Good morning. I've watched companies dump cash into AI last year, and that EY survey feels spot on—nearly all the big ones took losses, billions gone. 1. Healthcare keeps losing out, like with IBM Watson's mess, and education too, from too many pilots that don't match their data. Finance and manufacturing pull ahead, though, with 20-30% gains from simple automation that actually works. 2. CEOs aren't thrilled because those $1M-plus GenAI spends go mostly to patching problems, not profits—pilots never turn into daily workhorses. 3. Winners nail clear goals, clean data, and vendor teams over in-house builds; losers chase hype without linking to real jobs. Firms picking ready-made stuff double their wins compared to custom jobs. 4. The disappointment hits more from businesses skipping training and rollout than vendors talking big. Leaders forget to weave AI into everyday tasks, and it flops.
Faculty Member at The University of Texas at Austin McCombs School of Business
Answered 2 months ago
Question 2.) RESPONSE: "Many leaders are under pressure from boards to "do AI," which creates a rush to launch projects without discipline. I've seen technology companies mandate AI usage, so employees comply by using it to summarize meeting notes they don't actually need. It's Goodhart's Law in action. The metric becomes "AI adoption" rather than value creation. Meanwhile, real value is being created but not captured. Many organizations treat AI like an ERP implementation—a top-down, enterprise-wide system. But immediate AI ROI is bottom-up: individuals finding efficiencies in daily work. Most companies haven't figured out how to measure and capture that value back to the P&L." Questions 1.) and 3.) RESPONSE: "It's not industry-specific. Winners have discipline: they start with strategy, systematically identify high-value opportunities with a clear view of constraints, and have the cross-functional culture needed to execute. They also have foundational advantages, like strong master data governance, coherent system landscapes, well-designed workflows. Organizations without this infrastructure are trying to build AI on quicksand." Question 4.) RESPONSE: "Both, but here's what's interesting: overnight, every business software company became "AI-powered." There's a joke that every system now has an "AI button," which is usually just an LLM that summarizes text. But software overhype isn't new. What IS new is how many leaders lack the digital fluency to separate signal from noise. I've seen legacy automation projects relabeled as "AI" to secure funding or claim wins. Organizations are taking a haphazard approach by releasing tools without strategy, launching projects without assessing data quality or workflow readiness."
Q1. Industries that have existed for a long time, such as healthcare and manufacturers of old-fashioned products, suffer the most significant financial loss as compared with newer businesses using new technology. There are usually two main obstacles to emerging technologies making headway in these older industries; first, the technology itself, second, the disjointed nature of data and the risk factors associated with errors from technology require a large amount of manpower for oversight. This latter point creates significant margin erosion due to personnel costs associated with overhead from having to provide oversight. Conversely, newer businesses, such as e-commerce and customer service centers, are typically reaping their technology investments much faster than the aforementioned companies; e-commerce and customer service have greater data liquidity, much lower cost of failure, and therefore, will return their R.O.I. within weeks. Q2. The CIOs are not getting the R.O.I. they expected from AI technology because they are treating AI like any other software purchase and not fundamentally restructuring their underlying business processes to accommodate artificial intelligence. If you install a multi-million dollar "general AI" solution and your underlying business processes aren't re-engineered, you have wasted your money. The time saved by employees when using AI is typically consumed elsewhere as opposed to producing additional growth. Q3. Companies that have been successful at implementing AI technology typically focus on one high-frequency, high-friction issue to solve, as opposed to trying to address all problems within their organization. A good example is a logistics company we worked with that completely ignored the hype and focus on a comprehensive AI strategy. Instead, this organization focused solely on automating the document extraction process for the customs clearance process and was able to reduce processing times by 70% within a few weeks. Companies that attempt to build a "general-purpose assistant" that addresses all issues the same way are typically losing companies. Q4. The majority of the issues preventing companies from realizing the full value from their AI technology lies in the gap between these two experiences, as opposed to the "potential value" of any given AI product.
While building and launching several consumer- facing platforms, I have learned on the ground for companies running to deploy AI without clear use cases are burning through capital at the fastest clip. The industries that are showing positive return on investment are those where AI is being used to solve specific, measurable problems—fintech companies using AI for fraud detection or e-commerce firms improving recommendation engines. The frustration has a lot to do with companies and organizations seeing AI as something magical rather than as something that needs to be embedded strategically into existing workflows. In my product work, I've seen that successful AI implementations have an initial small scope driven by a clear target and scale is gradual—failures usually are made up of large upfront investments into broad AI initiatives with no way to measure success.
Founder, SME Business Investor, Property & Finance Specialist at Zanda Wealth
Answered 2 months ago
What I found is that many technology firms are losing the most because they are using AI like a shiny new toy rather than a tool for a specific job. In my line of work, I see firms dump millions into custom models when some kind of API would have worked just fine. It is like a whole car factory when all you need is a ride to the store. The true winners at this time are in high volume logistics and supply chain management. We assisted one of our clients in the area who initially saw results after only the first month of implementation using AI only to predict warehouse churn. They didn't try to "reimagine" their business but just fix one expensive leak. Speaking of that, I'd like to explain why most CEOs are not happy. Based on my years in the field, the 1 million dollar investment usually fails because it does not have a clear "exit" strategy for the old, manual processes. Companies add the cost of AI, but they don't actually remove the labor or the legacy software that the AI was meant to replace. That's why the ROI remains negative; you are just paying for 2 different ways to do the same task. I've seen that "disappointment" is rarely from being misled by vendors. From what I've seen, the problem is that businesses don't think about AI as a total change in the way staff work, but more like a software update. You can't just plug in a transformer and expect magic without rewriting your internal playbooks. If you are a busy executive, you know that the tool is only as good as the person who is holding it.
In my experience, the industry isn't what determines whether companies lose money on AI. It's their approach to AI. Most organizations treat generative AI like a traditional tech rollout, but they overlook the most important element: their people. You can buy the best AI tools in the world, but if employees aren't equipped and inspired to use them to reimagine their workflows, ROI never materializes. Companies in the technology industry can lose money because their culture may be tool first vs. human first, while organizations in less "AI mature" industries may see a return on their investment simply because they prioritize adoption, behavior change, and workforce readiness over AI capabilities. In my experience, AI success is more about the company's internal culture than their industry. The early winners are companies who engage their entire organization. They don't launch generic "use AI" initiatives. They focus on the value users can get out of using AI and give teams relatable, highly specific role-based guidance. For example, they will show how marketers use AI to draft campaigns; how analysts use it to model data; how sales teams use it to prepare for an upcoming sales call; and how Human Resource teams use it to build onboarding materials. Additionally, they don't expect IT to drive organization-wide adoption. AI can be used in so many ways, and IT won't understand the nuances of everyday work. So instead, successful organizations empower their champions, those users who are on the ground, actively using AI to reimagine their work, to tell their story to their peers. They spotlight real wins happening in real workflows, and they engage their leaders to model AI use themselves. That's what builds momentum and measurable ROI. Because when employees learn from peers who are using AI on actual workflows, adoption skyrockets because the use cases suddenly feel real, relevant, trustworthy and doable. And when leaders celebrate these small wins and create a culture of experimentation, organizations unlock the full value of AI and AI becomes part of their DNA, no matter what industry they're in.
1. Which industries are seeing losses vs. ROI? AI succeeds or fails based on data quality. Industries like healthcare and government often have trouble with return on investment (ROI). This is mainly because they don't have clean, organized data. Without good data, AI models can give unreliable results. On the other hand, e-commerce and fintech are winning because they operate in "measurable ecosystems." These fields hold vast user data and quick feedback loops. This setup makes it easy to see how AI helps with goals like preventing fraud or boosting sales. 2. Why are so many CEOs dissatisfied with ROI? The problem is that many leaders bought AI tools without first defining a specific business problem to solve or a way to track success. Simply buying an AI model isn't the same as fixing how a company works. Disappointment usually comes from unrealistic expectations and the high cost of setting the tech up. If a company can't connect an AI tool to clear results, it feels like a failed experiment. For example, knowing the cost of gaining a new customer is crucial. Without this, the investment seems less like a success. 3. What separates early winners from losers? The main difference is discipline. Winners start small: they test AI on one specific, narrow problem, measure the results, and only grow the project after they've proven it works. Losers often try to use AI everywhere at once without any safety rails. For example, using AI to chase "clicks" might get more traffic, but it can hurt a company's profits in the long run. AI only becomes valuable when it is tied to a specific way to make money and includes human oversight. 4. Is disappointment due to overselling or poor implementation? While AI is often marketed as a "miracle" fix, the real issue is organizational readiness. Think of AI as an amplifier: it makes a company's strengths better, but it makes its weaknesses worse. If a business already has messy data or doesn't know how to measure success, AI will only make those problems more complicated. The companies winning with AI treat it as a system that needs constant human improvement, not as a shortcut to success.
In healthcare, I've seen AI projects succeed and fail. Usually the problem isn't the AI itself, but companies trying to jam it into a messy system while dealing with regulations. At Superpower, when we just focused on one thing, like early biomarker detection, we saw real results. The broad, vague projects just burned cash. Honestly, most AI disappointment comes from companies not being ready, not from the product claims. If you have any questions, feel free to reach out to my personal email
1 / From what I've seen, industries like healthcare and higher ed often struggle more with AI ROI. A hospital system leader told me they sunk six figures into an AI chatbot for patient support, but it broke down quickly under real-world medical questions. On the flip side, ecommerce and logistics seem to be early winners--things like product recommendation engines or routing algorithms have clearer outcomes and faster validation. 2 / One issue is that executives buy into AI with the wrong expectations--it's not a plug-and-play fix. A friend in automotive invested in GenAI tools for marketing, but the real bottleneck was that no one in-house knew how to prompt or interpret the results. So, they had flashier tech but no improvement. 3 / The winners usually start small and integrate AI into a workflow they already understand. A business owner I met at a hospitality conference used AI to analyze guest feedback and shift their social posting strategy. It didn't cost much, but because it was connected to something they were already measuring--engagement and bookings--it felt like an instant win. 4 / It's both, I think. Vendors often oversell "magic" solutions, while buyers underestimate how much clarity and training is needed on their end. AI isn't a microwave: it's more like hiring a junior analyst--you need to guide, train, and review or it just adds noise.
1 / I've seen healthcare, retail, and education struggle the most with AI ROI. These sectors tend to adopt AI enthusiastically without aligned operational systems or clean data to support implementation. For example, in healthcare, deploying AI to predict patient outcomes or support diagnostics can backfire if electronic health records are inconsistent or staff aren't trained to interpret AI output. On the flip side, industries like logistics, finance, and precision manufacturing are seeing stronger returns because they're built on structured data and measurable outcomes--areas where AI thrives. 2 / One major reason for poor ROI is misalignment between expectations and actual solutions. Many businesses treat GenAI as a big-ticket innovation rather than a tool requiring integration, iteration, and team buy-in. I've seen this firsthand when companies invest heavily in integrated chatbots or content engines, only to find they don't meaningfully reduce workloads or improve output. Without concrete use cases and measurable KPIs tied to cost savings or productivity, the million-dollar spend yields little tangible return. 3 / Clear needs and structured data make all the difference. We've seen success when AI is applied to contained, measurable tasks--like using NLP models to streamline customer support analysis or applying generative tools to optimize product descriptions. Those efforts produce real time and labor savings. Early losers often leap straight to high-visibility use cases without assessing how AI plugs into their workflows. One example I recall is a brand trying to automate ingredient R&D with AI before securing reliable, standardized formulation data--so it never got off the ground. 4 / It's mostly a misuse issue. AI isn't underdelivering in capability--it's that too many companies lack the internal frameworks, training, or patience to deploy it effectively. We've had to rescope entire tech projects because the tools were solid, but the teams using them didn't understand how to feed or interpret the outputs. So yes, sometimes it's overhyped, but more often it's under-supported. Strong AI outcomes require strong change management.
Which is losing vs. seeing financial returns: In our experience working with companies on their AI readiness initiatives, those struggling to succeed chase them as shiny objects without understanding their strengths and core limitations. While AI has made creating software easier, the fundamentals of implementing commercially viable initiatives have not changed. The ones that are successful invest time in readiness, building target operating models, and assessing financial feasibility before developing, implementing, or committing to long-term obligations. With the average investment in GenAI incentives exceeding $1 million per business, fewer than 30% of CEOs say they're happy with their ROI. In our experience, this is consistent with most technology implementations, especially those driven by technical teams without assessing commercial viability. The aim of these initiatives is to test the promise of technology, especially by assessing the commercial feasibility of inflated promises. The reality is obviously different from what's promoted in vendors' marketing materials. This rate would improve once the understanding matures and Gen AI technologies move from the experimentation phase to the productization phase. Factors that set apart early winners are those with a clear business case and a proven history in a very similar context. In one engagement, a mid-market non-profit wanted to implement AI to solve all of its current process problems. They were told that AI can work with data and processes in any form. They didn't need to change. They could successfully develop the technical models, but unfortunately, business users didn't feel comfortable replacing their current workflows and systems, as they knew that, while the technology looked promising, it wouldn't produce the same outputs as their current systems. AI disappointments stem from misleading claims, overly hyped marketing, untrained users, and unreasonable expectations of AI. If they follow the ground rules of technology implementations, they can create unbeatable business models with unfair competitive advantage. Bio: Sam Gupta brings more than two decades of expertise advising executives on business, AI, ERP, and technology transformation, and has been frequently cited by major business and industry media, including the BBC, TechTarget, and Yahoo Finance.
AI losses tend to happen in heavily regulated industries where technology is layered on top of complex systems without tightening the basics first. In sectors like financial services and healthcare, AI cannot shortcut compliance, identity checks, transaction monitoring, or reporting. When companies expect AI to replace controls instead of support them, errors create rework, legal exposure, and operational slowdowns. The ROI gap usually comes from unclear use cases. Many businesses invest in GenAI tools before tightening workflows. If the underlying process is messy, AI just scales the mess. Spending over $1 million without defining what success looks like almost guarantees disappointment. In our case at Swapped, operating across 150+ countries with 24/7 live support, we use AI narrowly. It helps sort, tag, and route support tickets and surface patterns across regions. It does not approve transactions or override compliance checks. That keeps decisions tied to real money human-controlled while reducing manual noise. The companies seeing returns are the ones using AI to reduce repeat work inside stable workflows. The ones losing money often expect AI to fix messy processes or speed up regulated decisions that still require oversight. Most disappointment isn't about AI being fake. It comes from overselling and poor implementation combined. AI works best when paired with clear boundaries, strong controls, and human oversight. It fails when it's treated as a shortcut.
In finance and fintech, AI only works when you put it into tools customers actually use, like calculators or risk assessors. I've seen some banks pull ahead, but plenty of companies struggle, especially when the value to the customer isn't obvious. Our own AI tools flopped at first, but once we changed the rollout and explained what it did, people started actually using it. It's not about the hype, it's about having a clear plan to fit AI into day-to-day work. If you have any questions, feel free to reach out to my personal email
I've noticed manufacturing and logistics struggle with AI returns while fintech does better, mostly because their data is cleaner. The biggest trap is when companies jump into generative AI without a clear goal. The costs get out of control fast. Honestly, the companies that succeed start small with pilot programs and actually spend time training their teams. It's more about changing how people work than just the technology itself. If you have any questions, feel free to reach out to my personal email
From what I see in commercial real estate, companies waste money on AI when systems don't connect or staff aren't trained. We had property evaluation tools that promised efficiency but just added manual steps, raising our costs. I'd tell any company to start with one small project. Make sure the tech actually does the job before you roll it out everywhere. If you have any questions, feel free to reach out to my personal email
1 / I've seen some retail and fashion brands rush into AI expecting it to "automate creativity" -- and that rarely ends well. You can't fast-track emotional connection through a model. On the flip side, logistics-heavy industries like shipping or supply chain get clearer returns, because AI handles patterns and predictions better than feelings. 2 / Many brands pour money into AI without deeply understanding what problem they're solving. It becomes about trend-chasing, not purpose. ROI suffers when leadership treats AI like magic instead of a tool that still needs human clarity, data hygiene, and heart-centered strategy. 3 / The winners bring AI into harmony with human experience -- not in place of it. I've seen brands that use AI to better understand customer sentiment across language barriers, then shape design from those insights. But if you just want AI to replace people and slash costs, you won't build trust or create beauty -- just noise. 4 / It's both. AI gets sold as a miracle, then handed to teams who don't know how (or why) to integrate it. But also, I've sat in rooms where executives spent millions on AI tools without involving creatives, customer service, or even marketing. That disconnect is the real cost.
In construction, I see companies lose money on AI when they buy the tool first and figure out the problem later, or try to jam it into their old, clunky systems. We've actually made money applying it to the boring but necessary stuff, like predicting when a machine will break or tracking job progress. My advice? Start small, solve one real problem, and make sure your team actually knows how to use the new thing. If you have any questions, feel free to reach out to my personal email
President & CEO at Performance One Data Solutions (Division of Ross Group Inc)
Answered 2 months ago
You know, I see companies burn money on AI because they automate tasks that aren't their real problem. One client spent a lot automating basic customer support emails, but their actual issue was that their data was a mess. It didn't pay off at all. When they switched to using AI to fix the data, things stabilized. It usually comes down to just focusing on the wrong thing. If you have any questions, feel free to reach out to my personal email
At Tutorbase, I see schools and hospitals have a tougher time with AI. They're tangled in regulations and stuck with old computer systems. Tech and finance companies can test things fast and see what's making money. I think the problem is we expect too much too soon. You have to start small, train your people, and treat AI as a long-term play, not a quick fix. If you have any questions, feel free to reach out to my personal email
Professional services firms are bleeding cash on AI because they're automating the wrong things. I watched a consulting shop buy tools that cut proposal writing time in half, but revenue went down anyway. Their clients never paid for speed; they paid for the strategic thinking in those hours. Cramping that work into a shorter time just makes them appear to be cheaper. Manufacturing is different. They get returns as AI solves problems that impact the bottom line. A parts manufacturer that I advised installed predictive maintenance systems that identified problems with equipment days before they broke, avoiding production shutdowns that cost thousands per hour. The system paid for itself in a matter of months. Winners replace tasks that are expensive, every single time they do them. Losers attempt to replace judgment calls based on years of pattern recognition. I saw one firm spending big money on AI to save analyst hours reviewing clients, but the AI could see patterns, but missed context that only humans catch after hundreds of cases. People were still needed for verification, so instead of savings, the AI was still an overhead. A distribution company that used AI to optimize routes removed hours of work for the dispatchers and experienced returns in fuel saving in months. Route math has one correct answer, but strategic analysis requires human experience. The real problem is business buying AI before their data is ready. A medical billing company bought some prediction software to identify problems with claims early, but their records were maintained in four systems with inconsistent formats and missing fields. They spent 18 months scrubbing data before AI could run. Results sold by vendors with no auditing infrastructure. Companies that have native digital operations and clean data pipelines are seeing ROI as they avoid the "two-year clean-up tax" that kills returns for companies stuck on legacy systems.