Like the dot-com era at the turn of the century, the current artificial intelligence (AI) moment has a last-mile problem. Back then, the bottleneck was insufficient telecommunications infrastructure. Limited fiber-optic coverage left consumers with slow connections, which impeded digital transactions and the delivery of valuable offerings. Today, shareholders and boards are pressing C-suites to generate dramatically higher value from organizational AI investments. After all, the technology offers mind-boggling capabilities and the tools are readily available, so what prevents organizations from generating benefits whose value far exceeds productivity boosts—and far exceeds the costs of implementing AI?
The answer is less about the technology than how it is implemented, supported, measured and managed. Root causes of last-mile AI shortcomings include ROI confusion; the prioritization of productivity gains over transformative improvements; slow responses to hyperscalers’ evolving pricing models; planning gaps; change management deficiencies; data quality shortcomings; and skills deficits.
Each of these last-mile bottlenecks affects revenue, expenses, the balance sheet and operations, and ultimately shareholder value, which means that navigating this internal AI delivery problem fits squarely in the CFO’s mandate. Clearing these hurdles requires CFOs to firmly grasp the last-mile problem, take responsibility for resolving it and create an execution plan in collaboration with all the requisite players.
Dissecting the Delivery Problem
A year ago, the prevailing view was that a $3 million AI investment would pay for itself via the efficiencies and headcount reductions it produced within a relatively short expectation window of less than a year. That self-funding assumption has collapsed, and payback looks more like a three- to four-year proposition. As finance groups square this reality with CEOs and boards demanding swift, outsized returns and financial markets losing patience, they should examine the following last-mile obstacles:
- Neither side of the ROI equation is clearly defined: On the investment side, AI costs do not end at the subscription fee. Token consumption, internal time, process changes, role redesigns and third-party services also contribute to costs. On the return side, organizations often neglect to track benefits—including time savings, accuracy improvements, innovation gains and customer experience enhancements—that elude straightforward calculation. (Some AI providers advocate adopting new metrics, such as cost per successful task, result dependability and other measures of completed work, to capture the true cost and value of AI outcomes.)
- An overemphasis on productivity gains persists: Eighty-four percent of the finance group’s AI spending focuses on productivity and process improvements while only 16% focuses on high-value use cases that “materially change business outcomes,” according to a CPA Practice Advisor article citing Gartner research. This imbalance largely holds throughout the rest of the organization.
- Consumption-based cost models caught companies flat-footed: Organizations remain in “Please explain it to me” mode in response to hyperscalers’ new pricing models, which in some cases carry costs that weren’t contemplated six months ago.
- Widespread usage, inadequate planning: While nearly three-quarters of CFOs report that their teams use AI, fewer than half of those teams operate under a formal AI plan, according to Protiviti’s Global Finance Trends Survey results. This gap between adoption and intentionality captures AI’s last-mile problem.
- The final 20% of AI enablement is the most difficult part: Prior to generative and agentic AI, most business processes were transformed by a combination of automation, outsourcing, robotic process automation (RPA) and low-code/no-code platforms. That holds true in finance groups, where as much as 80% of processes—primarily rules-based, repetitive activities—have been optimized. The remaining 20% of work, which is now being AI-enabled, is far more complex. These workflows are rife with exceptions and cross-functional handoffs. Taming this complexity is difficult, but it could potentially produce a fair share of the value of AI enablement.
In the CFO’s Wheelhouse
CFOs’ unwavering focus on ROI and value creation, commitment to data-driven decision-making, and growing role as innovation advocates make them well-suited to leading the way on enterprise AI investments. So, too, do the following rationales:
- CFOs control the budget and are held accountable for it: Finance leaders connect what’s being spent to the value those investments produce. A fragmented approach to organizational AI investments prevents those links from being made and subjects CFOs to board questions about unclear or insufficient AI ROI. Someone has to get their arms around AI spend and the value it drives, and that responsibility fits well within the CFO’s mandate.
- Measurement discipline is a core finance competency: In the rush to deploy AI, many companies sidestepped an analytically rigorous investment approach. Now that the issues are clearer, CFOs should encourage business leaders to step back and ask where the investment is being made and what’s expected from it and then use that information as a basis for investment approval and ongoing ROI measurements to validate current investments and pave the way for others. Finance teams are also equipped to understand and calculate cost per token, tokens per request, tokens per workflow, API call charges and related cost measures that have recently entered the picture.
- Finance’s cross-functional credibility helps replace fragmented deployments with a holistic approach: Two of the most common AI deployment missteps are the absence of a holistic execution approach and a disproportionate focus on productivity benefits. The two questions we encounter most often in our AI work at Protiviti are, “Who owns AI?” and “What’s my ROI?” As financial stewards and providers of financial insights to the rest of the business, CFOs can coalesce IT, engineering, risk management and operations around a defined AI strategy and road map. This is especially true of finance groups that have established financial operations (FinOps) groups to instill financial accountability in variable, pay-as-you-go cloud computing investments.
Turning AI Into a Long-Term Value Driver
E-commerce and telecommunications companies eventually solved their last-mile problem, though it took years to do so. CFOs must make immediate headway on their organization’s last-mile AI problem sooner than that, and the following actions offer a starting point:
- Assign each AI investment an owner, an approved investment cost, an approved operating budget and measurable expected outcomes: These basic designations help companies forecast, govern and measure AI investments and returns. Benefits can be quantified across at least four categories: productivity gains, revenue growth, cost reduction and risk reduction. A complete ROI calculation sequence involves establishing the AI investment baseline, defining desired business outcomes, quantifying the financial benefits of those outcomes, calculating AI unit economics and then using those inputs to calculate ROI.
- Segment the AI portfolio by investment purpose: The three most common AI investments are made to drive (1) workforce productivity (e.g., chat solutions), (2) operational excellence (e.g., AI built into tech applications) and (3) growth and customer impact (e.g., custom-built agents). This segmentation enables finance leaders to apply their capital allocation expertise by aligning the organization’s investments with planned measurable business outcomes.
- Reset board expectations: Board members and CEOs who greenlit AI investments based on self-funding assumptions need to recalibrate their expectations. CFOs should lead difficult conversations about earlier misjudgments while identifying more realistic payback horizons. The underlying message: The returns will materialize if the company operates with greater financial discipline. Proactive engagement with the board and business leaders also ensures alignment on risk appetite, investment priorities and governance standards. Aspects of this boardroom conversation may need to be incorporated into investor communications and earnings calls.
- Establish a formal plan: Protiviti’s Global Finance Trends Survey results also indicate that AI adoption has outpaced planning; in fact, companies that have yet to adopt AI are planning more deliberately. A formal AI strategy and implementation plan helps convert scattered usage into a measurable investment portfolio.
- Treat data quality and change management as line items rather than assumptions: Navigating the last mile of AI enablement is nearly impossible without best-in-class master data governance and heightened attention to change management, legacy system integrations and changing regulatory requirements.
- Be realistic about talent investments—and returns: Many organizations default to upskilling existing teams rather than hiring outside talent. Those skills and other must-have talent categories, including change agents and “finance technologists,” are scarce. Yet neither sourcing approach is inexpensive or fast; each requires a thorough cost-benefit analysis. One thing is clear: It is critical to define the new roles, behaviors and cultural attitudes as well as the required skill sets that enable full realization of the value proposition of AI investments.
- Don’t forget to address technical debt and its implications: Aging ERP platforms, customized applications and disconnected legacy systems make AI integration complex, expensive and slower to implement than expected. The board may not understand this point. CFOs should work with their C-suite peers to understand the nature and extent of the technical debt they are inheriting so that the appropriate assumptions are incorporated into cost models.
By embedding agile financial controls, maintaining transparent communication, resetting expectations and prioritizing initiatives that align with business strategy and drive sustainable value, CFOs can help their organizations navigate that tough last mile. They are best equipped to guide AI transformations to successful completion and maximize financial outcomes for the organization.
