Why are Enterprises still struggling to Harness AI?

Artificial intelligence is no longer new. For at least three years, it has been the centrepiece of global corporate strategy decks, boardroom discussions, and investor narratives. Yet, outside the technology sector, AI adoption has often delivered underwhelming results. Many large enterprises are still running pilot programs, evaluating use-cases and thinking about how to scale. They are struggling to find return on investment and the executive leadership does not want to invest in large projects without a path to ROI.

The AI revolution has been touted as the next industrial revolution. Instead, for the majority or organisations, it has become a cycle of experimentation and hesitation. In this post, I try to find the answers to why this is so and what needs to be done to address this issue. I have used various research reports, white papers, case studies to dig deep on this topic (sources cited at the end) and come up with a strategy recommendation.

The Reality Check: Ambition Outpacing Adoption

Yesterday, I came across an FT article that said Rightmove’s stock fell 15% as it announced huge investments in AI without clear pathways to ROI. According to the 2024 McKinsey Global AI Survey, only 23% of organisations have managed to deploy AI solutions at scale. IBM’s Global AI Adoption Index 2024 reports that while 60% of enterprises have some form of AI initiative underway, only less than one-third are realising business value from it. The gap between ambition and adoption persists, suggesting that technology alone is not the barrier — capability, clarity, and confidence are.

In too many enterprises, AI has been treated as an experimental tool rather than a strategic enabler. Executives approve pilots because the world expects them to, not because they have a coherent AI vision. Without a unifying strategy, projects live and die within departments — never maturing into enterprise programs that transform the business.

A corporate example of this disconnect is Salesforce’s Agentforce. While touted as a next-generation business transformation platform, adoption metrics and stock performance have yet to align with expectations more than one year after the product launch.

Why Enterprises Are Struggling

Behind this struggle lie several structural challenges that go beyond technology itself. Below are the top three reasons that stand out from the various analyst research reports, case studies and market surveys.

1. The absence of a strategic narrative

Most organisations do not have a clearly articulated AI vision. Instead, initiatives are scattered — a chatbot here, a predictive model there. Without a cohesive narrative linking AI to business outcomes, progress remains fragmented.

2. Data debt and legacy integration issues

AI is hungry for data, but enterprise data is often siloed, inconsistent, or trapped in decades-old infrastructure. Integrating these systems demands governance and standardisation — often the least glamorous, yet most essential part of AI readiness.

3. Cultural and capability gaps

AI transformation is not only a technological shift but a human one. A Deloitte 2024 survey found that 70% of executives see “organisational readiness” as a more critical challenge than algorithmic accuracy. The human factor — training, process alignment, and ownership — is often underestimated.

So who is to blame for the misses above? The buck must stop at the executive leadership for these issues. It’s the role of the C-Suite level personnel at the top of the chain to drive these within their organisations when it comes to setting the narrative, addressing systemic/data issues and managing change within the enterprise.

But the story is not entirely grim. Let’s look at some data facts to understand who’s moving better at this.

Sectoral Analysis on Scaled AI Adoption across Enterprise

Based on analyst reports and market data, the following sectors are leading in AI adoption at scale:

The accompanying pie chart illustrates how sectors differ in realising tangible AI benefits.

Source: McKinsey Global AI Survey 2024, IBM Global AI Adoption Index 2024, PwC AI Outlook 2025.

The financial sector remains the leader — partly because data is its native asset and regulatory frameworks have encouraged early experimentation with automation and risk analytics.

• Financial Services use AI extensively in fraud detection, algorithmic trading, and credit modelling. JP Morgan’s AI-driven transaction monitoring now processes 12,000+ data points per second, reducing false positives by 40%.
• Manufacturing has become a laboratory for predictive analytics. AI-driven maintenance, digital twins, and quality control have helped firms like Siemens reduce downtime by 20–25% and improve throughput.
• Retail is seeing a slow but steady transformation — AI in demand forecasting, product recommendation, and dynamic pricing. Walmart, for example, now runs AI models across 300+ data pipelines to optimise pricing and promotions in real time.
• Healthcare and Pharma, though constrained by compliance, are achieving breakthroughs in diagnostics and R&D, with Pfizer and Novartis using AI to cut early-stage drug discovery time by up to 40%.
• Energy & Utilities use AI to anticipate grid failures and optimise production schedules. Shell’s predictive maintenance AI monitors over 10,000 sensors, saving tens of millions annually.
• Logistics firms such as UPS and Maersk are applying AI for routing and container optimisation.

The contrast between these successes and the rest of the enterprise landscape shows that AI success is not about industry type — it’s about intent, clarity, and execution discipline. Next, let’s take a closer look at a couple of the success stories to understand what does good look like when it comes to harnessing AI.

What AI Success Actually Means

Success in AI is not measured by how many models are built, but by how deeply AI capabilities become embedded in business workflows.

Unilever - Applying AI across supply-chain and marketing

• What they did: Unilever has documented dozens/hundreds of AI capabilities across demand forecasting, inventory optimisation, real-time replenishment and marketing personalisation. They moved strongly to cloud-native data platforms and integrated retailer point-of-sale data into forecasting models, enabling near-real-time adjustments.
• Why it works / outputs: Unilever reports fewer stockouts, lower waste in seasonal product lines (e.g., ice-cream), faster content production for marketing, and claims measurable cost savings tied to supply-chain AI. The key enabler: data integration with customers + cloud architecture.

Colgate-Palmolive - Using GenAI for product innovation, individual productivity and knowledge synthesis

• What they did:
• MIT Sloan reports said they Colgate-Palmolive are synthesising consumer insights using GenAI to provide instant answers to business questions.
• The report also found that they could combine one AI system that surfaces unmet consumer needs with another proprietary AI system that develops new product concepts to meet those needs.
• They have setup an AI Hub, which hosts internal versions of OpenAI and Google LLMs as well as image-generating models. Like many enterprises, these tools are deployed internally, allowing employees to leverage proprietary knowledge safely in their AI prompts.
• Why it works / outcomes:
• Reports indicate that Colgate-Palmolive has market research cut from days to hours, accelerating innovation by 30–50%.
• AI fragrance models matched consumer panels with 85–90% accuracy, reducing the 70% failure risk of new products.
• AI-optimised promotions improved sales by 15–20%, while digital ads delivered 79% more unique visitors at lower cost.

Common success factors visible in these case studies:

1. Clear, measurable KPIs (downtime %, fuel or miles saved, inventory turns).
2. Strong data plumbing (sensors, cloud, integrated external data).
3. A durable operating model that embeds AI outputs into daily decision-making (maintenance scheduling, route assignments, replenishment decisions). These are not one-off proofs-of-concept — they are production systems with governance and feedback loops.

Now let’s use the learnings from these case studies to craft a strategy that can scale AI initiatives from pilots to actual organisational programmes that reap benefits.

The Three-Point Play: Converting Pilots into Payoffs

Based on a synthesis of insights from success stories as well as analysts like McKinsey, Deloitte, and PwC’s 2025 AI Outlook, three priorities consistently distinguish organisations that achieve rapid ROI from those that remain stuck in pilot mode.

1. Define AI around problems, not technologies

The most successful AI initiatives begin with a clear business challenge — for instance: “How can we reduce customer churn by 10% this year?” or “How can we cut maintenance downtime by 15%?” Once that problem is defined, you then determine how AI (or automation, or generative models) can serve it. In contrast, starting with “Let’s buy X tool because it’s trendy” often leads to technology chasing use-cases rather than solving them. By anchoring the initiative in a measurable business outcome, you clarify scope, prioritise effort, and set up for value.

2. Build scalable foundations — data, governance, and leadership alignment

AI at scale requires more than a model—it needs clean, governed data and a robust architecture that supports reuse and extension. Think of data not as an IT by-product but as a structured enterprise product: with defined owners, quality controls, metadata, lineage, and integration. At the same time, leadership must reflect this—business lines, technology teams and operations must share accountability for AI outcomes, not leave it to a lone “data science team”. When leadership is aligned, data becomes accessible, processes become consistent, and the path from pilot to production becomes clearer.

3. Govern for scale & speed (operating model + measurement)

Create an AI operating model (centre of excellence + decentralised product teams + funding guardrails) and track ROI continuously. According to Gartner/KPMG, many pilots stall because there’s no productisation, no consistent funding, no measurement. When you combine these three priorities, you significantly increase the chance that a pilot will turn into an enterprise-wide value stream.

Action plan to scale AI

I have worked with many large multi-national / fortune-500 organisations on disparate IT-enabled programmes over the last 20+ years. I have seen many projects succeed as well as fail during this time. The lessons remain the same, even for AI projects in question at the moment. In my view, organisations must do the below correctly so as to realise their AI related ROI goals.

• Define clear, measurable goals and tie project funding to them
• Create an AI operating model (CoE + product teams + funding runway)
• Deloitte’s scaling playbook and multiple guides on AI CoE creation also recommend a central capability that enables business teams, provides MLOps and data platforms, and transitions successful pilots into production services. McKinsey’s “six dimensions” map also supports the central role of operating models.
• Sort out your data plumbing early (data flow across systems, data strategy and architecture, ETL, etc.)
• Data is the enabler for AI, this is well-known fact. Yet, many organisations struggle with sorting out their data. Deloitte and McKinsey recommend prioritising data lifecycle and platform investments; Gartner shows projects fail when data is poor or hard to access. Unilever case study showed that platform/cloud investments preceded broad scaling.
• Embed AI into the flow of work (build hybrid teams: humans + AI agents)
• To reap AI benefits, it is best to embed it into the flow of work and not let it remain a sidebar activity. AI agents should assist, help humans to do things better and even act on there own (within defined boundaries) to add value.
• Measure and iterate continuously - think AI DevOps
• Don’t wait for the best product, the market and industry would have moved on if you wait for months and years to launch an AI based solution. Instead, iterate (and fail) fast, measure, learn and course-correct continuously.
• Prioritise change management (training, user trust, incentives).
• THE most important factor. Drive cultural change in your organisations, human-centric change, lasting behavioural change to embed AI deeply in how the enterprise functions.

Why AI’s Promise Must Go Beyond the Few

Today, most of the economic value from AI accrues to a handful of technology giants — firms that already command the data, compute power, and talent ecosystem (read Google, Nvidia, Microsoft, Apple, Meta and Amazon). For AI to become a true general-purpose technology, it must deliver measurable benefits to every layer of the economy — manufacturing, logistics, services, and even the public sector.

When AI begins generating ROI for smaller and mid-sized firms — when it enhances supply chains, streamlines back-office operations, and helps create inclusive value — that’s when the AI revolution will have truly begun. Until then, it remains a paradox: a technology that promises universality but delivers concentration.

References

• McKinsey & Company. The State of AI: How organisations are rewiring to capture value. March 12, 2025. https://www.mckinsey.com/~/media/mckinsey/business%20functions/quantumblack/our-insights/the-state-of-ai/2025/the-state-of-ai-how-organizations-are-rewiring-to-capture-value_final.pdf
• McKinsey & Company. The State of AI: Global survey. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
• Boston Consulting Group (BCG). AI Adoption in 2024: 74% of companies struggle to achieve and scale value. Press release, October 24, 2024. https://www.bcg.com/press/24october2024-ai-adoption-in-2024-74-of-companies-struggle-to-achieve-and-scale-value
• Boston Consulting Group (BCG). Where’s the Value in AI? (PDF). https://media-publications.bcg.com/BCG-Wheres-the-Value-in-AI.pdf media-publications.bcg.com
• Deloitte. State of Generative AI in the Enterprise. https://www.deloitte.com/uk/en/issues/generative-ai/state-of-generative-ai-in-enterprise.html
• C3.ai / Shell. How Shell scaled AI predictive maintenance to monitor 10,000 pieces of equipment globally. https://c3.ai/blog/how-shell-scaled-ai-predictive-maintenance-to-monitor-10000-pieces-of-equipment-globally/ hai-production.s3.amazonaws.com
• Unilever. How AI is transforming Unilever Ice Cream’s end-to-end supply chain. 2025. https://www.unilever.com/news/news-search/2025/how-ai-is-transforming-unilever-ice-creams-end-to-end-supply-chain/
• Accenture & Unilever. Unilever and Accenture join forces to establish a new industry standard in generative-AI powered productivity. 2024. https://newsroom.accenture.com/news/2024/unilever-and-accenture-join-forces-to-establish-a-new-industry-standard-in-generative-ai-powered-productivity.htm
• Financial Times. Article: “AI’s awfully exciting until companies want to use it” (covers Jefferies/Gartner findings, Rightmove example). https://www.ft.com/content/74e31d3e-4b50-43b2-9aa2-e53f41b776a8