Introduction

Let us start with the most important number in business technology right now.

According to PwC's 29th Global CEO Survey, which polled 4,454 chief executives across 95 countries, 56% of organisations report neither increased revenue nor reduced costs from their AI implementations over the past 12 months. TechCrunch

Fifty-six percent. More than half of every company that has invested in AI across every industry, every country, every company size has seen nothing on their bottom line.

And it gets worse. Despite $30 to 40 billion in enterprise GenAI investment, a stunning 95% of organisations are achieving zero measurable return, according to a new report from MIT. TechRadar

While nearly eight in ten companies report using GenAI, just as many report realising no significant bottom-line impact. McKinsey calls this the GenAI paradox. Tech Startups

The GenAI paradox. Every company is using AI. Almost none of them can prove it is making them money.

This is the most important business technology story of 2026 not because AI does not work, but because most companies are implementing it in exactly the wrong way. And the gap between the companies getting real returns and the majority getting nothing is widening every month.

Here is the complete, honest picture and precisely what the businesses succeeding with AI are doing differently.

The Scale of the Problem — By the Numbers

Before diagnosing what is going wrong, it is worth sitting with the full weight of how widespread this problem is.

According to widely cited CEO survey data from 2026, more than half of chief executives say their companies have seen neither significant cost savings nor revenue gains from AI to date. Only a small minority report success on both fronts. The rest occupy a grey zone, where productivity feels higher but financial impact remains hard to prove. IBM

MIT's The GenAI Divide: State of AI in Business 2025 found a staggering 95% failure rate for enterprise generative AI projects, defined as not having shown measurable financial returns within six months. Crescendo AI

At scale, only about 5% of companies achieve substantial AI ROI, while 35% report partial returns. Average payoff reaches approximately 1.7 times investment, with 26 to 31% cost savings registered in functions like supply chain, finance, and client operations. Tech Startups

Most respondents reported achieving satisfactory ROI on a typical AI use case within two to four years significantly longer than the typical payback period of seven to twelve months expected for technology investments. Only 6% reported payback in under a year, and even among the most successful projects, just 13% saw returns within twelve months. Tech Startups

The picture that emerges is consistent across every major research organisation PwC, McKinsey, MIT, Deloitte, Nvidia, IBM all telling the same story. AI is being deployed everywhere. Returns are appearing almost nowhere. And the timeline for returns, when they do appear, is far longer than most businesses expected or budgeted for.

Moreover, tolerance for poor returns is running out, as CEOs, boards, and investors are making it clear they want to see demonstrable ROI on AI initiatives. 61% of the 3,700 senior business leaders surveyed feel more pressure to prove ROI on their AI investments now versus a year ago. And 53% of investors expect positive ROI in six months or less. Crescendo AI

There is a collision coming between investor expectations positive ROI in six months and the reality that most successful AI implementations take two to four years to deliver meaningful returns. Understanding that collision, and navigating it intelligently, is the most important strategic challenge facing business leaders in 2026.

Why AI Is Failing — The 7 Real Reasons

The failure is not random. Across every piece of research, the same root causes appear repeatedly. Here are the seven reasons AI investments are delivering zero return and why each one is so common.

Reason 1 — Using General AI for Specific Problems

One of the core reasons that companies fail to see ROI is the rift between horizontal and vertical AI usage. Horizontal AI, sometimes called general AI, refers to AI that can be used across a number of roles and industries. McKinsey found that many enterprises use horizontal AI models, such as copilots and chatbots, with nearly 70% of Fortune 500 companies using Microsoft 365 Copilot. Although 40% of workers find these general AI models extremely or very helpful, the benefits they provide tend to be spread out across employees — and, as a result, are less visible in terms of top or bottom-line results. Tech Startups

Microsoft Copilot is the perfect example. Millions of companies are paying for it. Employees use it daily. Productivity feels better. But when the CFO asks for the number — the actual, quantifiable, financial impact nobody can produce it. The productivity gain is real but diffuse. It shows up as slightly faster emails, slightly better document drafts, slightly quicker research spread across thousands of employees in ways that never accumulate into a visible P&L line.

Vertical AI, or domain-specific AI solutions tailored to a specific company, industry, or use case, has a higher potential for direct economic impact. Tech Startups A custom AI model that automates a specific, high-volume business process invoice processing, customer churn prediction, quality control in manufacturing delivers concentrated, measurable impact that shows up directly in cost or revenue.

Reason 2 — Starting With Technology, Not Problems

The generative AI gold rush has led many companies to prioritise implementing AI as soon as possible without thoroughly analysing use-case goals. According to a Gallup poll from late 2024, only 15% of US employees report that their workplaces have communicated a clear AI strategy. Marina Danilevsky of IBM notes: "People said, 'Step one: we're going to use LLMs. Step two: What should we use them for?' This disconnect between hype and functionality costs companies millions in lost time and resources." Tech Startups

This is the AI equivalent of buying a truck before deciding what you need to transport. The technology comes first. The problem definition comes second or never. The result is AI implementations that are technically impressive and commercially irrelevant.

The 5% of programmes that do succeed in deploying AI seem to focus on one issue. Some large companies and younger startups are excelling with AI because "they pick one pain point, execute well, and partner smartly with companies who use their tools." Startups following this blueprint have seen revenue "jump from zero to $20 million in a year." TechRadar

One pain point. Execute well. Partner smartly. That is the entire formula.

Reason 3 — Investing in the Wrong Use Cases

Companies were using AI for the wrong assignments. The research shows that AI works best with back-office tasks with a high return-on-investment, like administrative and repetitive functions. However, more than half of the funds spent on AI projects tried to use the technology for sales and marketing two areas that still need human involvement and have a lower ROI. TechRadar

This is counterintuitive and important. The most visible AI use cases marketing content, customer-facing chatbots, sales outreach are also the lowest-ROI ones. The highest-ROI AI applications are the unglamorous back-office ones: invoice processing, data reconciliation, compliance monitoring, quality control, demand forecasting. These are not exciting to demonstrate to investors. They are extremely effective at moving the bottom line.

Reason 4 — Adding AI Without Changing Workflows

Companies that report strong AI ROI tend to share a common trait: they redesign workflows rather than simply adding tools. They ask how work should change when AI is available, not how AI can fit into existing processes. This often involves unglamorous work. Data must be standardised. Decision rights must be clarified. Exception handling must be redesigned. Human oversight must be defined. When workflows are redesigned, AI outputs flow directly into systems of record. Decisions are made faster. Errors are caught earlier. Throughput increases in ways that affect cost and revenue. Only then does AI become visible in the P&L. IBM

Giving an employee an AI tool and expecting productivity to improve is like giving someone a faster car and expecting them to arrive earlier without changing what time they leave or which route they take. The tool is only as valuable as the process it is embedded in.

Time saved does not equal money saved unless leaders act on it. IBM This sentence should be printed and framed in every executive's office. If AI saves your customer service team two hours per day but you do not reduce headcount, increase service volume, or redirect those two hours to higher-value work the two hours disappear into the organisation and produce no financial return.

Reason 5 — Poor Data Foundations

AI depends on clean, accessible, and timely data. Enterprises with fragmented systems or inconsistent governance often spend more time preparing data than generating insights — stalling progress before ROI can be measured. CNBC

AI is only as good as the data it runs on. A machine learning model trained on inconsistent, incomplete, or inaccurate data will produce inconsistent, incomplete, and inaccurate outputs. For many businesses, the real AI investment they need to make is not in AI tools it is in data quality, data integration, and data governance. Without that foundation, every AI initiative is building on sand.

Companies seeing returns have invested in the underlying capabilities that enable AI effectiveness including data accessibility, technology environments built for integration, clearly defined AI roadmaps, and formalised processes for responsible AI deployment. As PwC's global chairman Mohamed Kande noted at Davos: "Somehow AI moves so fast that people forgot that the adoption of technology, you have to go to the basics. The companies that are seeing benefits from AI are putting the foundations in place. It's about execution, not technology." TechCrunch

Reason 6 — Measuring the Wrong Things

Successful programmes are treated like hiring a new employee: requiring structured onboarding, ongoing coaching, and performance tracking. Growth-focused companies are 2.6 times more likely to have rigorous tracking of AI value across the organisation, and 68% report clear metrics to measure innovation ROI effectively. Tech Startups

Most companies measure AI adoption how many employees use the tool, how many queries are processed, how many documents are generated. These are activity metrics, not outcome metrics. The only metrics that matter for ROI are business outcome metrics: revenue per employee, cost per transaction, customer retention rate, error rate, time-to-close, customer acquisition cost.

If you cannot draw a direct line from your AI implementation to a measurable change in one of these numbers, you cannot claim ROI regardless of how impressive your usage statistics are.

Reason 7 — Treating AI as a Project, Not a Capability

AI ROI does not fail because the technology underperforms. It fails when organisations expect value to appear without changing how work gets done, how decisions are made, or who owns outcomes. The companies seeing durable returns treat artificial intelligence as an operating capability, not a collection of experiments. Tech Startups

A project has a start date, an end date, a budget, and a deliverable. A capability is something your organisation can do continuously, improving over time, embedded in how you operate. The companies with zero AI ROI are running AI projects. The companies with strong AI ROI are building AI capabilities.

What the 12% Are Doing Differently — The Success Framework

The survey's most striking finding is not the 56% with zero ROI. It is the 12% roughly one in eight CEOs who report that AI has delivered both cost savings and revenue gains. What separates them from the majority is not the quality of AI tools they use. It is how they approach implementation. TechCrunch

Here is the exact framework the successful 12% are using:

Step 1 — Start With the P&L, Not the Technology

Matt Marze, CIO of New York Life Group Benefit Solutions, has been getting positive AI returns all along. The key? Pursuing and prioritising AI deployments based on the anticipated value each will produce. "From the start we wanted to be a technology, data, and AI company to drive unparalleled experiences for our customers, partners, and employees. So all along the value question, the ROI was very top of mind." Marze and his executive colleagues approach AI investments "the same way we think about all our investments" considering how they'd impact the company's earnings plan. Crescendo AI

Every AI initiative should begin with a specific financial hypothesis: if we implement this AI solution, we expect to reduce cost X by Y% within Z months, or increase revenue A by B% within C months. If you cannot make a specific, testable financial hypothesis before you start, do not start.

Step 2 — Pick One Pain Point and Execute Completely

Pick one pain point, execute well, and partner smartly with companies who use their tools. TechRadar The companies failing at AI are running five pilots simultaneously, none of which reach full production. The companies succeeding pick one high-value use case, implement it completely including workflow redesign, staff training, and outcome measurement and use the returns from that success to fund the next initiative.

Marze prioritises AI initiatives for areas that are AI-ready in terms of available data, systems, and skills; uses returns from those to fund subsequent initiatives; and designs AI systems in ways that allow for reusability so that subsequent projects can get off the ground more efficiently. Crescendo AI

Step 3 — Redesign the Workflow, Not Just the Tool

When workflows are redesigned, AI outputs flow directly into systems of record. Decisions are made faster. Errors are caught earlier. Throughput increases in ways that affect cost and revenue. IBM

For every AI tool you implement, map the complete workflow it touches. Identify exactly where AI output enters the process, who acts on it, what decision it informs, and what business outcome that decision affects. If AI output does not flow directly into a decision that affects revenue or cost redesign until it does.

Step 4 — Focus on Back-Office Before Front-Office

AI works best with back-office tasks with a high return-on-investment, like administrative and repetitive functions. TechRadar Invoice processing, data entry, report generation, compliance checking, quality control, scheduling these are the unglamorous applications that deliver the fastest, most measurable returns. Once you have proven ROI in back-office applications, you have the credibility, the data infrastructure, and the organisational capability to move into more complex front-office applications.

Step 5 — Fix the Data Before Adding the AI

The companies that are seeing benefits from AI are putting the foundations in place. It's about execution, not technology. TechCrunch Audit your data quality before selecting AI tools. If your customer data is inconsistent across systems, if your financial records are not reconciled, if your operational data lives in spreadsheets that nobody maintains fix those problems first. The AI can wait. The data cannot.

Step 6 — Measure Outcomes, Not Activity

Define your success metrics before you start not after. Choose metrics that connect directly to the P&L: cost per transaction, revenue per customer, error rate, time-to-resolution. Review them monthly. If the numbers are not moving in the right direction within six months, either the implementation is wrong or the use case is wrong. Both are fixable but only if you are measuring.

Step 7 — Make the CEO the AI Champion

As ROI remains elusive, CEOs are becoming more directly involved. In earlier phases, AI strategy was often delegated to CIOs or innovation leaders. In 2026, chief executives increasingly see AI as a personal mandate. IBM

Nearly 100% of trailblazers report deeply engaged C-suites versus only 8% of laggards. Tech Startups One hundred percent versus eight percent. The single most powerful predictor of AI ROI is whether the CEO is personally driving the AI agenda not delegating it to the technology team and waiting for results.

The Timeline Reality — Setting Honest Expectations

One of the most damaging aspects of AI hype has been the implicit promise of rapid returns. The reality, as the data consistently shows, is different.

Initial returns usually appear within 6 to 18 months as efficiency gains. More meaningful financial impact tends to emerge over 18 to 36 months, while enterprise-level ROI and competitive effects typically require 3 to 5 years. Tech Startups

Some 84% of CEOs predict that positive returns from new AI initiatives will take longer than six months to achieve. Crescendo AI Yet 53% of investors expect positive ROI in six months or less. This expectation gap is the source of enormous pressure and the driver of the premature cancellation of AI initiatives that were, in fact, on track.

The honest framework for setting AI investment expectations:

Months 1-6: Infrastructure, data preparation, workflow redesign, initial implementation. Cost without visible return.
Months 6-18: First measurable efficiency gains. Productivity improvements visible at team level. Cost savings beginning to appear in specific functions.
Months 18-36: Meaningful bottom-line impact. ROI beginning to clearly exceed investment. Competitive differentiation becoming visible.
Years 3-5: Enterprise-scale transformation. AI-enabled capabilities that competitors without AI cannot match.

This timeline is not a counsel of patience it is a call for strategic clarity. Know which phase of the journey you are in. Measure the right things for that phase. Do not cancel an 18-month initiative at month 8 because it has not yet delivered the enterprise-level returns that typically require three years.

What This Means for Kenyan Businesses

The AI ROI crisis is a global phenomenon but it has specific implications for Kenyan businesses that are worth addressing directly.

The good news: Kenyan businesses starting their AI journey now can learn from the expensive mistakes that Western enterprises have already made. The research on what works and what does not is comprehensive, freely available, and deeply instructive. You do not need to spend millions on failed pilots to learn these lessons they are documented in PwC, McKinsey, MIT, and Deloitte reports that you can read today.

The strategic opportunity: Companies can't demonstrate measurable returns will struggle to justify continued investments, while those with clear ROI stories will pull further ahead. The experimental phase is over. Substack In Kenya's competitive business environment, the companies that implement AI strategically starting with high-ROI back-office use cases, building proper data foundations, and redesigning workflows rather than adding tools will compound advantages over competitors who are either avoiding AI entirely or implementing it without discipline.

The most accessible starting points for Kenyan SMEs:

Automated invoicing and financial reconciliation directly relevant given the eTIMS compliance environment we covered earlier. AI-powered accounting tools that automate transaction categorisation, reconciliation, and tax preparation deliver measurable time and cost savings that are easy to quantify.
Customer service automation on WhatsApp — Kenya's dominant business communication channel. AI-powered WhatsApp business assistants that handle Tier 1 customer enquiries, order status, appointment booking, and FAQ responses deliver measurable cost savings and customer experience improvements with relatively short implementation timelines.
Inventory and demand forecasting for retail and distribution businesses, AI demand forecasting that reduces stockouts and overstock situations delivers direct, measurable gross margin improvement.
Document processing and data extraction any business handling significant volumes of forms, contracts, or applications can dramatically reduce processing time and error rates with AI document processing tools.

Each of these use cases has three things in common: clear financial hypothesis, measurable outcome metrics, and workflow integration that does not require rebuilding your entire technology stack.

The Bottom Line

The message is simple: the experimental phase is over. Substack

The story of AI in 2026 is not one of failure, but of unfinished transformation. IBM

AI works. The data is clear on this the 12% of companies seeing strong returns are not lucky, and they are not using better technology. They are implementing with discipline, measuring with rigour, and treating AI as an operating capability rather than a technology experiment.

The 56% with zero ROI are not failing because AI is overhyped. They are failing because they started with technology instead of problems, invested in horizontal tools instead of vertical solutions, added AI to broken workflows instead of redesigning them, and measured activity instead of outcomes.

The fix is not a better AI tool. The fix is a better strategy.

Every Kenyan business that is investing in AI or considering it should read the PwC, McKinsey, and MIT research. Not to be discouraged, but to understand exactly what separates the companies winning with AI from the majority that are not. The lessons are available. The opportunity is real. The question is whether your business will implement with the discipline that ROI requires.

Want to implement AI in your business with a clear ROI framework — not just tools? Talk to rawlinzdesigns.co.ke— we build AI-powered digital systems for Kenyan businesses with measurable business outcomes, not technology for technology's sake.

Further Reading:

Why 56% of Companies Say AI Has Delivered Zero ROI — And Exactly How to Fix It