Autonomy Is as Autonomy Does

Autonomy is--by definition--the defining characteristic shaping autonomous and non-autonomous artificial intelligence systems, and for a number of reasons, including human trust in autonomous systems, AI agents are going to be more common than agentic AI for some time, it is reasonable to forecast. 

There are a few exceptions. By definition, self-driving vehicles such as those operated by Waymo already have to be autonomous, as they must navigate an environment which is never static or predictable. 

Cybersecurity functions and some supply chain functions also operate “best” when autonomous and modifying their behavior in response to new conditions. 

AI agents do not normally modify their behavior without human oversight. 

The caveat is that it sometimes might be hard to differentiate between an AI agent and agentic AI, as AI agents gradually add learning functions, as will undoubtedly happen. 

Characteristic	Autonomous Agentic AI	AI Agents
Autonomy	High level of independence, can make decisions without constant human oversight1 3	Operate within defined parameters, often following predefined rules and scripts3
Goal Orientation	Focuses on achieving long-term, complex goals3	Primarily task-oriented, executing specific functions3
Adaptability	Highly adaptive, can adjust strategies based on new information or changing circumstances1 3	Limited adaptability, often require reprogramming for new tasks3
Learning Capability	Continuously learns and refines decision-making processes3	Often limited in learning capabilities3
Complexity	More sophisticated, integrating advanced technologies for reasoning and problem-solving3	Generally simpler, designed for well-defined tasks3
Decision-Making	Evaluates multiple factors and potential outcomes before making decisions3	Typically follows a straightforward, rule-based decision-making process3
Environmental Awareness	Actively perceives and adapts to the environment3	Operates within a controlled environment with limited interactions3
Proactivity	Responds proactively to changes, anticipating needs and problems2 3	Typically reactive, responding to specific inputs3
Planning	Can plan and sequence actions to achieve specific goals5	May have limited or no planning capabilities
Tool Usage	Can use various tools and capabilities to perform tasks effectively5	May have limited or predefined tool usage

Also, even today, agentic AI and AI agent capabilities overlap, some would argue. Both autonomous agentic AI and AI agents have some level of decision-making capability. While agentic AI has more advanced and independent decision-making processes, AI agents can also make decisions within their defined parameters.

Both types of AI can execute tasks autonomously, although the complexity and scope of these tasks should differ. Agentic AI can handle more complex, multi-step tasks, while AI agents typically focus on specific, predefined functions.

Both agentic AI and AI agents can interact with their environment. Agentic AI has a more sophisticated ability to perceive and adapt to changing circumstances, but AI agents can also respond to inputs.

While agentic AI has more advanced learning capabilities, some AI agents can also improve their performance over time based on new data and experiences.

Both types of AI are designed to achieve specific goals, but agentic AI arguably will be used to manage long-term, complex  goals, while AI agents focus on more immediate, task-specific objectives.

What is the Tipping Point for Linear and Streaming Video?

How much market share loss can linear subscription services take to competitive “live TV streaming” services before the share of market share loss breaks the linear business model? 

It’s an existential question for linear TV service providers without the ability to create their own live TV streaming services. At the moment, in the U.S. market,  that might appear most clear for Fox, which does not have a significant live TV streaming alternative. Disney and Comcast already have significant live TV streaming operations: Disney with Hulu+ and FuboTV; Comcast with Peacock; Paramount with Paramount+.  Warner Brothers Discovery might develop Max as such a venue. 

Disney owns ABC; Comcast owns NBC; Paramount owns CBS. Each of those TV broadcast networks have streaming versions, with limited market share, it can be argued. 

The existential issue hinges on how much more market share erosion linear video subscription services can take before the business case is unworkable. Perhaps we might argue that local broadcasters already have their own distribution (“free” over the air broadcasting). 

The greatest danger lies ahead for networks that have relied on multichannel distributors (cable TV, satellite, telco TV) for distribution. 

Looking only at the 10 or so largest video distributors, it appears that live TV streaming could already have as much as 30-percent market share, principally held by YouTube TV and the combined Hulu+ and FuboTV. 

Company	Brand	Total Subscribers	Technology
Charter Communications	Spectrum	13,000,000	Cable
Comcast	Xfinity	12,800,000	Cable
TPG	DirecTV	11,300,000	Satellite
YouTube (Google/Alphabet Inc.)	YouTube TV	8,000,000	IPTV
EchoStar	Dish Network	5,590,000	Satellite
Disney	Hulu + Live TV	4,500,000	IPTV
Cox Communications	Contour	3,050,000	Cable
Verizon	Fios	2,880,000	Fiber
EchoStar	Sling TV	2,140,000	IPTV
Altice USA	Optimum	2,100,000	Cable
Total subs		65,360,000
Linear subs		47,070,000
Streaming subs		12,500,000
Market Share Linear		72%

Using current market share, we might argue that 30 percent share loss is troublesome and growth destroying, but not an immediate case of unprofitability. The issue is more that stress will accelerate with additional share loss, as both subscription and advertising face revenue shrinkage and undoubtedly profit shrinkage as well. 

Beyond that, it is hard to predict what the tipping point--in terms of market share--will be, even accounting for industry consolidation and other profit-enhancing measures. 

Several decades ago, one might well have made the argument that a rural cable TV provider would be “out of business” if its take rates (a market share proxy) dropped from 90 percent to 70 percent. Recent contract negotiations between Charter Communications and key programming suppliers have seen Charter executives arguing they would abandon the multichannel video business entirely rather than pay the rates key programmers were demanding. 

More recently, small and rural cable TV firms have gotten out of the video subscription business, which has always favored operators with scale. 

And video streaming services are a prime example of the “direct to consumer” trend in video content distribution, much as DTC has been a trend in retailing and other content businesses. 

Industry	Example	Growth Indicator
Content	Netflix, Disney+	Subscriber growth, revenue growth, market capitalization
Retailing	Warby Parker (eyewear), Allbirds (footwear), Glossier (beauty)	Revenue growth, customer base expansion, brand loyalty
Software	Slack (workplace communication), Zoom (video conferencing)	Subscription revenue growth, customer acquisition cost, market share
Food & Beverage	Blue Apron (meal kits), Nespresso (coffee)	Customer subscription growth, repeat purchase rates, brand loyalty

Will AI Cure Decision Overload, or Increase It?

Will artificial intelligence provide significant relief from what some call “decision distress? Maybe. 

Decision distress is said to be the psychological and emotional discomfort leaders and people experience when making important decisions. And the flood of data now available to organization leaders likely doesn’t help. 

Some 72 percent of business leaders say overwhelming data has prevented them from making decisions altogether, while 91 percent report that the data deluge limits their organizational success. 

And some 85 percent of leaders have experienced decision distress, questioning or regretting their choices, which erodes their confidence in making timely decisions.

According to Oracle, 74 percent of people say the number of decisions they make every day has increased 10 times over the last three years, while 78 percent face more data availability than ever. 

Fully 86 percent say the volume of data makes decisions in their personal and professional lives much more complicated. About 59 percent say they face a decision dilemma--not knowing what decision to make--more than once every day.

Some might believe AI will take care of many big decisions leaders currently make based on “gut feeling” rather than analysis of data. But some of us suspect, even with better data, decisions will still involve a large element of choice, risk and judgment, despite the additional insights data can provide. 

Coaches can incorporate statistical probabilities into decisions to kick a field goal or go for a first down (and hopefully an ensuing touchdown) near the red zone. But there is risk, either way, no matter what the statistics suggest “should” be done in that situation. 

source: Qlik 

Better insights from data are a good thing. But AI is unlikely to be a substitute for big decisions that are inherently risky. For that reason, it might remain unclear how much AI actually can reduce feelings of anxiety about making big decisions. 

Perhaps not so surprisingly, given those statistics, 70 percent of survey respondents say they have given up on making a decision because the data was overwhelming.

At least in principle, AI should help, to a degree, by organizing and drawing conclusions from masses of data. Will that lead to more confident decision making? Again, maybe. If leaders trust the conclusions drawn from the data; the data sources; the AI tools making recommendations and their own instincts. 

Nothing is going to help an indecisive leader. And it stands to reason that AI will more commonly be relied upon to make small decisions with limited downside or upside. 

The really-big decisions are still going to be surrounded by anxiety and concern. They are, by definition, decisions with huge personal and organizational implications. 

Still, AI should help--not by making decisions risk free--but by providing an additional layer of information and insight. 

Some of those areas might include strategic matters, such as forecasting market trends, opportunities and risks. AI might likewise aid in financial areas such as resource allocation and other financial decisions.

Competitive analysis; product development and risk management are other areas where more conclusions drawn from AI analysis of unstructured data, for example, could be helpful. 

Less controversial (because it often is less risky) are AI assistance for operational efficiency improvements; human resources; compliance tasks or customer service.

Where AI will not help is freeing leaders from the responsibility for making big decisions and the consequences of those decisions, even if some fancifully believe AI can replace the functional role of CEOs, for example. 

Such arguments essentially reduce the CEO role to that of “making decisions.” One suspects that is only partially true. Organizations are full of humans. And that means the CEO role also includes a whole lot of non-rational “leading, creating meaning and dealing with humans” functions that AI can't replace.

AWS, Azure, Google Cloud Market Share: Definitions Matter

Compared to Amazon or Alphabet, Microsoft has a greater percentage of its revenue generated by “cloud” services, in large part because Microsoft’s core business is applications and software, where Alphabet is a search company fueled by advertising and Amazon is an e-commerce company powered by retail sales. 

In other words, Microsoft is able to directly categorize its application suite revenues and operating system revenues as “intelligent cloud” revenues, whereas Alphabet cannot claim its ad revenue in that same way. 

Neither can Amazon categorize its e-commerce revenue directly as “cloud” service revenue. 

That does not necessarily or directly mean Microsoft is “ahead” of Alphabet or Amazon as a supplier of cloud computing services. It is more a reflection of the fact that Microsoft can characterize significant product sales as direct “cloud services” revenue. 

source: S&P Global Intelligence 

As noted by Microsoft, Azure revenues are closer to AWS than most generally believe. 

Some analysts have attempted to make like-to-like comparisons between Azure, AWS and Google Cloud revenues that do not include Azure operating system and end user software revenue. Estimated that way, the gap between AWS and Azure is wider.

source: Statista