AI’s Secret Weapon: Turning Usability Into Market Domination
Jordan Zeiger, Tensility Intern and Undergraduate Student at Cornell University College of Engineering
Armando Pauker and Wayne Boulais, Managing Directors at Tensility Venture Partners
Introduction
The exponential growth in adoption of AI, as seen with Large Language Models (LLMs) and software development tools often leads to conversations about models, model size, or compute power. However, some of the key reasons behind the stunning growth of this generation of AI products are far less flashy than embedding matrices and multi-modal agentic workflows. The real story is about usability. AI has unlocked an entirely new kind of ease of use where complexity is hidden behind natural language interfaces. Product designers and software architects are now able to build novel experiences by using AI as a layer to simplify, automate and hide complex tasks and workflows. In this blog, we explore the factors behind this growth and the lessons that can be applied to future startups in the space.
We will cover three main points:
How ease of use compresses the adoption curve
How to create a Total Addressable Market (TAM) force multiplier
How Product Led Growth (PLG) fits in
We also show how the classic market penetration S-curves and adoption curves are affected by these trends.
Compressing the Adoption Curve
The Generative Pre-Trained Transformer (GPT) architecture was not the first AI transformer ever created. Other transformers, such as BERT (Bidirectional Encoder Representations from Transformers), predate GPT. BERT itself, created by Google, was a major innovation in the world of AI and NLP (Natural Language Processing) when it was released because it was a pre-trained model that greatly improved the ability for AI to do sentence-level prediction, be context aware, and perform accurate summarization.
However, BERT and its later offshoots did not create much of a stir outside a cadre of NLP practitioners, even though by 2020 BERT quietly powered almost all of Google’s search queries behind the scenes because of its superior abilities to understand context [1]. While powerful, BERT required a high level of expertise in NLP to use and implement, so its popularity grew only within a constrained circle of expert-level AI NLP practitioners and academics. It makes sense that when OpenAI announced ChatGPT in late 2022, the team there also expected a quiet research preview. GPT-3.5, a next-word predictor with the ability to understand context, was not internally considered a major leap forward in capabilities. However, the true breakthrough was the UI design and immediate usability [2].
Figure 1 compares the public interfaces of OpenAI’s ChatGPT prompt box and Google’s search box. No setup. No technical knowledge required. No learning or onboarding. Just a textbox on a basic white background. Both accepted natural language text input from many languages and delivered immediate results.
Figure 1. ChatGPT and Google interfaces side by side. Both are extremely simple textboxes.
That frictionless interface turned a pre-trained model with 50K tokens, a feed-forward mechanism and multi-headed attention into something any layman could use without understanding any of the inner workings of this new thing called a “Large Language Model.” Adoption was magical, instant, viral, and word-of-mouth-driven, unlike the earlier BERT and other NLP tools targeted to experts.
This simplicity of use led to an unprecedented acceleration in the adoption curve as shown in Figure 2. The blue curves illustrate how an easily usable AI interface can compress typical adoption rates with a slower, smaller peak before AI and a steeper, higher peak afterward, indicating accelerated and broadened user uptake.
Figure 2. Chart comparing market dynamics before and after the integration of AI. The blue curves illustrate adoption rates, with a slower, smaller peak before AI and a steeper, higher peak afterward, indicating accelerated and broadened user uptake. The yellow S-curves represent market penetration over time, showing a traditional saturation limit "Before AI" and an expanded Total Addressable Market (TAM) "After AI." A green arrow highlights the TAM growth enabled by AI, while the Y-axis contains a discontinuity to emphasize the substantial jump in market penetration.
Lesson: Ease of use in AI products accelerates the pace of the given innovation. AI can make a product easier to use for both experts and laymen simultaneously.
Creating a Force Multiplier for TAM
One of the common pitfalls in developing products based on deep technical expertise is the desire by the product creators to prove to their users how advanced and flexible the innovation is by exposing feature after feature, screen after screen, and minute control after minute control. This leads to nuanced integrations and long times to value. The net effect is that these products experience a constrained TAM because the customer set never expands beyond the expert users with the time and energy to learn and configure complex systems.
The history of MySQL, the iconic relational database management system, is a great example. The product was created in 1995 and became open source in 2000. By 2001, the product had more than 2 million installations [3]. MySQL is known for specifically prioritizing ease of use. The descriptors “ease” and “easy” for MySQL are mentioned fourteen times in this MySQL overview article by Oracle [4]. At the time of its release, MySQL was not the most feature-rich or enterprise-ready database. However, it prioritized simple setup, intuitive defaults, and out-of-the-box functionality. The company had an internal “15 minute” rule, making sure a user could be up and running in 15 minutes [5]. Those characteristics enhanced the open source business model to open a whole new audience for relational databases beyond expert database administrators - a classic TAM expansion. The introduction of MySQL coincided with the need for databases to power ecommerce and websites during the internet boom. Now home and professional users could use a database without needing as much advanced knowledge and experience.
StackBlitz, a browser-based Integrated Development Environment (IDE) for web development, is an example of TAM expansion in the current boom of no-code/low code AI-driven software writing tools. Their original product was based on significant hard engineering work. The advancement and use in the last five years of modern browser technologies like WebAssembly, WebContainers, and WebGPU, and the advancement of laptop and desktop computer hardware, made it possible to run complex development environments instantly without installation, configuration, or downloads.
However, adoption was lagging. Despite the modern technology stack, StackBlitz was failing. Then, the team released Bolt, an AI agent layered on top to make building web applications much easier. Suddenly, non-developers could build apps just by describing what they wanted in natural language, going from vague ideas to working, running code in seconds. The interface below in Figure 3 is reminiscent of ChatGPT and Google in Figure 1 with a clean textbox for input with a monochromatic background. Bolt reversed the company's trajectory by unlocking a whole new user base and subsequently making it one of Silicon Valley's most exciting startups [6].
Figure 3. The Bolt interface, a clean textbox for input with a monochromatic background.
Bolt didn’t just improve StackBlitz. It transformed it and shifted the product from a specialized tool to a general-purpose platform for anyone with an idea.
Lesson: By making the product usable by a broader audience, AI products that simplify tasks previously relegated to experts can massively expand their TAM.
Product Led Growth Leads the Way
Product Led Growth (PLG), where growth is primarily driven by the product, is not a new concept. The term was coined by Blake Bartlett of OpenView in 2016 during the enterprise SaaS (Software as a Service) market boom [7]. The strategy has been successfully used by many companies such as Hubspot, Airtable, and Slack. Here the product is its own marketing because it drives user engagement and retention. Several business models develop around a PLG strategy, but most center around a freemium model where the user gets to try a free product (either forever in a limited fashion or for a limited time or for a limited number of uses) in the hope of upgrading them to a paid tier in the future. The open source model is similar at a high level.
All of the current AI products with hyper growth - especially in the no-code/low code software code generation space - also employ a PLG approach. Cursor is an AI-native IDE that reimagines programming as a conversation. Lovable is an AI tool for non-technical users who want to make websites using plain English without writing any code. All of these are similar to StackBlitz Bolt described above.
The PLG philosophy is embedded in how these tools market and present themselves. Cursor reached a million users [8] purely through PLG. The product was so intuitive that its users did the distribution through word of mouth. Cursor's business model is along the freemium path with “try before you buy” offering: generous free tokens, no upfront cost. The marketing and product pages have virtually zero technical terms. Just direct, human language inviting users to try out the product right from the home page with phrases like “Build your idea” or “Create with AI.” Clearly, these tools are not selling a complex technical vision to the most technical people, instead they are broadcasting the message to everyone who has an idea for a software project to just give the product a try. People can explore the product and only pay later. This lets users build trust with the tool without risk—a key feature in PLG-led adoption.
Lesson: PLG is a key component in accelerating the adoption of the current hyper-growth AI products.
Conclusion
Figure 2, shown again below for ease of reference, brings these ideas together. Before AI, both the adoption curve (blue) and market penetration curve (yellow) followed a traditional path - slow to grow and limited in scale. Post-AI, however, well-designed products can steepen the adoption curve dramatically and raise the ceiling of market penetration. The green arrow marks this shift, driven by AI’s ability to collapse technical barriers behind natural language interfaces.
Figure 2. Chart comparing market dynamics before and after the integration of AI. The blue curves illustrate adoption rates, with a slower, smaller peak before AI and a steeper, higher peak afterward, indicating accelerated and broadened user uptake. The yellow S-curves represent market penetration over time, showing a traditional saturation limit "Before AI" and an expanded Total Addressable Market (TAM) "After AI." A green arrow highlights the TAM growth enabled by AI, while the Y-axis contains a discontinuity to emphasize the substantial jump in market penetration.
AI is not just a technical upgrade, it’s a usability revolution. By collapsing complexity behind intuitive interfaces, AI acts as a powerful abstraction layer that makes sophisticated tools accessible to non-experts. This radically expands the total addressable market (TAM), turning niche products into mass-market platforms. Tools like ChatGPT, StackBlitz, and Cursor show how even incremental AI capabilities, when paired with radical ease of use, can drive exponential adoption.
When combined with product-led growth strategies such as frictionless onboarding, free trials, and viral sharing AI-enabled usability becomes a growth engine. In an AI-driven world, ease of use is no longer just good design, it’s a strategic necessity.
Citations
https://searchengineland.com/google-bert-used-on-almost-every-english-query-342193
https://www.linkedin.com/pulse/chatgpts-product-led-growth-case-study-virality-ux-scaling-5puaf/
https://www.oracle.com/mysql/what-is-mysql/?utm_source=chatgpt.com
https://fourcornerstone.com/mysql-classic-edition-business-read-intensive-applications/
https://openviewpartners.com/blog/inventing-product-led-growth