Companies in regulated industries are typically subject to strict compliance requirements, and legacy system upgrades are required for AI-based processes.
Rapid advances in AI are outpacing enterprise adoption, creating a widening “deployment gap” as most organisations struggle to move AI projects beyond pilot testing into core business operations.
Speaking at the company’s Investors AI Day, Infosys co-founder Nandan Nilekani had noted that due to the rapid pace of AI innovation, technology advancement is racing ahead of enterprise deployment, creating a widening gap between model capability and real-world implementation. This, he described, as a deployment gap.
Similarly, Piyush Goel, CEO & Founder of Beyond Key, observed that 88 per cent of companies claim to have used AI at some point, but most have not found success through the implementation process.
“Most AI implementations remain in the pilot stage, with organisations still testing how to integrate the technology into workflows. The next step is to determine how companies integrate AI into their core workflows and business processes, and how they leverage existing infrastructure,” he said.
Many organisations still measure AI success through model accuracy, tool usage, or localised efficiency gains. According to Biswajeet Mahapatra, Principal Analyst, Forrester, there is a disconnect because executives expect revenue growth or margin impact, while very few firms can tie AI initiatives directly to profit and loss outcomes at the enterprise level.
No strong base
A lack of unified infrastructure and strong data architecture remains a major hurdle for organisations implementing AI, as many global enterprises still operate with fragmented data systems, Goel said.
AI also requires a unified data layer, industry-standard modern cloud-based infrastructure that supports transaction processing and real-time analytics, and organisations must have strong governance frameworks to support the successful implementation of AI.
“Without these foundational layers, AI models can’t access the clean and reliable data for their implementation. While an organisation’s legacy systems impede the ability for AI applications, the major contributor to the successful implementation of AI is organisational – disconnected ownership of the data, unclear governance of the data, and lack of alignment between it and the organisation’s business operations,” he said.
Big upgrades needed
Sector-wise, AI stays experimental where outcomes are hard to measure, liability is high, or workflows are fragmented. This includes public sector, healthcare delivery, heavy industry operations and highly customised back-office processes. Banking, finance, technology, telecom and digital native companies are early adopters of new technologies because they already utilise data-driven approaches to deliver their products and services.
Industries like public service, manufacturing and segments of regulated healthcare are still conducting pilot tests with AI. The reason for this lack of progress is due to the complexity of making AI operational versus the interest level. Companies in regulated industries are typically subject to strict compliance requirements, and legacy system upgrades are required for AI-based processes.
“It becomes operational fastest where data is already digital, and feedback loops are short: fraud/risk in financial services, customer service/contact centres, digital commerce/marketing, and software engineering. Firms with strong platform engineering and product operating models also deploy faster than firms organised around projects,” said Ashish Banerjee, Sr Principal Analyst at Gartner.
‘Frontier keeps moving’
He added that there is no single catch-up date because the frontier keeps moving. Most enterprises will reach baseline operational maturity in 12–18 months, assuming sustained investment and leadership attention. Broad, repeatable deployment across many functions is across 3–5 years, especially in regulated or legacy-heavy environments.
Forrester, on the other hand, expects 2026 to be a year of correction rather than acceleration, with meaningful catch-up occurring through 2027 as enterprises move from hype-driven experimentation to disciplined deployment, delayed spending, and production-focused AI that prioritises trust, governance and measurable value over speed.
Published on March 8, 2026
