How AI Changes Software Business Models

Artificial intelligence will change our experience of “applications” and “software” much as cloud computing changed the way people use apps and software. Perhaps nothing about AI experiences will change so much as deliberate actions on our part are replaced by automated actions conducted by agents on our behalf. 

Implications for Consumers	Description
Hyper-Personalization at Scale	Software evolves from offering generic interfaces to providing experiences tailored to individual behavior, preferences, and context. Streaming services, for example, use AI to not only recommend content but also to continuously optimize the entire user journey in real-time (Source 3.3).
Shift to Agentic Interfaces	Users will rely less on visiting websites or navigating apps and more on asking AI agents or virtual assistants to perform tasks on their behalf (e.g., "book me the cheapest flight next month," "draft a marketing email"). This means brand interaction and loyalty may shift from the app's visual design to the seamlessness of the agent's performance (Source 3.2).
Proactive Functionality	Apps will shift from being reactive to proactive. They will anticipate needs and offer solutions before the consumer realizes the problem. Examples include predictive analytics flagging potential issues (e.g., a subscription service detecting a user is at-risk of churning and offering a deal instantly) (Source 3.1).
Privacy and Transparency Concerns	The increased depth of personalization requires collecting and analyzing vast amounts of personal data. This amplifies consumer concerns about how their information is being used and potentially compromised, making AI ethics and transparency a core component of building user trust (Source 3.7).
Potential for "Experience Constraint"	Over-reliance on personalization risks creating "echo chambers," where the AI constrains the user's experience by only showing them content or options that align with past behavior, potentially limiting discovery or the formation of new preferences (Source 3.6).

Business models might change as much as they did when replaced by cloud-based supply: “shrink-wrapped software” purchases replaced by subscriptions; placed-based access by “anywhere access;” operational expense substituted for capital expense. 

Aspect of Transformation	Cloud Computing Model Shift	AI/Generative AI Model Shift	Source Link
Fundamental Business Model	Capital Expenditure (CAPEX) to Operating Expenditure (OPEX). (Move from owning to renting IT infrastructure).	Product Sales to Insights/Agent-as-a-Service. (Move from selling software to selling autonomy and outcomes).	The Cloud Shift, The AI Shift
Core Value Delivered	Agility, Elastic Scalability, and Cost Efficiency. (The ability to scale infrastructure instantly).	Hyper-Personalization, Prediction, and Automation. (The ability to perform tasks and anticipate needs).	Cloud Transformation, AI's Impact on Innovation
Pricing Model	Subscription (SaaS) or Usage-Based (IaaS/PaaS) on a per-resource/per-seat basis.	Usage/Consumption-Based, Value of Insights, Dynamic Pricing, or Outcome-Based Billing (Source 2.3).	Cloud Cost Optimization, AI-Driven Profit Centers
Impact on Consumer Experience	Accessible from Anywhere (Mobility) and increased application reliability (uptime).	Autonomous Agents providing 24/7, proactive, and Real-Time Functionality.	Cloud Benefits, AI in Customer Experience

Are ISPs Overselling the Value of Higher Speeds?

In the communications connectivity business, mobile or fixed, “more bandwidth” is an unchallenged good. And, to be sure, higher speeds have enabled new applications. 

But it might also be fair to argue that, beyond a certain point, “more bandwidth” supplied to consumer users of mobile apps and devices has reached something of a point of diminishing returns. 

As a test, I have spent months on a fixed network connection that rarely exceeds 70 Mbps in the downstream and perhaps 7 Mbps in the upstream. 

Do I experience the difference between the symmetrical gigabit connection I am used to? Yes. But has my work or other use cases been unpleasant or unworkable? No.

And even if I detect some difference when on a PC, do I experience any difference on my mobiles when connected to Wi-Fi? Not really. 

That has been a shocking realization. 

Don’t get me wrong. I still favor higher speeds and more bandwidth. As time goes by, the cost to supply higher capacity is no greater than supplying less bandwidth once did. 

If one’s network is built to supply symmetrical 5 Gbps, then supplying lower speeds does not really cost much, if anything. The shock has been that the usefulness of today’s networks is so high that even limited bandwidth supplies high value, and does not seemingly impair the “typical” user experience. 

The caveat, of course, is that I have no need to continually upload large files, and as part of the test, have made sure there is only a single user on the test account, and have typically connected only two devices simultaneously, typically using only one device actively. 

For starters, mobile applications are designed to work efficiently even on sub-optimal network conditions (using data compression, caching, and low-resolution defaults), so absolute “highest capacity” network access is less important. 

And though data delivery matters, it often is the device’s processing speed that matters more, in terms of supplying a satisfying user experience. 

Generation	Theoretical Peak Bandwidth (Downlink)	Typical User Bandwidth	Key Applications/Use Cases Enabled
1G (1980s)	∼2.4 kbps	N/A	Analog Voice Calls (The first truly mobile phone system)
2G (1990s)	∼64 - 144 kbps (GSM/GPRS/EDGE)	∼9 - 50 kbps	Digital Voice Calls, SMS (Text Messaging), Basic MMS (Multimedia Message)
3G (Early 2000s)	∼384 kbps (Initial) up to 21 Mbps (HSPA+)	∼0.5 - 2 Mbps	Mobile Internet Browsing, Sending/Receiving Large Email Attachments, Basic Video Streaming, GPS / Location Services
4G (2010s)	∼100 Mbps (Initial LTE) up to 1 Gbps (LTE-Advanced)	∼10 - 50 Mbps	High-Definition (HD) Video Streaming, Real-Time Online Gaming, Video Conferencing, Cloud Services, App-based Ride-Sharing
5G (Late 2010s/Present)	∼1 - 10 Gbps (Peak mmWave)	∼100 - 500+ Mbps	Ultra-HD (4K+) Streaming, Ubiquitous IoT (Internet of Things), Massively Scaled AR/VR Experiences, Cloud Gaming with minimal latency, Advanced Autonomous Vehicles

At a certain point (often cited around 10-20 Mbps for high-quality video streaming), human perception limits the value of ever-increasing speed. 

Once a webpage loads in under a second or a video streams instantly, a further increase from 100 Mbps to 1 Gbps offers little discernible benefit to the typical user for those common tasks.

My point is that it is shocking how good access networks now are; how optimized the apps are and how fast the latest devices actually process. 

It has changed my evaluation of value-price relationships for access networks, both fixed and mobile. I still prefer gigabit networks. On the other hand, I am well aware that in many instances, all that bandwidth is unnecessary. 

So different value-price decisions are rational. Higher speeds remain “nice to have.” But beyond a (to me) shockingly low point, higher speeds are not necessary. 

That is quite a shift from the days when I used to pay $300 a month for a 512 kbps connection, and thought that was money well spent. But apps did not use video; streaming music was likewise unavailable; real-time apps were few and far between and devices were much more limited in terms of onboard processing. 

Year	Device Era	Typical CPU Speed (Clock Rate)	Typical RAM	Key Architectural Change/Use Case
1996	Early PDAs/Communicators (e.g., Nokia 9000)	∼20 – 33 MHz	2 – 8 MB	Transition from basic cell phone to early data/email device (Intel i386-based).
2002	Feature Phones / PDA Hybrids (e.g., Pocket PCs)	∼150 – 200 MHz	32 – 64 MB	Shift to dedicated mobile CPUs (e.g., ARM, Intel XScale); basic multimedia.
2007	First Generation Smartphones (e.g., iPhone 1, Nokia N95)	∼400 – 620 MHz (Single Core)	128 MB	Launch of modern Mobile OS (iOS, Symbian); Web browsing, early App ecosystem.
2010	Early Android / High-End Smartphones (e.g., Samsung Galaxy S)	∼1.0 GHz (Single Core)	512 MB	Standardization of the 1 GHz clock speed; Advanced mobile gaming, HD video.
2012	Multi-Core Transition (e.g., Samsung S3, iPhone 5)	∼1.0 – 1.5 GHz (Dual to Quad Core)	1 GB – 2 GB	Introduction of multi-core processors (SoC); smoother multitasking, 64-bit architecture begins.
2015	4G LTE Flagships (e.g., iPhone 6S, Samsung S6)	∼1.5 – 2.0 GHz (Quad to Octa Core)	3 GB – 4 GB	Focus on high-resolution displays (4K video recording); 64-bit architecture becomes standard.
2020	5G Flagships	∼2.5 – 3.0 GHz (Octa Core)	8 GB – 12 GB	Integration of dedicated AI/Neural Processing Units (NPUs); Advanced computational photography, early AR/VR experiences.
2025	AI/Advanced 5G	∼3.0 – 3.5 GHz (Octa Core+)	12 GB – 16 GB+	Peak clock speed growth plateaus, emphasis shifts to core count, specialized accelerators (AI/ML), and energy efficiency.

NIMBY Slows Infrastructure Deployments

Local opposition to infrastructure projects (“not in my backyard”) remains a key constraint for any number of initiatives related to housing, energy transmission, power generation, data center siting. Even when citizens agree that “more” of any type of infrastructure is needed, those same citizens oppose local development. 

Local community opposition has become one of the primary non-technical constraints on adding new generation in the U.S., especially for utility‑scale wind, solar, and some gas projects. 

Evidence from developer surveys and studies suggests that opposition now contributes to delays for roughly half of large renewable projects and to cancellation for roughly one‑third, with typical delay times of about a year.​

A Lawrence Berkeley National Laboratory survey of wind and solar developers found that about one‑third of projects in the past five years were cancelled and roughly half experienced delays of six months or more, with local zoning, grid connection problems, and local opposition identified as the main causes. 

Community opposition and restrictive local ordinances often are viewed as leading reasons for delays and cancellations, on par with interconnection issues.​

Project delays tied to community opposition average around 11 months for solar and 14 months for wind, extending already long four‑to‑six‑year development timelines and raising sunk costs by roughly 200,000 dollars per megawatt of capacity when projects are delayed or cancelled, according to the World Resources Institute. 

In 2023, legal and regulatory trackers documented significant opposition to 378 renewable energy projects across 47 U.S. states, and identified nearly 400 local and close to 20 state‑level restrictions that are severe enough to effectively block projects in many jurisdictions, according to developers. 

Cost, regulations, zoning and changing policies all play a role in shaping the pace of infrastructure development, but citizen opposition also matters.