Editorial
The Gist
- AI investments surge. Businesses and venture capitalists are pouring billions into AI, but many still struggle to justify the ROI of these investments.
- What's the long-term vision? Nearly half of AI adopters face difficulties proving value, leading to questions about the long-term economic feasibility of AI systems.
- Strategic value focus. As AI use cases evolve, companies must prioritize long-term strategic gains over short-term ROI to fully leverage AI’s potential.
There continues to be a tremendous amount of funding going into the AI market.
AI investments are occurring across all sectors. Many businesses are rerouting their capital for AI experiments, and venture capitalists are placing big bets on any start-up that has AI in its name.
Meanwhile, employees are rapidly testing AI tools in search of productivity gains. As a result, the impact of AI on the public sector will be significant. According to BCG, the potential productivity improvements in the public sector from generative AI could be worth $1.75 trillion per year by 2033.
Goldman Sachs anticipates that tech firms, corporations and utilities are set to spend around $1 trillion on capital expenditures in the coming years to support AI. Gartner predicts that AI software spending will surge to $297.9 billion by 2027, up from $124 billion in 2022.
According to TechCrunch, in the first half of 2024 alone, more than $35.5 billion was invested in AI start-ups globally. Five of the six venture rounds exceeding $1 billion in the first half of 2024 were raised by AI companies. Twenty-eight of these companies have raised more than $100 million.
This behavior is not surprising, given the promise of unprecedented efficiency, enhanced customer experience and a competitive edge.
AI Investments Skyrocket While ROI Remains Uncertain
With all these investments pouring into AI, there is a restlessness beginning to appear regarding ROI expectations. The use cases that could fully justify these immense AI investments have yet to surface. There are questions emerging about the economic feasibility of these AI systems. At some point, the justification for investments in AI systems must create value that exceeds their cost.
Among 700 AI adopters surveyed by Gartner, almost half of them indicated challenges with demonstrating the value of AI investments. As these corporate AI budgets continue to swell rapidly in the coming years, CFOs will find themselves under rising pressure to demonstrate the return on investment for such expenditures.
We’ve seen some of this before as the early phases of cloud adoption struggled to justify the ROI for migrating applications and data to the cloud. And in many cases, the CFO ended up with huge consumption bills.
At some point, we will see the investments in AI flatten as LLMs mature and more use cases make it into production. For now, expect AI investments to continue and look for companies to focus on the long-term strategic value from AI.
Keep the Big Picture in Mind With AI Investments
Technology has a history of changing the human dynamic, mostly for the good. Consider some of the most impactful technology transformations such as electricity, telephones, televisions, personal computers, internet, mobile phones and cloud computing. They have all initially created disruption and uncertainties, but in the long term they have provided significant lasting value. AI is on the same course.
However, when you are in the eye of the storm of a technical transformation, it's frequently hard to see and understand the broader impact. It’s the fear of the unknown that results in dismissing the technology as a fad or something that is too risky to adopt. Can you imagine our ancestors grasping what it means to have electric-powered lights let alone a mobile phone connected to the internet?
This is not unrealistic behavior as many new technologies are not perfect at launch. When electricity first launched, there were some real dangers as wires were not routed atop poles as they are today and instead routed on the ground. Occasionally switches would create sparks.
Mindsets also took a while to adapt as some people didn't want electricity in their homes and wouldn't let their children near any electric devices. Many people thought it was a passing fad and continued to wire their homes with gas lights as a backup.
What has changed over time is the speed at which these technology transformations are occurring. It took 2.4 million years for our ancestors to control fire and use it for cooking, but it only took 66 years to go from the first flight to humans landing on the moon. ChatGPT took just two months to reach 100 million users, the fastest consumer app ever. Today’s world cycles are extremely fast.
Related Article: Generative AI Timeline: 9 Decades of Notable Milestones
Is AI the Next Tech Bubble?
The current AI boom has drawn comparisons to the “dotcom bubble” of the late nineties and early 2000s. A series of high-profile IPOs from Netscape, Yahoo, Amazon and eBay fueled a frenzy in the IPO market culminating in 1999 when there 446 IPOs. If you consider the amount of VC funding going into AI right now, you can see similar patterns emerging.
The enthusiasm of the dotcom era was similar to today’s AI rush resulting in companies like CMGI hitting a peak valuation of $41 billion. However, it turned out CMGI was nothing but a shell company that owned a bunch of distressed assets. They lost 99% of their value when the bubble finally burst.
The bubble peaked in March 2000. It finally popped when the world realized that much of the venture funding was going to internet companies with no viable business model. These companies were using VC money to buy networking and infrastructure equipment from Cisco, servers from Sun Microsystems, storage from EMC, Oracle database licenses and PCs from Dell, fueling a tech boom alongside the rise of internet stocks. Sound familiar?
The dotcom era was fueled by a paradigm shift called the network effort, best described by Metcalfe’s Law (co-inventor of the Ethernet), which says that the value of a network increases exponentially as the connected users increase linearly. This was fueled by social media platforms, messaging platforms, online marketplaces and discussion forums.
In today’s AI era, NVIDIA is the poster child of the AI boom. NVIDIA’s annual revenue for its Fiscal Year 2024 was $60.9 billion, a 125% increase from 2023.
Over the last decade, Nvidia GPU AI-processing power is claimed to have grown 1000-fold. But these performance gains are different from the Moore’s Law era when engineers relied on the physics of smaller chips to drive performance. Instead, the latest NVIDIA Hopper architecture with its Transformer Engine uses a dynamic mix of eight- and 16-bit floating point math tailored to the needs of today’s generative AI models.
The similarities in infrastructure buildouts are startling. During the dotcom bubble, Telco providers overbuilt telecom capacity before the bust wiped out many investors. But over time, much of that capacity was put to good use as applications began to scale.
Consider the impact that ChatGPT has had today versus when Netscape was launched in the dotcom era. While ChatGPT gained more users than the Netscape browser at the time, the ability to surf the Web was pretty remarkable. This is very similar to the euphoria that has surrounded ChatGPT and the other Large Language Models (LLMs).
Looking back, the early web experience was limited to basic brochure-style sites, and it took some time for more advanced business models and online commerce to become an everyday reality. ROI was as elusive then as it is for AI now. But the long-term impact is certainly understood now.
Webinar
Sep
25
CMS Briefing: A Live Look at What’s Next in AI-Driven Platforms
Learn how leading organizations are using AI‑driven tools to publish faster, personalize smarter and stay secure.
Register
Webinar
On demand
Ready or Not: How Data-First Organizations Are Unlocking Agentforce Potential
Learn how to cut through the noise, activate Agentforce and build a Salesforce AI strategy that actually delivers.
Watch Now
Webinar
On demand
AI in Customer Service: Faster Resolutions, Happier Customers
Don’t let rising demand burn out your team. See how to build a smarter, more resilient support org.
Watch Now
Webinar
On demand
From Hype to High-Impact CX Strategies That Actually Scale
Turn buzzworthy AI and outsourcing trends into measurable CX wins with fresh 2025 data.
Watch Now
Webinar
On demand
Insights to Action Rethinking the Contact Center for Real Business Impact
Join our exclusive webinar to hear CX executives share their innovative strategies for transforming service delivery.
Watch Now
Webinar
On demand
Accelerating Healthcare Ops with AI
Sign up to see how your peers are putting AI to work — and making it stick.
Watch Now
So, is there a bubble in AI? Very likely. But will it persist like dotcom? Absolutely!
Automation vs. Human Augmentation
It’s possible that we are focused on the wrong outcome. Maybe our obsession with creating human-like AI needs to be refocused. Our goal should not be around full automation and replacement of workers but instead human augmentation to improve productivity and extend creativity.
The more human-like a machine is, the more likely it can be a substitute for human labor. The more labor that AI replaces, the more wages will go down and the less economic power people will have. This creates a trap famously dubbed the “Turing Trap” by MIT professor Erik Brynjolfsson. Brynjolfsson makes an important distinction in the way we deploy artificial intelligence.
Augmenting human capabilities can lead to new possibilities. Automation simply provides a reduction of time and cost to do what we have always done. Companies see a much easier path to obvious and immediate benefits if they choose to do the same thing faster and cheaper. Augmentation is a much more complicated task, requiring a different set of skills to innovate and do something differently.
The solution gets back to how we measure ROI. If the success of AI is purely measured by the cost-benefit of labor reduction, the long-term effects of job displacement are huge. However, if we begin to think in terms of enhancing working productivity and output, we are actually extending the value that AI can provide.
Related Article: The AI Adoption Mindset: Augmentation and Collaboration
How to Measure the ROI of AI Investments
AI ROI is difficult because it is a greenfield technology. Many AI use cases are being tried for the first time. The experimentation and iterative processes mean that ROI is in a constant evolution due to the fiddling with data, models and training, which all take time. And even if an AI application works well in one department, it might not work well in another part of the business. Very few established sets of best practices exist to consistently deliver value from various AI projects.
In these early phases of AI project experimentation, it is likely going to be difficult to assess hard ROI such as cost savings, revenue uplift or risk mitigation. It could take several years and many production rollouts to quantify “needle-moving” metrics.
The other issue is that measuring ROI on AI projects is different from traditional IT projects. A traditional IT project might use enterprise software solutions — which can more readily measure the ROI — for discrete business problems. Most AI programs impact enterprise processes, business models and revenue enablement, which makes it hard for these traditional ROI models to work.
Additionally, the costs associated with adopting AI are often amorphous. There are investments in AI infrastructure, AI tools, data preparation, data science and human capital — all of which are required to enable an AI solution. Despite these challenges, it is still critical that organizations create some measurable benchmarks by which they can hold AI projects accountable.
AI also has some use cases where ROI measurements are easier to obtain. For instance, ecommerce is one area where revenue uplift could be measured through the impact of AI-based recommendation engines. Insurance underwriting can use AI to determine the creditworthiness of loan applicants who have thin credit profiles or who lack traditional credit scores. The ROI of AI and chatbots in the contact center can determine how many fewer call center resources have been required to resolve a customer’s concerns. Finally, using anomaly detection for automating fraud detection reduces risk in banking and financial services.
Create Framework of AI KPIs
For AI projects that don’t have straightforward AI measurements, it might be useful to create a framework of certain KPIs to monitor. These can include, but not limited to:
- Workflow automation: This can include what percentages of processes are fully or partially automated, the number of low-value processes that have been fully automated and the number of errors a process creates.
- Productivity gains: This could measure the percentage of co-pilot adoption, task execution speed, code generation volume and sprint velocity of teams with and without co-pilots.
- Service levels: This could be measuring metrics such as speed, accuracy or precision, for example metrics like customer issue resolution, quality levels, customer retention and NPS.
- Adoption rates: Measuring adoption rates and usage statistics of AI tools is a good proxy measure to AI ROI.
To realize the full benefits of AI, organizations will need to continue to push AI solutions into production, measure their effectiveness and pivot as required. Maturing skills and capabilities will be critical to scale.
Related Article: Investing in AI: The Road to Customer Experience Nirvana?
Will AI Become a Utility?
There is a point at which AI technology becomes the underpinning of everything we do. It’s already embedded in many of the applications we use every day -– your phone, your car, your home, your email and much more. It would be hard to navigate your daily routine and not touch AI in some fashion.
As AI becomes more ubiquitous, it may be inevitable that it functions like a utility like water and electricity. Water and electricity are essential for human well-being, economic development and environmental sustainability.
AI may also become indispensable across industries as it continues to enable automation, enhance decision-making and drive innovation. The cloud service providers have embedded AI into their platforms and tools making it accessible in a utility-like model, where businesses and individuals can use AI capabilities on demand, paying only for the resources they consume — similar to the model electric and water companies use.
As a utility service, the ROI model for AI then shifts from measuring just traditional financial terms for AI projects to a more ubiquitous measurement across applications that touch the enterprise. Many AI applications deliver intangible benefits, like better decision-making or enhanced customer satisfaction, which can be difficult to quantify directly but still provide significant value. The dynamic nature of AI means its potential value may not be fully realized until the technology fully matures or is integrated more deeply across organizations. In the coming years, AI investments will need to prove their worth beyond the initial hype.
AI will eventually be viewed as a utility due to its growing ubiquity, on-demand availability and infrastructure-based delivery. Measuring the ROI of AI as a utility requires a combination of traditional economic measures (i.e., cost savings, revenue growth and productivity) and broader considerations like social impact, long-term innovation and environmental sustainability.
Learn how you can join our contributor community.
