Markets underestimating complexity of enterprise AI rollout: Cognizant AI Officer Hodjat
Read Time:3 Minute, 4 Second
Babak also said that enterprises do not need perfectly organised data before deploying AI, as agent-based systems can work across fragmented environments.
- Enterprise AI adoption is slower and more complex than expected
- Deep integration and engineering ensure safe, reliable AI at scale.
- AI boosts productivity but requires ongoing organisational effort
The market may be moving faster than enterprise reality when it comes to artificial intelligence (AI) adoption, with companies underestimating the engineering complexity required to deploy AI safely and reliably at scale, according to Cognizant’s Chief AI Officer Babak Hodjat.
Enterprises are increasingly experimenting with Generative AI (Gen AI) and agent-based systems, but integrating these technologies into legacy environments requires deep design, domain expertise, and organisational understanding rather than plug-and-play deployment, Hodjat told Moneycontrol.
“AI is an engineered discipline,” he said in an interview on the sidelines of the Nasscom Technology & Leadership Forum 2026, adding that making systems safe and reliable demands careful architecture and a detailed understanding of client processes.
While expectations may currently be “displaced,” he said the industry will gradually recognise the effort needed to operationalise AI inside complex enterprises.
The comments assume significance after a recent rout in global IT stocks, triggered by breakthroughs from AI research firm Anthropic, has intensified fears of AI-led disruption to outsourcing. Markets are pricing in rapid automation, even as industry leaders caution that enterprise adoption will be slower and far more complex than expected.
Hodjat’s remarks add to a growing chorus of tech leaders rejecting the view that end-to-end AI plugins can quickly replace enterprise work that has traditionally powered the IT industry.
Cognizant CEO Ravi Kumar S, Infosys Chairman Nandan Nilekani, Wipro Chief Technology Officer Sandhya Arun, and a plethora of tech leaders have recently said that enterprise AI adoption requires deep integration, data readiness, governance layers, and significant engineering effort, making large-scale automation far more gradual than market expectations suggest.
Also, read: Nandan Nilekani flags implementation gap, says AI models outpacing real-world use
Productivity gains, but more work ahead
Hodjat pushed back against concerns that AI-led productivity improvements could shrink technology services demand, arguing that rising efficiency has historically expanded the scope of work rather than reduced it.
Programming tools and software development practices have continuously improved productivity over decades, he added, and AI represents a faster continuation of that trend rather than a structural break.
“We expect ourselves to be more productive, clients expect us to be more productive, but at the same time we are expected to do more,” he said, describing AI as a “force multiplier” that enables companies to solve larger enterprise problems.
According to him, enterprises have only begun applying AI in narrow use cases such as software development lifecycle (SDLC), service desks, and natural language processing tasks like document extraction and translation.
Significant opportunities remain untapped in areas such as process discovery, knowledge acquisition across disparate systems, etc.
AI without perfect data
A common assumption among enterprises is that data must first be fully harmonised before AI deployment.
Hodjat disagrees.
He said multi-agent systems can operate across fragmented data environments by placing AI agents on top of existing systems and orchestrating interactions dynamically. While such setups may initially be less efficient, they allow organisations to begin deriving value immediately while modernisation continues in parallel.
In large enterprises built through acquisitions, data is often spread across multiple systems.
Agentic AI can bring these systems together in real time for users, even if the backend remains fragmented.
“The highest priority is solving the user problem first,” he said, adding that deeper consolidation can follow later.
Happy
0 %
Sad
0 %
Excited
0 %
Sleepy
0 %
Angry
0 %
Surprise
0 %
0
0
Average Rating