Every major technology company has an AI story right now. The question investors, boards, and frankly a lot of CTOs are quietly asking is: where’s the return? AI ROI — the actual, measurable business value extracted from artificial intelligence investment — has become the central tension in enterprise tech in 2024. New research from EY puts hard data behind what many already suspected: most companies are spending heavily on AI and struggling to prove it’s working.
- AI ROI remains elusive for most tech firms despite heavy investment — EY research shows execution gaps are the core problem.
- Companies that clearly define AI ROI targets before deployment are significantly more likely to report measurable returns.
- Leadership alignment and data readiness are the two biggest internal barriers slowing AI value creation across the industry.
- Tech companies treating AI as infrastructure rather than a project are pulling ahead of peers on measurable business outcomes.
Table of Contents
The AI Spending Surge Nobody Can Fully Justify Yet
Spending on AI tools, infrastructure, and talent has accelerated at a pace the industry hasn’t seen since the early cloud migration wave. Hyperscalers like Microsoft, Google, and Amazon are pouring tens of billions into AI compute. Enterprise software vendors from Salesforce to ServiceNow are rebranding entire product lines around AI capabilities. And the companies buying these tools? They’re spending too — on licences, integration, training, and the consultants to help them figure out what they’re actually doing.
Yet across all that activity, the EY research surfaces a striking disconnect. The majority of technology companies report that they’re investing in AI. Far fewer can point to clearly defined, financially measurable outcomes from that investment. AI ROI, it turns out, doesn’t come automatically with the purchase order.
This isn’t a technology failure. The tools work — often impressively. It’s an execution problem, and EY’s findings suggest it’s more widespread than most companies would admit publicly.
What Separates the Winners from the Rest
The companies reporting genuine AI ROI share a few consistent traits that are worth examining closely. First, they set outcome targets before they deploy anything. That sounds obvious — almost insultingly so — but it’s apparently not the norm. Too many AI initiatives start with the technology and work backwards to a business case, which is roughly the opposite of how you’d approach any other capital investment.
Second, the leaders in AI returns have invested seriously in their data foundations. AI models are only as useful as the data they’re working with. A company running a large language model on fragmented, inconsistent, or poorly governed internal data isn’t going to get transformational results — it’s going to get plausible-sounding outputs that can’t be trusted at scale. Data readiness isn’t a sexy investment, but EY’s research is consistent with what practitioners have been saying for years: it’s the unglamorous prerequisite for everything that follows.
Third — and this one matters more than it might seem — the companies getting strong AI ROI have genuine executive alignment. Not just a CEO who mentions AI in every earnings call, but leadership teams that have agreed on what they’re trying to achieve, who owns outcomes, and how success gets measured. Without that, AI projects tend to proliferate across business units without coordination, duplicating effort and diffusing accountability.
AI ROI and the Infrastructure vs. Project Mindset
One of the more interesting frames to emerge from EY’s analysis is the distinction between companies that treat AI as a project and those that treat it as infrastructure. The project mentality produces pilots — carefully scoped, often successful in isolation, rarely scaled. The infrastructure mentality builds AI into the operating fabric of the business: into workflows, decision loops, product development cycles, and customer interactions.
The analogy to cloud is apt here. Companies that got the most from cloud computing weren’t the ones that migrated a handful of workloads and called it done. They were the ones that redesigned how they built and deployed software around cloud-native principles. AI ROI appears to follow a similar logic. You don’t extract maximum value by bolting AI onto existing processes. You get it by rethinking those processes with AI as a core assumption.
That’s a harder organisational lift, and it explains why the gap between AI leaders and laggards is widening even as overall adoption numbers look broadly similar across the industry. Adoption is easy to claim. Transformation is harder to fake.
The Measurement Problem Is Real
Part of the AI ROI challenge is genuinely methodological. How do you attribute a productivity improvement to a specific AI tool when a dozen other variables changed at the same time? How do you measure the value of a decision made faster, or a customer issue resolved without escalation, or a piece of code written in half the time? These gains are real, but they’re diffuse in ways that don’t map neatly onto a quarterly earnings line.
Some companies are building internal measurement frameworks from scratch — essentially creating new accounting categories for AI-driven value. Others are leaning on vendors to provide metrics, which introduces obvious conflicts of interest. GitHub, for instance, publishes data on Copilot’s impact on developer productivity, but those numbers come with caveats that don’t always make it into the boardroom presentation.
EY’s position, unsurprisingly for a professional services firm, is that companies need structured governance frameworks and defined KPIs from day one. That’s sound advice. It’s also advice that costs money to implement properly — which is its own small irony in a conversation about return on investment.
What the Technology Sector Gets Wrong About AI Value
There’s a particular failure mode that seems common in the technology sector specifically, as distinct from other industries. Tech companies — especially those that build software — tend to overestimate their own sophistication with AI because they’re comfortable with technology in general. There’s an assumption that because your engineers are smart and your data pipelines exist, the AI ROI will follow naturally.
It doesn’t. Building internal AI capabilities requires different skills than building customer-facing software products. Evaluating AI model behaviour at scale, managing model drift, maintaining human oversight on high-stakes decisions, integrating AI outputs responsibly into business processes — these are genuinely hard problems that don’t yield to general engineering talent alone.
The EY AI research hub and similar analyses from McKinsey and Gartner all point in the same direction: the technology sector is not immune to the execution failures that slow AI value creation in every other industry. If anything, the confidence that comes with being a ‘tech company’ may be a mild liability.
Where This Is Headed
The current period feels like the late 1990s internet moment — not in the bubble sense, but in the sense that everyone can see something important is happening, most organisations are moving, and the gap between those moving intelligently and those moving reactively is about to become very visible in financial results.
Analyst pressure on AI returns is already building. On Q2 2024 earnings calls, investors pushed Microsoft, Alphabet, and Meta harder than ever on AI monetisation timelines. That pressure will intensify. Companies that have built genuine AI ROI frameworks — with clear metrics, strong data foundations, and organisational alignment — will have concrete answers. Everyone else will be talking about the pipeline.
The uncomfortable truth is that most of the technology industry is still in the pipeline stage. That’s not a reason for pessimism — the returns are real for the companies doing this right. But it does mean the next 18 to 24 months will do more to separate genuine AI leaders from AI storytellers than anything that’s happened so far.
Source: EY
Frequently Asked Questions
Why is AI ROI so difficult to measure in the technology industry?
AI ROI is hard to pin down because gains are often indirect — faster workflows, reduced errors, better decisions — rather than a clean revenue line. Without pre-defined success metrics tied to specific business outcomes, firms end up with impressive demos and ambiguous financial impact.
What does EY’s research say about which tech companies are getting returns from AI?
EY’s findings point to companies that align AI initiatives with clear business objectives and invest in data infrastructure from the start. Those treating AI as a strategic capability — not a one-off tool — are the ones consistently reporting tangible productivity and revenue improvements.
How long does it typically take to see AI ROI after deployment?
Timelines vary widely, and meaningful financial returns from enterprise AI deployments can take considerable time to materialise. Early wins in automation and cost reduction can appear faster, but transformational impact on revenue or competitive positioning takes longer.
What are the biggest internal barriers to achieving AI ROI?
Key blockers commonly include poor data readiness — meaning fragmented, low-quality, or siloed data — and a lack of leadership alignment on what AI is actually supposed to achieve. Without both, even well-funded AI programmes tend to stall.
