AI startups face a critical decision: build AI features on existing B2B SaaS products or create entirely new AI-native applications? This choice determines their ability to compete with Big Tech or fail to gain market share. To win against Big Tech, AI startups must create an entirely new paradigm that makes today’s SaaS obsolete. This means building applications as if none of today’s SaaS solutions will exist in five years.

This article explores:

• Why building AI extensions to existing SaaS is a losing strategy against Big Tech
• How frameworks from Thomas Kuhn and Clayton Christensen explain successful paradigm shifts
• Three practical tactics for AI startups to create their own paradigm, ranked from easiest to hardest

There’s speculation about whether B2B AI applications will be extensions of existing SaaS or create a new experience that replaces existing SaaS.

If AI apps are extensions, Big Tech will catch up to B2B startups’ innovations before they can catch up to the incumbent’s distribution. Big Tech has an advantage due to its high profit margins, which allow it to invest in internal innovation teams and acquire startup competition.

So how can startups create a completely new experience that replaces existing SaaS? Two books provide a playbook: Thomas Kuhn’s “The Structure of Scientific Revolutions” and Clayton Christensen’s “The Innovator’s Dilemma.” Both provide frameworks for paradigm shifts and disruption in science and technology.

Let’s define the steady state before a paradigm shift. According to Kuhn, “normal science” occurs within an established paradigm—accepted theories, methods, and standards. Scientists work within this framework to solve puzzles and expand knowledge, leading to cumulative progress. In the tech industry, Christensen defines “sustaining innovations” as incremental improvements that enhance existing products for high-end customers. Established companies excel at this, continually refining their offerings to maintain market leadership.

Applying these theories, the existing suite of SaaS is the “pre-AI” paradigm. Big Tech has successfully sustained SaaS innovation within this paradigm.

Next, let’s define a paradigm shift. Kuhn posits that scientific theories can’t explain anomalies, leading to a crisis. This intellectual turmoil gives way to a new paradigm, as a revolutionary idea replaces the old framework. Kuhn’s examples include Copernicus’s heliocentric model and Lavoisier’s chemical theories, which sparked significant scientific revolutions. Christensen describes “disruptive” innovations in technology that start in niche markets with simpler, cheaper products initially underperforming. Over time, these innovations improve and capture a broader market, eventually displacing incumbents. Christensen’s example is the transition from cassettes to CDs, where CDs initially served a niche market before becoming dominant.

In this case, I consider the end-user complexity of incumbent SaaS solutions to be an anomaly, while AI represents disruptive innovation. LinkedIn certifications, which prove proficiency in software, reflect this complexity, known as “knobs and dials” software, where users must understand unique commands and settings. In an enterprise environment, admins must manage this complexity, adding further burden.

To avoid falling into the trap of extending the existing paradigm of knobs and dials, startups should consider Kuhn’s winning playbook. Each tactic varies in difficulty, as it becomes more detached from existing solutions:

• Easy: Don’t build a plugin. Kuhn’s analogy: Scientists create new tools and apparatuses to observe relevant facts in their paradigm. Real-world threat: Big Tech will build it themselves.
• Medium: Don’t build an app requiring outside subscriptions. Kuhn’s analogy: Scientists invest in more machines to make nature align with the paradigm theory’s predictions. Real-world threat: Customers won’t afford it, and Big Tech will bundle it.
• Hard: Build an app that assumes none of current SaaS solutions will exist in 5 years. Kuhn’s analogy: Scientists assume constants that make the paradigm theory applicable in the real world. Real-world threat: If you don’t believe in the paradigm shift, your problem-solving will be limited by the existing paradigm’s rules.

While the “hard” option may be extreme, it’s a crucial shift to avoid solving enterprise needs with the current “jigsaw pieces.”

AI’s novel approach to solving gaps in SaaS applications achieves the anticipated outcome. However, AI startups working with the current assumptions and jigsaw pieces will fail because they complete the pre-determined picture of Big Tech, reinforcing what’s already known, instead of creating new paradigms.

Devil’s Advocate

Satya Nadella, CEO of Microsoft, asserts that the current Software as a Service (SaaS) suite is pivotal to a customer’s AI evolution because it will serve as the “specialized canvases” utilized to modify AI-generated artifacts. For instance, a user can request Copilot in Excel to generate a financial report in natural language, similar to how they would communicate with an analyst. Subsequently, the user can view and modify the report in Excel. While there is minimal manual effort, Excel remains a valuable canvas for this workflow. Nadella effectively argues that customers will derive the greatest benefit from AI that extends the existing paradigm.

However, I perceive a potential flaw in this argument. If existing SaaS products like Excel are essentially domain-specific CRUD applications designed for creating, reading, updating, and deleting artifacts such as spreadsheets, then why do customers not need to redefine the underlying data structure and schema to optimize performance in an agentic experience rather than a click-through experience? While it is feasible to utilize an LLM to learn the existing xscl artifacts to facilitate information access, ultimately, the data stored in an .xlsx file’s schema is centered around a manual click-based CRUD experience and will suffer in terms of speed and accuracy compared to a database specifically designed for agentic experiences. 

Sources:

• The Innovator’s Dilemma, with a New Foreword: When New Technologies Cause Great Firms to Fail
• The Structure of Scientific Revolutions: 50th Anniversary Edition
• https://open.spotify.com/episode/640nHdVvWgOp8tYShOpXH1?si=t_ECsGJdQc6UfltGd4-3wg

If AI apps end up as extensions to existing SaaS, Big Tech will easily catch up to a B2B startup’s innovations before that startup can catch up to the incumbent’s distribution. Big Tech has an additional advantage due to its extremely high profit margins. High margins create a large cash coffer that can be spent hedging against any new technology that could disrupt core business. Big Tech hedges by investing in internal innovation teams and acquiring startup competition.

So how can startups, against all odds, create a completely new experience that truly replaces existing SaaS? There are two books that provide a playbook: Thomas Kuhn’s “The Structure of Scientific Revolutions” and Clayton Christensen’s “The Innovator’s Dilemma.” Both provide frameworks for paradigm shifts and disruption in science and technology.

First, let’s start by defining the steady state before a paradigm shift. According to Kuhn, “normal science” occurs within the confines of an established paradigm—a set of accepted theories, methods, and standards. Scientists work within this framework to solve puzzles and expand knowledge, allowing for cumulative progress and collaboration as they build on each other’s work without constantly debating foundational principles. Similarly, in the tech industry, Christensen defines “sustaining innovations” as incremental improvements that enhance existing products to meet the needs of high-end customers. Established companies excel at this, continually refining their offerings to maintain market leadership.

Applying these theories to the question at hand, I view the existing suite of SaaS as the “pre-AI” paradigm. Big Tech has been successful in sustaining the innovation of SaaS products within this paradigm.

Next, let’s define a paradigm shift. Kuhn posits that occasionally in science, “anomalies” arise that existing scientific theories cannot explain, leading to a crisis. This intellectual turmoil eventually gives way to a new paradigm, as a revolutionary idea gains acceptance and replaces the old framework. Kuhn provides the example of Copernicus’s heliocentric model and Lavoisier’s chemical theories, both of which sparked significant scientific revolutions. Similarly, Christensen describes “disruptive” innovations in technology that start in niche markets with simpler, cheaper products that initially underperform compared to established solutions. Over time, these innovations improve and capture a broader market, eventually displacing incumbents. Christensen provides the example of the transition from cassettes to CDs, where CDs initially served a niche market before becoming the dominant technology.

In this case, I view the end-user complexity as the anomaly in incumbent SaaS solutions, and AI as the disruptive innovation. This complexity is reflected in the LinkedIn certifications, which prove someone has taken a course on how to use the software. I refer to this as “knobs and dials” software, where a user interacting with the software has to understand a set of commands and settings unique to that application. In an enterprise environment, admins have to worry about the user misusing the “knobs and dials,” adding additional complexity to the management of these solutions.

Now, let’s revisit our question: how can startups ensure they are not falling into the trap of extending the existing paradigm of knobs and dials instead of building their own? Here is my interpretation of the winning playbook Kuhn would share today. Each tactic varies in how difficult it will be to effectively implement in a customer’s environment, as the tactic becomes further detached from existing solutions:

Easy: Do not build a plugin.

Kuhn’s analogy: Scientists figure out novel ways to observe the relevant facts in their paradigm (e.g., star positions in astronomy or wavelengths in physics) and to do this, they create new tools and apparatuses.

Real world threat: Big Tech will build it themselves.

Medium: Do not build an app that requires any outside subscription.

Kuhn’s analogy: Scientists try to make nature line up with the paradigm theory’s predictions. This involves investing in more machines.

Real world threat: Customers won’t be able to afford it, and Big Tech will just bundle it.

Hard: Build an app that expects none of the current SaaS solutions will exist in 5 years.

Kuhn’s analogy: Scientists assume constants that make a paradigm theory applicable in the real world, such as Avogadro’s number in chemistry or Boyle’s law in physics.

Real world threat: If you don’t believe in the paradigm shift completely, your problem-solving will be shackled by the existing paradigm’s rules.

While one could argue that (3) is extreme, I’d argue it’s a critical mentality shift to avoid solving the needs of enterprise with the current set of “jigsaw pieces.”

Using AI to solve the gaps in the existing suite of SaaS applications is achieving the anticipated in a novel way. And the AI startups that work with the current set of jigsaw pieces and assumptions will fail because they are completing the pre-determined picture of Big Tech, proving what is already known, instead of creating their own new paradigm.

This content is password protected. To view it please enter your password below:

Password:

This content is password protected. To view it please enter your password below:

Password:

This content is password protected. To view it please enter your password below:

Password:

This content is password protected. To view it please enter your password below:

Password: