Though Automation is a Given, Business Model Disruption is Big AI Upside

One frequently hears that artificial intelligence is going to change work, jobs and productivity by  automating tasks, enhancing efficiency, and providing predictive insights not feasible at low cost in the past. And all that is likely to happen. 

Study Title	Date	Publisher	Key Conclusions
How Will AI Affect Jobs	2024	Goldman Sachs	AI could replace the equivalent of 300 million full-time jobs, but may also create new jobs and boost productivity.
AI Will Transform the Global Economy. Let's Make Sure It Benefits Humanity.	2024	International Monetary Fund (IMF)	AI could impact 60% of jobs in advanced economies, with half potentially benefiting and the other half facing lower demand or displacement.
Which U.S. Workers Are More Exposed to AI on Their Jobs?	2023	Pew Research Center	Higher-paying, white-collar jobs are more likely to be exposed to AI, but the impact on job loss or gain is uncertain.
Why there will be plenty of jobs in the future - even with AI	2024	World Economic Forum	While AI may eliminate some jobs, it could also create new ones and increase productivity, leading to overall job growth.
The Impact of AI on Jobs	2019	UK Government (PwC analysis)	AI will likely shift job demand, creating roles in healthcare, IT, and research while reducing positions in administration and customer support. Regional and skill-level impacts vary significantly.
Artificial Intelligence and Jobs: No Signs of Slowing Labour Demand	2023	OECD	AI adoption is increasing demand for specialized skills and roles, without causing overall job losses yet. Task automation drives shifts in job nature rather than outright displacement.
The Impacts of Artificial Intelligence on Jobs	2017	Emerald Publishing	AI will automate both routine and nonroutine tasks, emphasizing human-robot collaboration. Skill transformations and enhanced adaptability will be key.
Future Jobs: AI's Impact on Manufacturing	2022	Springer	In China, AI in manufacturing reduced repetitive tasks but promoted job creation in skilled areas, aligning with a shift toward higher productivity.

Perhaps even greater disruption could happen if AI enables a big shift in revenue models. For example, how did the internet affect most content businesses?

Traditional newspapers and magazines saw steep declines in print ad revenue as readers--and advertising revenue--migrated online. Sure, one might argue that a shift to virtual rather than physical distribution--in some cases--led to lower production costs. 

But the biggest change was a shift from “essentially non-targeted advertising to “targeted” advertising venues, with a higher ability to specify audience on a wider range of behavioral characteristics. And that was a huge negative for legacy media. 

Also, online content aggregators and independent publishers have intensified competition for attention and audiences​, away from legacy media. 

The music industry experienced a collapse of physical media sales, which attacked the revenue which provided 85 percent to 90 percent of total revenue.

In video and film, linear formats suffered as on-demand formats grew. So revenue shifted from legacy intermediaries (TV broadcasters, cable TV services and linear networks) to  direct-to-consumer alternatives. 

And the video industry also was affected by the shift away from physical media sales. Global physical home entertainment revenue fell from $25 billion in 2004 to $5 billion by 2020, according to the OECD. 

So far, it is unclear whether--or how--AI could similarly disrupt revenue models in various industries. But that potentially will have a bigger impact than any efficiency-related AI gains.

Even if We Can't Define "AI," We Can Use It

We might agree that defining “artificial intelligence” is difficult. But lots of common and foundational concepts likewise are very-difficult to impossible to define in ways that most of us would accept. We are simply going to learn to use the tools without being able to define it with rigor and high degrees of shared meaning, in all likelihood. 

Concept	Description	Example/Insight
Love	Encompasses multiple types (romantic, familial, platonic); varies across cultures; both emotion and choice	Parent's love continues despite child's actions, suggesting it's more than just positive feelings
Intelligence	Involves multiple faculties (logical, emotional, social); culturally relative; constantly debated in science	Someone might excel at abstract math but struggle with practical problems
Art	Boundaries constantly shift; subjective appreciation; includes both traditional and conceptual forms	Is a child's drawing art? Is an AI-generated image art?
Time	Both physical and psychological component; experienced differently by different people; philosophical debates about nature	Feels slower when waiting, faster during enjoyment
Health	Balance of physical, mental, social wellbeing; cultural variations; constantly evolving standards	Someone might be physically fit but mentally stressed
Success	Highly personal; culturally dependent; changes over life stages; multiple dimensions	Career success vs. personal fulfillment
Consciousness	Self-awareness varies; subjective experience; philosophical zombie problem	Different states like sleep, meditation challenge definitions
Happiness	Mix of temporary pleasure and lasting contentment; individual variation; cultural differences	Can someone be happy while suffering?

Linear TV Enters Early Stages of "Harvesting"

Since all observers agree the linear TV (“live TV”) subscription business is dwindling, we should expect consolidation of service providers, as that happens in virtually all declining industries. The objective for such moves is to reduce cost and harvest profits for as long as possible.

The most-recent example is the new joint venture combining live TV assets owned by Disney (Hulu+Live TV) with Fubo, where Disney will own 70 percent of the asset, but Fubo leaders will manage the business.

Both the Fubo and Hulu+Live TV brands will continue to exist and be marketed.

But that will not be the end of the consolidation, as other assets eventually are combined. The logical combinations (assuming antitrust issues are addressed) are cable assets merging or satellite assets being combined. The other logical move are combinations of streaming assets. 

Company	Brand	Subscribers	Type
Charter Communications	Spectrum	14,122,000	Cable
Comcast	Xfinity	14,106,000	Cable
DirecTV	DirecTV, DirecTV Stream, U-verse TV	11,300,000	Satellite/IPTV
YouTube (Google)	YouTube TV	7,900,000	vMVPD
Dish Network	Dish Network	6,471,000	Satellite
Fubo	Hulu + Live TV plus Fubo	6,200,000	vMVPD
Verizon	Fios	3,012,000	Fiber
Cox Communications	Contour	2,695,000	Cable
Altice USA	Optimum	2,262,000	Cable
Dish Network	Sling TV	2,055,000	vMVPD

Just as discussions about future mergers of linear video programming assets (“cable channels”) now are happening, so too will distributor mergers have to happen, as the market continues to shrink. 

Like it or not, linear video content and distribution assets now are cash cows to be milked as the category inevitably declines. 

Wipro Survey Suggests Business Leaders are Moving to Cloud-Based AI Operations

A Wipro report on cloud computing finds enterprise leaders are moving towards cloud-based platforms for artificial intelligence use cases. That probably makes sense to many as projects move from pilot stage to full implementation, which requires scaling the compute infrastructure. 

source: Wipro

While not absolutely confirming the magnitude of hyperscale cloud computing leader investments in AI infrastructure, the survey of some 500 U.S. and European business leaders tends to confirm the need for additional infrastructure investment to support AI operations. 

Microsoft, for example, has announced it will spend about $80 billion on AI infrastructure in its 2025 fiscal year (ending in June 2025) after spending about $53 billion on data center infrastructure in 2023 and 2024. 

It has been estimated that Microsoft, Meta, Amazon, Alphabet and Apple alone invested up to $210 billion in infrastructure (including AI infra) in 2024 alone. 

Company	2024 AI Infrastructure Investment
Microsoft	$50 billion
Meta	$30-37 billion
Amazon	$75 billion
Alphabet	$38.3 billion
Apple	$10.8 billion

 S&P Global Ratings expects market annual spending for artificial intelligence, including traditional AI (machine learning) and generative AI, will expand to nearly $650 billion by 2028 from less than $200 billion in 2023, a compound average growth rate in the high-20 percent area. 

The agency projects the AI market will account for nearly 15 percent of total global IT spending by 2028, including semiconductors, hardware, software, and IT services.

 

source: S&P Global 

AI Capex by Hyperscalers is a Reasonable Bet on Market Growth

While there is no guarantee that the huge capital investments by hyperscale cloud computing providers will pay off, most of us might agree that enterprises are going to be spending more on generative artificial intelligence in coming years, as seen in facilities and servers. 

Global spending on data center construction is forecast to reach at least $49 billion by 2030. 
Capital spending on procurement and installation of mechanical and electrical systems for data centers is likely to exceed $250 billion by 2030. 

On the demand side, cloud computing sales are expected to rise to $2 trillion by 2030, with generative AI accounting for $200-300 billion (10-15 percent) of that spending by enterprises.

The total addressable market for cloud services (including AI "as a service") is projected to grow at a 22 percent compound annual growth rate from 2024 to 2030. 

S&P Global Ratings expects market annual spending for AI, including traditional and generative AI, to expand to nearly $650 billion by 2028 from less than $200 billion in 2023.

So, in principle, assuming much of the actual server hosting for those applications can be captured by the hyperscalers, the investments will prove financially rewarding. 

      source: Wall Street Journal, IDC

If Time is Money, and IT Saves Time, is that ROI?

A survey by IDC commissioned by Microsoft focusing on ways Copilot saves time and therefore increases productivity. It’s a good example of the familiar information technology “return on investment” exercise where we assume “time is money” and that new “technology saves time.”

By definition, such methods cannot capture benefits such as 

• Improved accuracy and quality of work

• Increased employee morale and satisfaction

• Enhanced customer service

• Better decision-making.

But those metrics are likewise hard to quantify for knowledge or office work. But advocates keep trying. 

Lumen Technologies estimates Copilot saves sellers an average of four hours a week, equating to $50 million annually. In healthcare, Chi Mei Medical Center doctors now spend 15 minutes instead of an hour writing medical reports, and nurses can document patient information in under five minutes. 

Pharmacists are now able to double the number of patients they see per day. In retail, AI models help Coles predict the flow of 20,000 stock-keeping units to 850 stores with remarkable accuracy, generating 1.6 billion predictions daily. Microsoft provides 200 such examples.  

source: IDC, Microsoft 

If you have been around such productivity estimates before, you know that the estimates are produced fairly simply: estimate time saved by workers, then multiply by the salaries of those workers. 

If you have worked in sales of information technology products to business customers, and have made such arguments yourself, you also know that buyers discount such claims. 

IDC says generative AI usage jumped from 55 percent of entities using it in 2023 to 75 percent in 2024.

For every $1 a company invests in generative AI, the ROI is $3.7 times, while some leaders using generative AI claim a returns as high as 10 times. 

On average, AI deployments are taking less than eight months and organizations are realizing value within 13 months, IDC reports. 

The ROI of generative AI is highest in financial services, followed by media and telecommunications (including mobility), retail and consumer packaged goods, energy, manufacturing, healthcare and education.

The primary way that organizations are monetizing AI today is through productivity use cases.