AI Seems to be Displacing Some Amount of Other Enterprise IT Spending

Among the other effects artificial intelligence might be having on enterprise information technology spending, it appears there is a shift underway from application software to AI. 

The Boston Consulting Group finds enterprises are deliberately reallocating budgets from mature categories such as enterprise resource planning  and traditional application software to fund AI, cloud, and security initiatives. 

The Information Services Group predicts enterprise AI spending will grow nearly six percent in 2025, while overall IT budgets are expected to increase by less than two percent.

Researchers at A16Z argue that enterprise AI spending is now competing directly with traditional enterprise software purchases.

IDC meanwhile argues that 66 percent of all software spending will go toward AI-enabled applications and platforms through 2028. Other studies confirm the general trend. 

Study/Article	Evidence of Shift from Application Software to AI
BCG IT Spending Pulse (2025) 1	Budgets squeezed in mature software to fund AI
ISG Study (2025) 2 3 4	AI spending outpaces overall IT budget growth
A16Z CIO Survey (2025) 5 6	AI now part of core IT budgets, not just innovation
IDC AI Spending Guide (2024-2028) 7	Majority of software spend shifting to AI-enabled apps
S&P Global SME IT Spending (2025) 8	AI spending intent higher than for other software categories
TechTarget, SiliconANGLE (2025) 9 10	AI is a top IT buyer priority, surpassing traditional software
Tangoe State of the Cloud (2024) 11	AI drives cloud and IT budget increases

Anthropic Claude Leads in Developer Market

Lots of businesses in many industries use enterprise customer revenue models successfully. And, so far, that seems true for Anthropic, whose Clause chatbot is pitched to enterprise customers, compared to OpenAI, which tends to be thought of as the consumer user leader. 

In fact, some might argue Claude has staked out a lucrative position in software development. The argument is that every major development platform (GitHub Copilot, Cursor, Replit) uses Claude as the preferred or default model.

So Anthropic leads the enterprise developer market.

source: SaaStr 

source: SaaStr 

Claude has 18.9 million monthly active users and 2.9 million mobile app users, about five percent  of ChatGPT’s user base. Yet this smaller, more focused audience generates 40 percent of OpenAI’s revenue.

Claude’s efficiency metrics:

• 16 million website visitors in January 2025

• Average session duration of 6 minutes with 3.73 pages viewed

• 37.2 percent of interactions from “computer and mathematical” sectors

• 79 percent of Claude Code chats focused on automated coding tasks

Hybrid Tech Transitions are Not Always Temporary

One historical way new technologies are adapted by legacy providers is a “hybrid” approach where the new methods are grafted onto the existing platforms, such as when steam engines were added to sailing ships. 

But such hybrids might not always be temporary or transitional. It is possible that hybrid vehicles (combining electric and internal combustion engines) are a lasting solution, not just a step toward full electrification, for example. Landline phones and mobile phones continue to coexist. Print and digital media are often blended as well. Physical bank branches and automated teller machines persist alongside digital banking, serving different customer needs and preferences. Compact fluorescent lamps (CFLs) and LEDs are used alongside traditional incandescent bulbs in many settings. In the search business, we see artificial intelligence chatbots being integrated with traditional search.

Since the early 20th century, railways have used diesel engines to generate electricity, which then powers electric motors.

We still use mechanical keyboards with our computing appliances. Some photographers use both analog and digital processes, such as shooting on film first and then using digital processing afterwards. Likewise, vinyl records coexist with digital audio.

Some students of technology change might argue that new technologies rarely replace existing systems immediately or completely. The companion issue are instances where both older and newer technologies become part of a "permanent" solution, rather than one new platform replacing another older platform completely.

That noted, the broad pattern of new technology adoption tends to feature several phases, where hybrid deployments make sense as a transitional move.

1. Augmentation - New technology enhances existing systems

2. Hybridization - New and old technologies operate alongside each other

3. Transformation - Systems reorganize around the new technology's capabilities

4. Eventual displacement - Original systems may become niche or obsolete.

So we might argue that search engines will evolve:

• AI-enhanced traditional search, where conventional search results accompanied by AI-generated summaries and insights

• Dual-mode interfaces offering both traditional keyword/filter searches and conversational AI interactions

• Graduated complexity handling where simple queries are handled by AI directly, while complex ones redirect to traditional search mechanisms

• Trust-verification hybrids, where AI generates answers while simultaneously providing source links for verification.

E-commerce platforms likewise might offer:

• AI shopping assistants alongside catalogs, offering both browsing and guided experiences

• Human-in-the-loop recommendations where AI suggests products but humans curate final selections

• Blended decision support where AI provides personalized advice while maintaining traditional filtering options

• Mixed reality shopping where AI visualization tools are integrated with traditional product photography.

Knowledge platforms will likely develop:

• AI synthesis with source transparency, offering AI summaries with clear attribution to original content

• Tiered expertise systems with AI handling routine information needs but connecting to human experts for complex topics

• Collaborative learning environments where AI tutors work alongside traditional educational content

• Memory augmentation where AI extends human knowledge rather than replacing learning.

So “disruption” might not always be as dangerous to incumbents as some might think. "Hybrid" adaptations will occur. But, over time, some hybird models might prove sustainable over the longer term.

When Robotaxis Will Displace Auto Rentals

Inevitably, people are going to wonder when, and under what circumstances, robotaxis are going to displace auto rentals, just as there was similar questioning when ride sharing first developed. 

And, in all likelihood, the decision matrix will be roughly similar. The most vulnerable segment will be short-duration rentals in dense urban areas, on short trips of one to perhaps three days, as often is the case for ride sharing as an alternative to car rental. 

Single travelers are probably more likely to use robotaxis than groups or families. 

But ride sharing still does not compete well with auto rentals in many instances, including trips that last more than several days; involve distance travel and travel to rural areas. 

For example, on an upcoming week-long trip with a distance from airport to destination of about 70 miles, one-way using Uber will probably be about $158 (fare plus tip), so assume double that for the roundtrip, or about $316. Renting a vehicle for the week (six days) will run about $328. 

Granted, that will include fuel charges and possibly some parking, but the $12 difference also includes local transportation for the whole six days. 

Actually, for that particular trip, which I take relatively frequently, even a short weekend visit works out better just renting a vehicle, rather than relying on ride sharing.

Travel Mode Cost Comparison

When Ride-sharing & Robotaxis Beat Auto Rentals

Cost-Effective Alternative

Marginal/Situational

Not Cost-Effective

Travel Scenario	Current Rideshare (2025)	Robotaxis (2027-2030)	Key Factors	Rental Car Advantages
LOCAL TRAVEL (Within Metro Area)
Single Person, 1-2 Days Business meetings, short visits	Cost-Effective $40-80/day total	Highly Cost-Effective $25-50/day total	No parking costs, insurance, or fuel. Short distances favor per-trip pricing.	Flexibility for spontaneous stops, privacy
Group (3-4 people), 1-2 Days Friends visiting, family events	Marginal $60-120/day total	Cost-Effective $35-70/day total	Per-person cost sharing makes robotaxis attractive	Group luggage space, split rental costs
Single Person, 3-7 Days Extended business, family visits	Marginal $120-280/week	Cost-Effective $75-175/week	Daily usage patterns matter. High-frequency days favor robotaxis.	Unlimited usage, storage for personal items
Group, 3-7 Days Family vacations, group trips	Not Cost-Effective $180-420/week	Marginal $105-245/week	Multiple daily trips with groups become expensive	Luggage capacity, convenience for families
OUT-OF-TOWN TRAVEL (Regional/Long-Distance)
Single Person, 1-2 Days Regional business, weekend getaways	Not Cost-Effective $200-400+ each way	Marginal $120-250+ each way	Distance is key factor. Under 200 miles may favor robotaxis by 2030.	Much cheaper for long distances, route flexibility
Group, 1-2 Days Weekend trips, events	Not Cost-Effective $300-600+ each way	Not Cost-Effective $180-375+ each way	Long-distance robotaxis remain expensive even with cost sharing	Significant cost advantage, luggage space
Single Person, 3-7 Days Extended regional travel	Not Cost-Effective $400-800+ total	Not Cost-Effective $240-500+ total	Rental becomes dramatically cheaper for extended regional stays	Unlimited local driving, much lower per-day cost
Group, 3-7 Days Family vacations, multi-day trips	Not Cost-Effective $600-1200+ total	Not Cost-Effective $360-750+ total	Rental cars dominate for extended out-of-town group travel	Massive cost advantage, convenience, storage
SPECIAL CONSIDERATIONS
Airport-Only Business Travel Fly in, meetings, fly out	Cost-Effective $30-60 total	Highly Cost-Effective $20-40 total	Airport parking costs ($15-30/day) make alternatives attractive	None - rideshare/robotaxis clearly superior
Dense Urban Areas NYC, SF, Chicago downtown	Cost-Effective Variable by usage	Highly Cost-Effective Variable by usage	Parking costs ($25-50/day) and congestion favor alternatives	Very limited - mainly privacy and storage
Rural/Remote Areas National parks, small towns	Not Available Limited service	Limited Availability May remain limited	Service availability is primary constraint	Often the only viable option

Key Assumptions & Variables

• Robotaxi Pricing: Assumes 30-40% cost reduction vs. current rideshare due to elimination of driver costs
• Rental Car Costs: $40-70/day base rate plus gas ($30-50/day), insurance ($15-25/day), parking ($0-50/day)
• Trip Distance: Local = within 50 miles, Regional = 50-300 miles, Long-distance = 300+ miles
• Usage Patterns: Assumes 2-4 trips per day for local travel, varies by scenario
• Market Maturity: Robotaxi projections assume mature service with high vehicle availability
• Parking Costs: Urban areas ($20-50/day), Suburbs ($0-15/day), highly variable by location
• Group Size: Assumes single occupancy for business, 2-4 people for leisure travel

Agentic AI Should Change Computing Infrastructure: Issue is How Much

Agentic artificial intelligence, eventually featuring teams of autonomous agents working in concert, should have some obvious impact on computing infrastructure. 

Chips will shift further in the direction of custom silicon. There will be more need for low-latency networking; more local or edge processing in addition to remote processing; more parallel and dynamic context to processing; distributed and fault-tolerant processing; more access to distributed databases. 

Specialized hardware (graphics processing units and field programmable gate arrays); more orchestration and more security also will be needed. Think perhaps of swarms of autonomous drones that have to work together, for example. 

In general, we will need “more:” more energy; more chips; more networking; more processing; more interworking and collaboration between autonomous systems. 

So how does that look for a firm such as Lumen Technologies, as a supplier of networking? Perhaps nobody doubts that “more” capacity will be needed, and might be needed in some different locations. 

The issue might be “how much” AI networking requirements actually change market demand, aside from the obvious “more capacity” that is continually needed. 

For starters, Lumen is doubling its intercity fiber mileage; upgrading bandwidth to 100 Gbps and 400 Gbps, using self-provisioning for enterprise customers, with plans to upgrade to 1.2 Tbps to 1.6 Tbps. 

Lumen also is building private networks that connect data centers owned by hyperscalers. But it might be the change in where capacity is needed that will change most. For some time, networking capacity has been driven both by the need to interconnect data centers and the need to make more bandwidth available in the access network so end users are connected with sufficient bandwidth and low-latency services. 

Agentic AI does not necessarily change that situation. Data center interconnection will drive developments in the network backbone. And AI used by edge devices will continue to rely on “on the device” local processing. But requirements for more edge processing in addition to “on the device” will likely mean more regional data center computing and therefore more bandwidth of a regional nature. 

Whether peer-to-peer requirements lead to more meshy architectures remains to be seen. But to some extent agentic AI simply continues other trends such as needs for more symmetrical bandwidth in the access network. As upstream bandwidth became more important as users started routinely uploading images and video, so agentic AI will additionally create more need for bidirectional capacity as local processors and actions combine with web services, software as a service platforms and application programming interfaces.

Barring a big change, such as Lumen somehow divesting its entire local telecom business, to become a latter-day Level 3 Communications capacity supplier, AI-driven requirements might be more incremental than disruptive.

As a financial matter, a Lumen that is a pure-play capacity provider might have 70 percent of present revenue, but a higher valuation. Some believe that could result in a Lumen valuation that is up to double what the firm presently commands, assuming "flawless execution" and probably also hinging on how the debt burden gets distributed.

Will AI Increase "Span of Control" and What Does That Mean for Middle Managers?

Most observers likely believe artificial intelligence will be used to increase span of control, at least to the extent that many supervisory functions can be handled by AI. That "should" create some opportunities to increase "span of control," which should also allow some moves to reduce middle management jobs.

“Span of control” refers to the maximum number of “direct report” humans that can be supervised by a single manager. That number tends to hover between five and eight, but can be wider or narrower depending on the industry and tasks to be supervised. 

But that assumption might not be universally applicable. If and where AI means job functions get more complex, or require even more exercise of "human" skills such as empathy, ethical judgment or complex coordination, then AI might actually decrease spans.

Still, that is likely the exception to the rule. Most observers likely believe that AI will make possible some increases in spans of control.

AI systems might  monitor performance metrics continuously, provide real-time feedback, coordinate schedules and resources and handle routine administrative tasks that previously required human managers. It remains to be seen. 

AI should be able to analyze patterns across larger groups of employees than any human manager could effectively track. So to that extent, middle manager value could decrease somewhat. 

Still, middle managers likely cannot be completely replaced. They translate strategic vision into operational tasks, provide coaching and development, facilitate cross-functional coordination, handle complex interpersonal dynamics and act as information conduits between levels. They also make nuanced decisions that require contextual understanding and emotional intelligence.

The issue is how AI might ultimately be able to become a substitute for some of those functions. 

Industry	Typical Span of Control (Manager:Direct Reports)	Characteristics & Rationale
Call Centers	Wide (1:10 - 1:20+, potentially even 1:100+)	High volume of standardized, repetitive tasks. Employees often follow scripts and procedures. Technology (CRM, call monitoring) assists with oversight, allowing managers to oversee many agents.
Manufacturing, Production	Medium to Wide (1:8 - 1:15+)	Tasks can be repetitive and process-driven, especially on assembly lines. Clear procedures and visual management tools facilitate supervision. Safety protocols often require close monitoring, but experienced workers may need less direct oversight.
Retail (Store Level)	Medium to Wide (1:5 - 1:15+)	Store managers often oversee a team of sales associates, cashiers, and stockers. Tasks are somewhat standardized, but customer interaction and merchandising can add complexity. Varies by store size and type.
Technology, Software Development	Narrow to Medium (1:4 - 1:7)	Work is often complex, requiring high levels of collaboration, problem-solving, and creativity. Teams are typically self-managing to a degree, but managers provide technical guidance, mentorship, and project oversight for highly specialized tasks.
Healthcare (Clinical/Nursing)	Narrow to Medium (1:3 - 1:10)	High complexity, critical decisions, and direct patient care. Requires close supervision to ensure patient safety, quality of care, and adherence to protocols. Nurse managers often have fewer direct reports due to the intensity and criticality of the work.
Executive Management, Leadership	Narrow (1:3 - 1:7)	Deals with strategic decision-making, complex problem-solving, and high-level collaboration. Requires significant individualized attention, coaching, and strategic alignment among direct reports (e.g., department heads, VPs).
Consulting, Professional Services	Narrow to Medium (1:3 - 1:8)	Projects are often unique and complex, requiring specialized expertise and close client interaction. Managers (project leaders) need to provide detailed guidance, quality control, and client relationship management for their team members.
Education (School Administration)	Narrow to Medium (1:5 - 1:10)	Principals oversee teachers, who in turn manage students. The complexity of managing various curricula, student needs, and administrative tasks often leads to a narrower span for administrators compared to highly standardized environments.
Financial Services ( Banking Branch)	Medium (1:6 - 1:10)	Branch managers oversee tellers, customer service representatives, and loan officers. While some tasks are standardized, customer interactions and regulatory compliance require a moderate level of oversight.

And though AI should tend to increase spans of control in most industries, AI could result in a decrease of spans in education and healthcare, two fields where productivity increases are hard to produce. AI might in many cases increase the span of control by reducing the managerial burden. 

But in some contexts it could decrease spans if roles become more complex or require closer oversight of human-AI collaboration. That might also tend to happen when empathy, ethical judgment or complex coordination is required, and that would tend to be true whether we are considering AI or simply span of control issues in general. 

Industry	Current Avg. Span of Control	AI Impact Direction	Projected Change	Reasoning
Manufacturing	5–10	Increase	10–15	AI enables predictive maintenance, process monitoring, reducing oversight needs.
Retail	8–12	Increase	12–18	AI-driven scheduling, inventory, and customer insights reduce manual supervision.
Healthcare	4–6	Mixed	3–7	Clinical AI reduces admin load, but patient safety and human empathy limit scalability.
Finance	6–10	Increase	10–15	AI automates reporting, fraud detection, and customer service functions.
Education	10–15 (academic), 5–8 (admin)	Mixed	8–20 (academic), 4–10 (admin)	AI grading and tutoring help faculty scale, but student engagement needs persist.
IT & Software	6–12	Increase	10–18	AI tools for code review, project tracking, and testing reduce managerial bottlenecks.
Construction	4–8	Slight Increase	6–10	AI in site planning and safety monitoring helps, but physical oversight still critical.
Legal Services	3–6	Increase	5–10	AI in discovery, contract analysis boosts support staff efficiency.
Public Sector, Government	5–9	Slow Increase	6–10	Bureaucracy limits rapid gains, but AI may help in records and citizen services.
Logistics, Transport	6–12	Increase	10–18	AI route optimization and fleet tracking allow for greater managerial reach.