, high-quality data, AI can generate inaccurate results, leading to poor business outcomes.
Moreover, AI adoption depends on trust and transparency within the organization. Employees need to understand how AI enhances their roles rather than replacing them.
“We believe in leveraging AI for modernization as well as embarking upon modernization to leverage AI,” Duddu said. “Identification of inefficiencies in processes and addressing them can include leveraging AI to bridge the gaps.”
He added fixing process inefficiencies would have a huge impact and lot of systems would need to change. So in his view, both should be done in parallel.
“We should identify inefficiencies, prioritize them based on cost-benefit analysis and identify AI-enabled features that can be added at varying points of process improvement even if they are sub-optimal if business benefits outweigh the costs,” Duddu said.
To ensure AI delivers the desired business outcomes, IT teams must adopt a process-first approach. Brown emphasized that AI should enhance business processes, not just automate them.
“AI investments should continuously improve workflows rather than just accelerate inefficient processes,” she said. “Process intelligence acts as a real-time feedback loop, ensuring AI remains aligned with business goals and adapts over time.”
Duddu said data fragmentation, lack of interoperability, resistance to change and nurturing the right skills are common challenges IT professionals should anticipate when modernizing operations with AI integration.
“Enterprises have still not adopted AI at scale and are still operating with pilots and tactical solutions,” he said. “The cost of change could be very high, and AI can help accelerate implementing the changes as well as reducing the costs of transformation.”
He recommended IT teams establish robust data governance frameworks, integrate AI into existing enterprise applications and provide upskilling opportunities for employees.
Additionally, responsible AI must be fully integrated into all processes from the outset, ensuring unbiased data, transparency and compliance with ethical guidelines.
“Addressing security and compliance from the outset will also prevent roadblocks during implementation,” Duddu said.
