Reasons Large Firms Miss AI Adoption Trends

This week at Fortune Brainstorm Tech, I sat down with leaders actually responsible for implementing AI at scale - Deloitte, Blackstone, Amex, Nike, Salesforce, and more. The headlines on AI adoption are usually surveys or arm-wavy anecdotes. The reality is far messier, far more technical, and - if you dig into details - full of patterns worth stealing. A few that stood out: (1) Problem > Platform AI adoption stalls when it’s framed as “we need more AI.” It works when scoped to a bounded business problem with measurable P&L impact. Deloitte's CTO admitted their first wave fizzled until they reframed around ROI-tied use cases. ➡️ Anchor every AI proposal in the metric you’ll move - not the model you’ll use. (2) Fix the Plumbing Every failed rollout traced back to weak foundations. American Express launched a knowledge assistant that collapsed under messy data - forcing a rebuild of their data layer. Painful, but it created cover to invest in infrastructure that lacked a flashy ROI. Today, thousands of travel counselors across 19 markets use AI daily - possible only because of that reset. ➡️ Treat data foundations as first-class citizens. If you’re still deferring middleware spend, AI will expose that gap brutally. (3) Centralize Governance, Decentralize Application Nike’s journey is a case study: Phase 1: centralized team → clean infra, no traction. Phase 2: federated into business-line teams → every project tied to outcomes → traction unlocked. The pattern is consistent: centralize standards, infra, and security; decentralize use-case development. If you only push from the top, you have a fast start but shallow impact. Only bottom-up ownership gives depth. ➡️ You can’t scale AI from a lab. It has to live where the business pain lives. (4) Humans are harder than the Tech Leaders agreed: the “AI story” is really a people story. Fear of job loss slows adoption. ➡️ Frame AI as augmentation, not replacement. Culture change is the real rollout plan. (5) Board Buy-In: Blessing and Burden Boards are terrified of being left behind. Upside: funding and prioritization. Downside: unrealistic timelines and a “go faster” drumbeat. Leaders who navigated best used board energy to unlock investment in cross-functional data/security initiatives. ➡️ Harness board FOMO as cover to fund the unsexy essentials. Don’t let it push you into AI theater. (6) Success ≠ Moonshot, Failure ≠ Fatal. - Blackstone's biggest win: micro-apps that save investors 1–2 hours/day. Not glamorous, but high ROI. - Nike's biggest miss: an immersive AI Olympic shoe designer - fun demo, no scale. Incremental productivity gains compound. Moonshots inspire headlines, but rarely deliver durable value. ➡️ Bank small wins. They build credibility and capacity for bigger bets. In enterprise AI, the model is the easy part. The hard part - and the difference between demo and value - is framing the right problem, building the data plumbing, designing the org, and bringing people along.

AI adoption slows inside large organizations because leaders mistake ambition for readiness, which wastes capital and weakens trust. Running multiple AI initiatives at the same time raises coordination effort and slows decisions, which reduces real business output. Teams are expected to change how they work while still hitting short-term targets leading to quiet failure. Leaders who overlook this end up with sunk costs and slower core operations. What is changing is that AI now requires sustained operational focus rather than short pilots because models affect workflows across the organization. This is happening because AI tools evolve faster than companies can adapt their ways of working. The result is fragmented adoption and falling returns when too many initiatives compete for attention. This means: fewer initiatives executed deeply deliver more value. Two takeaways for legacy business leaders Parallel AI rollouts fail because organizations can only absorb one major workflow change at a time. This creates hidden capacity limits that stall execution. Strong vendors do not fix weak integration because success depends on how work actually happens, not on the quality of the tool. Start with one revenue-critical or cost-critical workflow and map decision handoffs before introducing AI, because clarity lowers the risk of failure. Avoid deploying AI where ownership is unclear or metrics are disputed, because automation amplifies confusion. Using AI for appearances destroys value, while operating with AI requires focus and discipline. #AIAdoption #AIReadiness #EnterpriseAI #OperatingModel #LeadershipExecution #DigitalStrategy #ChangeCapacity #WorkflowDesign #AIGovernance #BusinessOutcomes #AIReality #FoundersPerspective

One in three companies are planning to invest at least $25m in AI this year, but only a quarter are seeing ROI so far. Why? I recently sat down with Megan Poinski at Forbes to discuss Boston Consulting Group (BCG)'s AI Radar reporting, our findings, and my POV.   Key takeaways below for those in a hurry. ;-)   1. Most of the companies have a data science team, a data engineering team, a center of excellence for automation, and an IT team; yet they’re not unlocking the value for three reasons:   a. For many execs, the technologies that exist today weren't around during their school years 20 years ago. As silly as it is, but there was no iPhone and for sure no AI at scale deployed at people’s fingertips.   b. It's not in the DNA of a lot of teams to rethink the processes around AI technologies, so the muscle has never really been built. This needs to be addressed and fast...   c. A lot of companies have got used to 2-3% continuous improvement on an annual basis on efficiency and productivity. Now 20-50% is expected and required to drive big changes. 2. The 10-20-70 approach to AI deployment is crucial. Building new and refining existing algorithms is 10% of the effort, 20% is making sure the right data is in the right place at the right time and that underlying infrastructure is right. And 70% of the effort goes into rethinking and then changing the workflows. 3. The most successful companies approach AI and tech with a clear focus. Instead of getting stuck on finer details, they zero in on friction points and how to create an edge. They prioritize fewer, higher-impact use cases, treating them as long-term workflow transformations rather than short-term pilots. Concentrating on core business processes is where the most value lies in moving quickly to redesign workflows end-to-end and align incentives to drive real change.   4. The biggest barrier to AI adoption isn’t incompetence; it’s organizational silos and no clear mandate to drive change and own outcomes. Too often, data science teams build AI tools in isolation, without the influence to make an impact. When the tools reach the front lines, they go unused because business incentives haven’t changed. Successful companies break this cycle by embedding business leaders, data scientists, and tech teams into cross-functional squads with the authority to rethink workflows and incentives. They create regular forums for collaboration, make progress visible to leadership, and ensure AI adoption is actively managed not just expected to happen.

Why #AI is not scaling in Enterprises? Do you agree ? AI is struggling to scale primarily due to diminishing returns from ever-larger models, data quality bottlenecks, infrastructural complexity, skills shortages, and organizational challenges such as lack of leadership alignment and change management. While early pilots often succeed, turning these into sustainable, enterprise-wide AI deployment faces significant systemic hurdles. The Core Reasons AI Is Not Scaling : 1. The Parameter Plateau and Diminishing Returns Modern AI has benefited from rapid scale, moving to models with billions (or trillions) of parameters. However, each successive jump delivers smaller improvements at exponentially higher costs, both computationally and environmentally. The shift now is toward smaller, smarter models using cleaner, highly curated data rather than just bigger architectures. 2. Data Quality, Pipelines, and Integration Poorly structured or inconsistent data, fragile data pipelines, and lack of data governance significantly impede the scaling of AI beyond proof-of-concept. Clean, well-labeled, domain-specific data is far more effective for training robust enterprise AI than massive volumes of uncurated information. Legacy application modernisation is essential. 3. Human Capital and Organizational Resistance There is a global scarcity of skilled AI talent able to build, deploy, and maintain large-scale AI systems. Additionally, resistance to change from within organizations—even at the leadership level—hampers efforts to operationalize AI, especially if the business value is not clearly demonstrated. 4. Operational and Technological major hurdles include: - High computational and storage costs as workloads expand - Model drift, requiring constant retraining and updates - Fragmented tech stacks leading to integration challenges - Compliance and security risks, especially with sensitive data. 5. Lack of Alignment with #Business Objectives Many AI projects fail to scale because they’re not tightly aligned with measurable business goals. This makes it difficult to demonstrate ROI or drive organizational buy-in, which is essential for enterprise-wide adoption. Keys to overcoming the "Scaling Barrier" -Focus on data quality and domain curation, not just #model size. -Build robust, standardized data pipelines and governance to avoid bottlenecks -Develop AI centers of excellence and invest in talent, fostering AI literacy organization-wide - Align AI initiatives with clear #business value, objectives, and ongoing executive sponsorship. - Invest in adaptable infrastructure and continuous model improvement processes to manage drift, compliance, and cost. Scaling AI is as much an organizational transformation as a technological one, requiring coordinated investment in people, processes, and data—not just algorithms and and compute power. It the same cycle we will need to adopt like the #Digital #Transformation

Big consulting firms rushing to AI...do better. In the rapidly evolving world of AI, far too many enterprises are trusting the advice of large consulting firms, only to find themselves lagging behind or failing outright. As someone who has worked closely with organizations navigating the AI landscape, I see these pitfalls repeatedly—and they’re well documented by recent research. Here is the data: 1. High Failure Rates From Consultant-Led AI Initiatives A combination of Gartner and Boston Consulting Group (BCG) data demonstrates that over 70% of AI projects underperform or fail. The finger often points to poor-fit recommendations from consulting giants who may not understand the client’s unique context, pushing generic strategies that don’t translate into real business value. 2. One-Size-Fits-All Solutions Limit True Value Boston Consulting Group (BCG) found that 74% of companies using large consulting firms for AI encounter trouble when trying to scale beyond the pilot phase. These struggles are often linked to consulting approaches that rely on industry “best practices” or templated frameworks, rather than deeply integrating into an enterprise’s specific workflows and data realities. 3. Lost ROI and Siloed Progress Research from BCG shows that organizations leaning too heavily on consultant-driven AI roadmaps are less likely to see genuine returns on their investment. Many never move beyond flashy proof-of-concepts to meaningful, organization-wide transformation. 4. Inadequate Focus on Data Integration and Governance Surveys like Deloitte’s State of AI consistently highlight data integration and governance as major stumbling blocks. Despite sizable investments and consulting-led efforts, enterprises frequently face the same roadblocks because critical foundational work gets overshadowed by a rush to achieve headline results. 5. The Minority Enjoy the Major Gains MIT Sloan School of Management reported that just 10% of heavy AI spenders actually achieve significant business benefits—and most of these are not blindly following external advisors. Instead, their success stems from strong internal expertise and a tailored approach that fits their specific challenges and goals.

AI in Telco Won’t Scale if Legacy Stays Telcos continue to announce AI transformation roadmaps. From GenAI in customer service to AI-RAN and self-optimizing networks, the ambitions are clear. Yet across the industry, most of these initiatives remain trapped in pilot mode. The reason is not model maturity or lack of talent. It is legacy infrastructure. A recent survey by Fierce Telecom found that 32% of operators cite legacy systems as the primary barrier to AI adoption. In parallel, Accenture reports that 66% of service providers identify technical debt as the top constraint to modernization. Over half of IT Telco teams spend more than 800 hours annually maintaining aging platforms. That is time diverted from deploying automated pipelines, training models, or integrating intelligent agents into production systems. Legacy showstoppers are happening every day. In 2024, a large Telco group partnered with a top vendor to implement its cognitive SON platform. The objective was to use AI to optimize power consumption, reduce interference, and improve network efficiency by up to 30%. But the project initially failed to scale. The AI system required real-time telemetry, dynamic network configuration access, and external data streams such as energy pricing. Core telemetry data was locked inside proprietary EMS platforms that did not support open interfaces. External data integration was blocked by outdated middleware layers. Configuration workflows still require manual validation due to rigid OSS processes. The model was fully functional, but the infrastructure was not. Only after the Telco replaced key legacy OSS components and re-engineered its data architecture did the AI deployment deliver measurable impact. Across the telecom industry, legacy systems dominate BSS, OSS, provisioning, and assurance layers. These platforms were not designed to support AI inference, real-time feedback loops, or autonomous operations. They were built to enforce transactional integrity, compliance, and control. As a result, they constrain AI deployments in both speed and scope. Enterprise-wide benchmarks reinforce this structural problem. 64% of large organizations still run over a quarter of their operations on legacy systems. In telecom, that percentage is likely higher and far more critical to daily network functionality. AI in telecom cannot scale on infrastructure that was never meant to support it. Until the underlying systems are modernized, even the best-designed models will remain boxed into isolated pilots. The path forward is not just about choosing the right algorithms. It begins with the architectural will to replace what no longer supports execution.

By the time someone says “our people aren’t ready for AI,” several things have usually already gone wrong. That framing sounds like a diagnosis, but more often it’s a way to stop asking questions. When you look more closely, the same patterns tend to appear. 1) Unclear boundaries Without clear guardrails, people fill the gaps with assumptions. Uncertainty creates hesitation, and hesitation gets labeled as resistance. 2) Decisions without users When AI is imposed rather than shaped with the people doing the work, disengagement becomes a form of self-protection. 3) Dismissed concerns When valid questions get shut down, people stop raising them. The issues don’t disappear, they just go underground. 4) Tool-first training A demo is not adoption. Without reinforcement, coaching, and connection to daily work, people default back to what feels safer and familiar. 5) Leadership misalignment Mixed messages create paralysis. When priorities shift week to week, waiting becomes the most rational choice. 6) Missing “why” People can adapt to significant change when the reason makes sense to them. Without context, AI initiatives feel arbitrary and disconnected. 7) Feedback with no follow-through Asking for input and doing nothing with it teaches people that their voice has no impact. They remember that lesson. 8) Reality gap AI strategies built in conference rooms often miss how work actually happens. That gap is where adoption quietly breaks down. 9) Readiness assumed Labeling people as resistant ends the conversation. Understanding what’s actually happening requires curiosity many organizations skip. When AI adoption stalls, it’s often a signal, not a flaw. A signal to look at how the change was led, not who is being blamed. What would you add? ——— ✦ ——— More on AI + Workforce Transformation → Janet Perez

A year ago, Marc Benioff stood in front of 45,000 Dreamforce attendees and said automating with AI was “easy and quick.” He even hinted that Salesforce might not need to hire more software engineers because AI agents could handle the work. Fast forward to Dreamforce 2025, the message is different. On the same stage, he admitted adoption “takes time” and that AI innovation is moving faster than customer readiness to deploy it in production. Salesforce even had to bring back a technical team to help customers adopt AI - a team Benioff had previously cut when the hype suggested it wouldn’t be needed. The technology itself isn’t the problem. Models are more capable than ever, and innovation hasn’t slowed down. The challenge is what happens after the demo. Large companies can’t just plug AI into their systems overnight. They need to restructure data, rewire architecture, and most importantly - build trust. Enterprises move cautiously for a reason. Without security and safety guardrails, robust testing, clear governance, and alignment to business goals, AI can’t be scaled responsibly. That’s why adoption is lagging: not because the tools aren’t powerful, but because the foundations of security, reliability, and oversight are still catching up. The models have demonstrated their capabilities. What’s missing is the trust, safety, and alignment that let businesses feel confident putting AI at the heart of their operations. Until those pieces are solved, adoption will never match the pace of innovation.

A directive came from leadership: “Everyone is now using AI.” Many in the company didn't change a thing. That's what happens when executives mandate AI tools instead of earning adoption. McKinsey has tracked transformation failures for decades. The conclusion hasn't changed: 70% of change initiatives fail. Employee resistance is the most common reason. Yet when tool adoption stalls, most executive responses look exactly the same: → write a stronger memo → set a harder deadline → make it mandatory The strategy is the same one that's been failing for decades. Here's what kills AI tool adoption and what works: 1/ You Created Compliance, Not Capability People use the tool enough to avoid trouble. They never use it enough to create value. Adoption metrics look fine. ROI never materializes. Reality: Gartner found only 38% of employees are willing to support organizational change today, down from 74% in 2016. 2/ You Told Them What, Not Why "We're moving to [tool]" is not a change strategy. People fill the vacuum with fear. Anxiety drives avoidance. Reality: Organizations are 3.5x more likely to succeed when leaders communicate outcomes before launching solutions. 3/ You skipped the pilot Full-organization rollouts collapse under the weight of edge cases. No feedback loop before the stakes are high. A loud failure poisons future adoption efforts. Reality: Companies are 3x more likely to succeed when they use piloting and prototyping to identify skill gaps. 4/ You chose champions, not skeptics Early adopters sell it to true believers. Skeptics aren’t persuaded, they’re overruled. The mandate hardens their resistance. Reality: Your critics, converted, become your most credible advocates. 5/ Training was a one-day event A 90-minute demo is not adoption. People default to old habits under pressure. New tools feel harder until they feel natural. Reality: Embedded coaching and workflow integration matter more than onboarding sessions. 6/ Middle management was left out Executives announce it. Managers determine if it actually happens. Adoption dies in the layer between the announcement and the work. Reality: McKinsey attributes 72% of transformation failures to inadequate management support. 7/ You measured deployment, not depth “90% of users logged in” is not adoption. Reality: The difference between a technology investment and a technology return is depth of use. What works instead: → Identify champions from mid-level teams, not just the executive floor → Tie tool use to outcomes people already care about → Measure depth of use, not just activation rate → Give managers a how, not just a what → Anchor adoption to performance reviews Zapier drove 97% company-wide AI adoption through hackathons & peer show-and-tells, not mandates. That gap between pilot and production is a people problem executives keep trying to solve with authority. Authority creates compliance. Engagement creates adoption. Save for future reference.

This week I’ve both ordered groceries with OpenAI's newly released AI agent Operator for the first time, and met 4 senior transformation leaders who are struggling to get basic Gen AI adoption in their respective organizations. The gap between AI development and human readiness is getting too large, too fast. This will be - is already starting to become - the greatest challenge change management professionals have ever faced. From my conversations with change and transformation leaders, a number of things stand in the way of realizing the grand visions .ai founders in Silicon Valley are spinning: 1. There’s a widespread general consensus that Microsoft’s Copilot (the most implemented AI POC of 2024) is not as good as people thought it would be, and that it’s not meeting ROI expectations on productivity at its current price point. 2. Gaining real workplace adoption of AI requires a “one step back, two steps forward” approach, which no one (employees nor leaders) is willing to take. For most, it’s still easier to just do it themselves when they’ve done it for 20 years, than learning how to ask an AI for support. 3. Most introductions of AI in enterprises kick off with extremely dull compliance and security training, which does a great job of dampening any excitement that existed about the topic in the first place. 4. The most senior executives in large enterprises are often the slowest adopters themselves, hosting grand town hall speeches about the need for operational efficiencies while keeping their own EAs. Total failure in leading from the front. 5. Generic AI adoption training doesn’t work. It needs to be use case specific, function specific, and much more personalized to unique user groups and their everyday tasks than the one-size-fits-none videos being blasted out about how to prompt engineer a pasta recipe in ChatGPT. But the absolutely biggest miss of all, in my own view, across the AI transformation programs I’m witnessing: The WHY is missing. People are asking their leaders: - Why should we adopt AI? - Save time and become more productive? Got it - and what will you do with that extra time I get? I don’t think you want me to spend more time with my family. - So that means I will work more - until the day I ultimately don’t work at all? And then what happens to me and my family? Humans are wired for storytelling, and the WHY in AI transformation stories does not speak to individuals at all. It speaks to Wall Street’s operating margin goals, not people’s life goals. And just like Wall Street don’t care about people, people don’t care about Wall Street. Until the WHY and the story changes, AI transformations will continue to struggle. Tomorrow I’m Zooming my grandmother to show her Operator and ask for her thoughts on all of this. I’ll be back here with our call notes.