Business Value and the Limits of ROI

This is the first in a series looking at a new structure for planning and investing in enterprise technology over the next few years. The introduction can be found here.

AI has made an old enterprise technology problem more visible: we continue to measure long-term transformation with tools optimized for shorter-term investment decisions. ROI still plays an important role in financial discipline, capital allocation, and accountability, but it answers a narrower question than executives ask when they use it to judge transformation. It can tell us whether a particular investment generated an economic return; it is much less useful in determining whether the enterprise is building the capabilities required to create sustainable value over many years.

Deloitte found that technology investments commonly carry expected payback periods of seven to twelve months, while satisfactory ROI from a typical AI use case takes two to four years. Only 6% of organizations reported AI payback within one year (yet 85% increased AI investment during the previous year, and 91% planned to increase it again). The theory behind ROI never required a one-year return, but enterprise practice has pushed technology investment toward shorter proof periods, widening the gap between when executives expect value to be demonstrated and when transformation can reasonably produce it.

As an example, AI's long-term value will depend on infrastructure, enterprise data and context, process redesign, workforce changes, governance, applications, and eventually operating and business models. Those investments happen at different times and create value at different rates, so requiring each component to demonstrate rapid ROI can produce a portfolio of individually defensible projects while preventing the larger transformation they were intended to support.

Deloitte Canada found that 90% of executives reported positive AI productivity impacts and 88% were confident they could measure ROI, but productivity gains and cost savings remained the most common measures. BCG found nearly nine in ten CEOs reporting some cost or revenue benefit from AI in targeted areas, yet only 26% had embedded AI within broader business transformation, and only 14% clearly defined P&L impact across all AI initiatives. The stronger performers were roughly seven times more likely to redesign workflows and reshape the business end-to-end, showing that targeted ROI can coexist with weak enterprise transformation.

ROI is therefore a poor primary measure of transformational business value when the objective extends beyond improving an existing activity. Legitimate returns from automation, lower costs, or faster processes show that an activity improved, but they do not establish whether the enterprise created a better operating model, changed how decisions are made, developed new sources of growth, or built capabilities competitors will find difficult to replicate. Business Value must capture observable improvement in how the enterprise grows, operates, decides, serves customers, manages risk, and adapts, including financial results, without reducing value to them.

The larger implication extends well beyond today's AI projects because AI is unlikely to remain a discrete technology category for very long. It is becoming part of the underlying foundation on which enterprise technology will operate for the next two or three decades, much as networking, the internet, cloud, and mobile eventually became embedded assumptions in modern computing. Models, vendors, and terminology will change, but enterprises will continue building systems that understand context, make decisions, act autonomously, interact with people and machines, and consume increasingly distributed computing resources.

Infrastructure decisions must therefore account for future compute, storage, memory, power, latency, security, and workload-placement requirements, not just the workload that justified them today. Application strategy increasingly includes whether existing software should be modernized, surrounded, decomposed, or retired as intelligence moves across the technology estate, while data strategy expands into enterprise context, semantics, business rules, institutional memory, and decision histories because future systems will need far more than records to operate effectively.

The measurement problem will persist after today's AI cycle because the next major technology will arrive before most enterprises finish absorbing this one. Robotics, new computing architectures, autonomous systems, edge intelligence, quantum, and technologies still considered frontier today will eventually enter the active planning horizon. Enterprises that evaluate each wave as an independent project risk rebuilding infrastructure, retraining organizations, recreating governance, and accumulating another layer of technical debt, while a more durable strategy invests in capabilities that can absorb changing technologies without forcing the organization to reinvent itself every time.

Many of those capabilities create option value rather than immediate return, which makes them difficult to capture through project ROI. A modernized application estate can reduce the cost of adopting the next technology, better enterprise context can improve every intelligent system that consumes it, stronger identity and control mechanisms can enable greater machine autonomy, and adaptable infrastructure can support workloads that do not yet exist. Evaluating those investments only against the first use case understates what the enterprise is actually buying.

This is where index metrics become more useful because they allow executives to preserve individual measures while still seeing overall direction. I began using weighted indexes almost two decades ago for that reason: the underlying measures remain visible, while weighting reflects strategic priorities and shows whether the organization is moving toward the intended outcome. For an enterprise going through AI transformation today, early weighting might emphasize process redesign, infrastructure readiness, usable enterprise context, workforce capability, governance, and organizational capacity to absorb change, then shift toward repeatable revenue, margin improvement, customer outcomes, market expansion, or new business creation as those capabilities mature.

That creates a better relationship between short-term accountability and long-term strategy because a productivity gain can justify continuing an initiative without proving that it will scale, a revenue increase can show demand without proving that the economics are repeatable, and cost reduction can create capacity without establishing that the operating model is better. ROI remains useful for evaluating those individual outcomes, while the broader measurement model determines whether they are collectively making the enterprise more capable of absorbing and creating value from whatever comes next.

Business Value therefore sits at the beginning of the enterprise technology taxonomy because infrastructure, technology, talent, operating models, and other capabilities will continue to change while enterprise value remains the objective. Business Value is produced across the taxonomy rather than within any single technology investment, which is why boards and executive teams need to know whether today's decisions are building an enterprise that can create greater, repeatable, and sustainable value through multiple technology cycles over the next 24–36 months and well beyond.

Enterprises need long-term direction, short-term strategies to advance toward Business Value, and measurements that show whether the organization is becoming more adaptable and more valuable as the technology underneath it keeps changing. ROI remains part of that system without carrying the entire definition of business value.

Here are some reading resources: