Enterprises Race to Scale AI Beyond the Pilot Stage

Key Points

• 80% of Fortune 500 companies use agentic AI, but only 6% — McKinsey's 'AI high performers' — attribute at least 5% of EBIT to the technology.
• Bupa cut its mobile app modernization timeline by 60% using AI-assisted engineering, lifting its app rating from 3.7 to 4.7 in the process.
• Jabil, operating across 100-plus global manufacturing sites, is prioritizing data integration as the prerequisite for any AI deployment.

Sources

Key Points

• Workflow redesign is the decisive variable: McKinsey's top AI performers are nearly three times more likely to have fundamentally restructured processes, not just layered AI on top of existing ones.
• The productivity paradox is real and urgent: 80% of employees report AI makes them more productive, yet only 37% of organizations see it move the needle on operating profit — a disconnect that signals value is being created but not captured at the company level.
• Data infrastructure is the bottleneck before AI, not after: both Jabil and Bupa discovered that the foundational work — integration, standardization, clean data architecture — had to precede AI deployment, not follow it.
• Scale creates a compounding advantage: large companies (over $1 billion in revenue) are scaling AI agents at nearly double the rate of smaller firms, and that gap grew significantly in just one year.
• Legacy modernization is no longer optional: end-of-life technology stacks are actively preventing AI adoption, and AI-assisted engineering tools are now cutting modernization timelines by 60%, removing a key excuse for delay.

Sources

Key Actors

Sources

winners

• Enterprise software vendors with deep integration capabilities — SAP, Salesforce, ServiceNow — as data unification becomes the prerequisite spend before AI can scale.
• AI-native systems integrators and modernization consultancies like Infosys, which can now promise 60% faster delivery timelines using AI-assisted engineering, dramatically improving their competitive economics.
• Large enterprises with over $1 billion in revenue, which are already scaling AI agents at twice the rate of smaller firms and are compounding that structural advantage with each budget cycle.
• Employees in AI-mature organizations who are being freed from manual data reconciliation and repositioned toward decision-making and insight work — a genuine quality-of-work upgrade.

losers

• Mid-market and smaller firms, which lag significantly in enterprise-wide AI scaling (22% versus 54% for large companies) and risk falling into a permanent capability gap as large competitors automate faster.
• Organizations still running isolated AI pilots without connecting them to workflow redesign — they are spending AI budget without generating EBIT impact, a position that becomes increasingly untenable as boards demand returns.
• Legacy technology vendors whose end-of-life platforms are actively blocking AI adoption, as companies like Bupa demonstrate that migration is now faster and safer than it has ever been, removing the inertia that kept legacy stacks entrenched.
• Business analysts and process workers in organizations that have not yet automated manual data reconciliation — the 400 hours of effort Bupa eliminated per project represents a category of work that is rapidly disappearing.

implications

• The next enterprise AI budget cycle will increasingly be structured as a workflow redesign budget, not a technology procurement budget, forcing CIOs and CFOs to co-own AI investment decisions in new ways.
• Data integration is becoming a competitive moat: companies like Jabil that establish a single trusted data backbone early will find AI adoption compounding in value, while laggards face a growing remediation cost.
• AI governance and agent oversight are moving from theoretical concern to operational requirement as agents take on consequential tasks — organizations that build governance frameworks now will face less disruption as regulation tightens.
• The talent market for legacy technology specialists will continue to compress sharply, making the cost of delaying modernization higher with every passing quarter.

minority report

• The 'workflow redesign' narrative may be overstating organizational agency: McKinsey's 6% of high performers could simply reflect companies that were already operationally excellent before AI, meaning AI is amplifying existing advantages rather than being the cause of them — which would imply that most companies cannot replicate the playbook regardless of intent.
• The 60% time reduction in Bupa's modernization may not generalize: the Xamarin-to-native migration was a relatively well-scoped codebase replacement, and AI-assisted reverse engineering becomes exponentially harder with truly complex, decades-old mainframe systems that dominate banking, insurance, and government — sectors where the modernization case is most urgent.
• Token costs as a constraint deserve more attention than they receive: 20% of McKinsey respondents say AI-related operating costs are already limiting their use of the technology, a figure that could rise sharply as agentic AI — which consumes far more compute per task than copilot-style tools — scales across enterprises.

Detected Trends

Sources

second order

• Enterprise software consolidation will accelerate: as data integration emerges as the decisive AI prerequisite, companies will rationalize their tool landscapes aggressively, concentrating spend with fewer, deeper platform vendors — a significant headwind for point-solution SaaS businesses.
• CFO and COO authority over AI budgets will grow at the expense of CTO-led technology procurement, as the McKinsey data reframes AI ROI as a workflow and operations problem rather than a technology selection problem.
• The consulting and systems integration market will bifurcate between firms that can credibly deliver AI-accelerated modernization timelines and those offering legacy project management playbooks — the 60% time compression Infosys demonstrated will become a baseline expectation, not a differentiator, within two years.

prediction

• Within 18 months, 'AI readiness' assessments — measuring data integration maturity, workflow redesign depth, and agent governance infrastructure — will become a standard component of enterprise M&A due diligence, as acquirers price the cost of AI remediation into valuations.
• Agentic AI governance frameworks will become a regulatory focal point by 2027, particularly in regulated industries like healthcare and financial services, as agents take on consequential decisions — companies that built governance early, as NiCE's Chandra describes, will have a significant compliance cost advantage.
• The share of large enterprises scaling AI agents in at least one function will surpass 60% within 12 months, up from 40% today, driven by the compounding effect of early integrations and falling implementation costs from AI-assisted tooling.

minority report

• The enterprise AI investment wave could face a sharp correction before it pays off at scale: with 20% of companies already reporting token costs as a constraint and AI budgets exceeding 10% of IT spend at 28% of organizations, the economics of agentic AI — which is dramatically more compute-intensive than earlier AI tools — may trigger a forced consolidation of use cases rather than the predicted expansion, temporarily reversing the scaling narrative.
• Workflow redesign as a source of AI ROI may be a correlation rather than a cause: McKinsey's high performers may simply be companies with stronger operational discipline, better data practices, and more capable leadership overall — meaning the prescription of 'redesign your workflows' could be systematically giving organizations the wrong lever to pull, while the actual variable is organizational capability that cannot be directly replicated through process initiatives.

What This Means

Sources

implications

• AI strategy and operational strategy must merge into a single planning process — organizations that maintain separate technology and business transformation roadmaps will consistently underperform those that integrate them.
• Governance and agent oversight infrastructure is the next critical investment gap: as agentic AI takes on consequential tasks, the organizations without explicit governance frameworks will face both regulatory risk and operational failure modes that are difficult to remediate at scale.
• The 'simplify-first, innovate-second' principle from Jabil applies universally — complexity is the silent tax on AI ROI, and every integration added without simplification compounds the cost of future deployments.

second order

• A new class of enterprise AI auditor will emerge — internal or third-party — tasked with assessing data integration maturity, workflow redesign depth, and agent governance quality as standard operational risk functions, analogous to cybersecurity audits today.
• Human workforce strategy inside AI-mature enterprises will shift from headcount reduction narratives toward capability repositioning — the Bupa example of eliminating 400 hours of manual BA effort to give analysts capacity for higher-order work is the template, not mass layoffs.
• Point-solution SaaS vendors without deep integration capabilities or platform-level data connectivity will face accelerating churn as enterprises consolidate around fewer, more interoperable platforms to support AI scaling.

minority report

• The entire enterprise AI scaling narrative may be structurally over-indexed on large, well-resourced organizations: every case study — Bupa, Jabil, McKinsey's high performers — involves companies with the capital, talent, and organizational complexity to pursue multi-year transformation programs. The practical playbook being articulated may be inaccessible to the vast majority of businesses globally, meaning the main effect of this moment is not democratizing AI value creation but concentrating it further in already-dominant enterprises — a dynamic with significant implications for market structure, competition, and policy that is largely absent from the current conversation.
• Treating AI agents as equivalent to human workers, as NiCE's Chandra proposes, may create dangerous accountability gaps rather than solve governance problems: human workers carry legal, ethical, and professional liability that agents cannot, and frameworks built on workforce equivalence may systematically underestimate the failure modes that emerge when agents operate in ambiguous, high-stakes situations without the contextual judgment humans apply intuitively.