Scaling AI agents a bottleneck for Indian enterprises: Dynatrace’s Rafi Katanasho

Scaling autonomous artificial intelligence (AI) agents within enterprises in India remains a challenge as businesses struggle to trust them with critical processes, believes Rafi Katanasho, Asia Pacific chief technology officer (CTO) and vice-president of solution sales at Dynatrace.

Dynatrace is an AI observability and application performance monitoring (APM) platform to manage cloud-native, hybrid, and multi-cloud environments.

Katanasho believes autonomous agents can fail silently, delivering poor business outcomes without triggering system alerts. “India has largely solved for access to AI. However, the bottleneck is the confidence in delivering AI in a production environment, at scale.”


The concern is agents can access systems, use tools and carry out tasks on behalf of employees, exacerbating the consequences of a wrong decision.

Katanasho said many companies are taking a “bolt-on AI” approach, automating smaller parts of existing processes instead of redesigning entire workflows around autonomous systems. Critical deployments, he added, require greater confidence on agents to behave as expected, even during situations they were not explicitly trained or instructed to handle.

Deloitte India’s State of AI in the Enterprise report released in March 2026 found 40% of Indian respondents reported significant or full AI usage, compared with 28% globally. At-scale deployment was strongest in product development, strategy and operations, and marketing and sales.

But deploying AI is different from giving agents the authority to take decisions and act without human intervention.

Why accountability matters?

Indian banks are already experimenting with AI agents for parts of know-your-customer (KYC) processes and credit validation. But Katanasho said companies are being cautious about extending automation to entire business processes because the consequences of errors can be significant.

He cited the scale of India's digital payments ecosystem to illustrate the risk. Even if an AI system is 99.9% accurate, he said, the remaining 0.1% could still create a meaningful challenge when applied across a large volume of transactions.

“All of a sudden, you’re not getting big errors, you're just getting bad business results,” Katanasho said.

For enterprises, the challenge is identifying when an agent is behaving unexpectedly, rather than simply checking whether the underlying application is running.

This is where enterprises need more accountability.

Katanasho said companies need to monitor how agents interact with one another, the models they use and the tools and enterprise systems they access. Tracking individual agents in isolation may not reveal where a mistake occurred or how it affected the final outcome.

The stakes are high in sectors such as banking, fintech and healthcare, where errors in automated processes can affect sensitive data, service availability and overall business operations.

“Autonomy requires authority. But authority without the truth is very dangerous,” he said. “You need an independent system to be able to observe all these interactions end to end and hold the outcome accountable, not just each agent accountable.”

Where’s the ROI?

For Katanasho, the question is no longer how much companies are spending on AI, but whether that investment is delivering measurable returns.

“There’s a lot of tokens being consumed here. But what is the outcome? What are the business outcomes from that investment going on?,” he asked.

He said boards and chief financial officers across India are increasingly asking how AI spending translates into business results, particularly as companies redirect money from traditional technology investments towards AI.

While employees may feel more productive using AI tools, he believes the bigger opportunity is in redesigning entire workflows around AI agents. “The ultimate value, in my view, from agentics specifically, is transforming that business process altogether. Process reengineering through AI.”

Currently, companies are automating low-risk, clearly defined processes where AI can deliver relatively predictable results. But Katanasho said the larger benefits could come from AI-native projects that rethink how a business process works from the ground up, rather than adding AI to existing systems.

He compared this transition to the shift to cloud computing, where businesses that redesigned their operations around the technology captured greater benefits than those that experimented with it.