Biju Krishnan is a seasoned AI strategist and governance expert with over 20 years’ experience. For the past 12 years he’s focused on big data, analytics, and AI. His career spans foundational technology roles at global leaders, giving him a deep, practical understanding of enterprise data challenges.
Biju now specialises in pragmatic AI strategy and governance, helping organisations elevate their AI maturity while ensuring compliance.
In this post, Biju challenges the conventional methods for measuring AI ROI. These rigid metrics, Biju argues, aren’t adequate for assessing the impact of transformative technology. Instead, organisations should embrace the principle of ‘doing more with less’ to capture AI’s strategic promise:
ACAIO I consulted for explained that their company’s CFO consistently blocked AI initiatives over return on investment (ROI) concerns, killing them before they could begin. I saw a Copilot rollout shelved because, even though the pilot showed efficiency gains, those gains could not be translated into a Profit & Loss (P&L) impact.
Leadership courts headlines with AI’s strategic promise, spotlighting the smallest proof-of-concept, yet inside the boardroom they default to ROI gatekeeping that, as the next sections reveal, systematically erodes that promise.
AI functions as a general-purpose technology, similar to electricity, the internet, or mobile computing. These technologies don’t just improve existing processes; they reshape entire industries and create capabilities that didn’t exist before.
History provides clear examples of the risks organisations face when they fail to recognise and adapt to transformative technological changes. For instance, the automotive industry initially dismissed Tesla’s electric vehicles as niche products without significant market potential. Established players focused on incremental enhancements to traditional combustion engines, overlooking the disruptive impact of electrification and software integration. Meanwhile, Tesla redefined the industry’s core by prioritising these technologies. Within a decade, what began as scepticism turned into an existential crisis for legacy automakers. Today, traditional car companies are urgently retooling factories and overhauling software systems as they strive to catch up with Tesla, which has already set new industry standards.
Even companies that emerged during the internet era are not immune to this threat. Firms like Uber and Lyft, once considered pioneers of innovation, are compelled to evolve or risk obsolescence. While Uber has established a capable AI engineering division with a strong track record, it now faces mounting pressure from competitors leveraging sustained technological investment. For example, Google’s ongoing commitment to self-driving technology since 2009 has resulted in Waymo, a fleet of autonomous taxis that poses a direct challenge to Uber’s business model. This underscores the importance of continuous innovation and the danger of complacency, even for those who once disrupted their own industries.
AI’s value compounds in the same way. Early benefits appear subtle: better decisions, faster experiments, new product features. Over time, these accumulate into structural advantages.
History provides clear examples of the risks organisations face when they fail to recognise and adapt to transformative technological changes.
Organisations develop capabilities they couldn’t buy or predict at the start. The companies that treated the internet as a cost centre instead of a strategic asset in 1995 never became the ones that defined 2005.
By mid-2025, research revealed a troubling gap: executives couldn’t translate generative AI initiatives into measurable ROI.
These statistics drew intense media coverage, amplified by the ongoing narrative of AI as the next industrial revolution. Yet this enterprise adoption failure is old news to anyone who worked in AI before ChatGPT. VentureBeat reported in 2017 that 87% of data science projects never reached production. A 2020 Forbes article found that only 15% of leading firms had deployed AI capabilities at scale.
Boards and C-level executives acknowledge AI’s strategic value, but doubt their ability to capture it. That scepticism forces CAIOs and CDAIOs to justify every investment, reinforcing the ROI gatekeeping that systematically erodes AI’s strategic promise.
Traditional AI ROI assumes model accuracy equals business value. A fraud detection system promises millions in savings from fewer false positives. A churn model identifies at-risk customers for retention campaigns. But these calculations ignore the messy middle. The fraud alert means nothing if investigators lack the capacity to act. The churn prediction fails if marketing can’t execute personalised outreach. The model’s value depends entirely on the business process it touches, yet finance teams attribute outcomes solely to the algorithm.
Generative AI productivity promises are no different. A client in the pharmaceutical industry piloted Microsoft Copilot with two hundred employees. The data showed an average of four hours saved per week per employee. Their business case multiplied hourly rates by the time saved and presented it to the CFO. He rejected it immediately. As a company veteran, he knew time savings don’t directly impact the balance sheet. However, a doubled Microsoft subscription does, and it would blow the IT budget.
Maybe a different angle would have worked better. Instead of pushing ROI, they could have pitched Copilot as workforce development, a low-barrier way to upskill current employees. The alternative is hiring AI-savvy people from scratch, with recruiting fees, onboarding time, and training on company knowledge. Rather than fighting for IT dollars, they might have approached HR for learning and development funds. That conversation probably wouldn’t trigger the same budget alarms.
The point I’m trying to make is that if you know the ROI argument you are making is wishful thinking, don’t expect the CFO to see it any differently. If your pitch doesn’t have a direct ROI link, find other pillars to support the investment that are hard to dismiss.
Internal alliances are also very important. Business leaders are eager to position themselves as adopters of futuristic technology. Instead of categorising your AI investments under IT or data, seek sponsors across the organisation to strengthen your case.
A BCG India executive shared a story about creative technology investment. Their client issued a sixty-page RFP for new CRM software.
Traditional IT suppliers responded with three-hundred-page proposals that mentioned AI at least a hundred times but were clearly the same old methodology repackaged.
Most suggested a two-to-threeyear timeline with around a hundred staff on the project. In contrast, a small insurance-specific software firm responded with a QR code linking to a sandbox that already implemented the client’s requirements, plus additional features from their domain expertise. They built the functional demo in two weeks and proposed a two-tothree-month implementation timeline with pricing tied to actual business outcomes. While not directly about AI development, this shows how relying on past experience to estimate costs can lead to massive overinvestment.
AI platforms can cut experimentation and proof-ofconcept costs by creating repeatable ways to build and run AI systems. Reinventing the wheel for each use case is how companies report no return on investment.
I like the 20/20 rule from Tobias Zwingmann: keep proof-of-concept experiments under twenty thousand dollars and twenty days.
With the right approach, doing more with less is achievable today. Cloud computing and AI-assisted coding make it possible, and investing in AI platforms helps you create repeatable deployment patterns.
If your organisation’s culture requires ROI discussions to get anything approved, you have to adapt your AI strategy to that reality.
A friend who is a CAIO at a large retailer faces this exact situation. He chose to focus GenAI on the shop floor instead of knowledge worker productivity. The CFO, a long-tenure employee, understands shop floor maths intimately and is receptive to anything that helps manage increased business with fewer people. Focusing on the shop floor, where staffing costs correlate directly with per-shop revenue, makes ROI arguments easier to formulate. Pick battles you can win rather than fighting on all fronts.
Despite the increasing prevalence of AI across industries, only a select group of organisations possesses the capability to clearly demonstrate the ROI from their AI initiatives. Although monitoring AI applications is gradually becoming an industry standard, many organisations still encounter major obstacles when attempting to connect these monitoring efforts to tangible business key performance indicators (KPIs).
For most, the challenge lies in integrating technical metrics with broader business outcomes. This disconnect makes it difficult to provide solid proof of the value created by AI projects. The task is further complicated by the existence of data silos within organisations, which hinder the seamless merging of relevant information.
Recognising these challenges, Chief AI Officers (CAIOs) should prioritise efforts to align business KPIs with MLOps and AIOps metrics. By working towards the integration of these data sources, organisations can gradually build a foundation that enables clearer demonstrations of AI-driven value. Over time, the availability of such integrated data will facilitate greater buy-in from company leadership, supporting more strategic investments in AI technologies.
While there is no one-size-fits-all solution, the following advice, drawn from field experience, can help secure crucial buy-in for AI initiatives:
Educate Leadership: Offer C-level education on AI’s strategic importance to ensure leadership alignment.
Reframe Value: Acknowledge that some AI initiatives cannot survive a traditional ROI discussion; their value must be framed as a strategic investment.
Creative Budgeting: Source budgets creatively from various organisational pools – such as R&D, HR, or marketing – rather than burdening dedicated AI allocations. Bring shadow IT initiatives into the light.
Lead with Business Metrics: Where possible, lead with business outcomes, keeping ROI calculations in reserve, especially when they rely on speculative assumptions.
Build Efficient Foundations: Invest in persona-specific AI platforms to reduce the costs of experimentation and production deployment, enabling you to achieve more with fewer resources.
Engage Innovative Suppliers: Partner with vendors who offer innovative operating and financial models.
Choose Your Battles: If ROI is unavoidable, shape your AI strategy around initiatives where the correlation to value is easily demonstrated.
The rise of generative AI, like ChatGPT, has finally given AI leaders a seat at the table. It is now up to us to leverage this position, moving beyond rigid ROI measurements to deliver the transformative strategic value that artificial intelligence promises.
Chief AI Officers (CAIOs) should prioritise efforts to align business key performance indicators (KPIs) with MLOps and AIOps metrics.
