How Startups Win – The 7 Powers Framework 

If the end state of your startup is just a GPT-API call away, where does your power lie? 

OpenAI has redefined the B2B landscape, evolving from model provider to ecosystem that rivals the platforms of tech giants like Apple and Google. Startups seeking to compete in the Generative AI (Gen AI) space face a daunting challenge: without the distribution, resources, and established brand of these tech giants, there’s limited viability for startups to successfully compete.

This essay introduces a strategic framework to help startup founders think about winning in an environment where software offers no moat.

We will explore how startups can find footing by applying Hamilton Helmer’s 7 Powers at the frontier of AI. This framework provides a blueprint for application-level strategies to carve out a durable competitive edge, transforming strategic thought into defensible traits.

We will adapt the 7 Powers framework across three critical layers, delineated across three different essays: the Application Layer, where customers engage directly with the product; the Model Layer, the core engine driving outcomes in General AI products; and the Execution Layer, showcasing what it takes to orchestrate a culture and process to bring the application and model layer to life. As the three essays will demonstrate, each layer is as valuable as the last, and each lever therein, combined with the right strategies, helps to generate power in an environment where none is granted lightly.

Central to our narrative is a case study of Perplexity AI, a smart search provider delivering expert-level responses to queries with grounded citations. Through Perplexity AI's journey, we will demonstrate how the strategic application of the 7 Powers at each layer can lead to power.

Our focus in Part I is on the Application Layer, spotlighting the art of product sense and the founders' pivotal role in introducing distinctive products to crowded markets. We will also assess the critical levers at a founder's disposal—value, usability, viability, and feasibility—and how their strategic use can strengthen a startup's market position.

With our journey outlined, let’s start with Part I: The Application Layer.

Here, differentiation begins with a deep understanding of product sense—a startup's first step toward building a digital moat. Within an environment where technological capabilities risk becoming generic, the thoughtful deployment of these technologies establishes a distinct domain for startups. Product sense thus extends beyond the basic functionality of the product to how it meaningfully connects with and surpasses user expectations, laying the foundation for a competitive edge.

The Application Layer: Product Sense and Startup Power

For early-stage companies, strong opinions coupled with strong visions from founders capture the attention of investors and captivate users. However, the seductive allure of obvious, low-risk ideas leads to vulnerabilities, as these ideas are the easiest to replicate and compete against.

To command power at the Application Layer, founders must confront a tough question: “What’s obvious to you and no one else?”

Securing power at this stage requires rigorous, often sobering, self-reflection: What unique advantages do you possess –experience, network, or insights – that make you best suited to build a startup that becomes the next unicorn?

Only through candid self-examination can a meaningful opportunity unravel. When guided by empathy, this introspection can unearth the holy grail of ideas: identifying a broken workflow – the most valuable thing a startup in this space can solve.

It takes diligence to find this Delphi-like inspired insight; this empathy-driven understanding leads to the pinnacle of achievement at the Application Layer: a refined product sense. Product sense, fostered by intuition and a keen perception of unmet customer needs, becomes the north star. It guides teams through the fog of limited data, allowing confidence in the instinctual recognition of customer needs, even before they can be fully articulated.

This approach attempts to refine the human form and, in doing so, earn the privilege of selling a solution, not just software.

Again, what’s obvious to you and no one else?

This innate understanding, our product intuition, undergoes refinement and validation within a structured framework known as product sense. This framework rests on four critical pillars:

1. Value Risk: What’s the value of the problem, and where does your early solution show elements of market pull?

2. Usability Risk: Will users figure out how to use it?

3. Viability Risk: What is the business case, and how does it integrate with your existing strategy and vision?

4. Feasibility Risk: Do you have the skills or network to build this in your vision? 

Why Start with Product Sense?

Creating billions of dollars in enterprise value, by nature, is hard. While some unicorns start as simple ideas (bundling and re-bundling), their value is defined by their lasting impact on customer pain points.

At this juncture, product sense is indispensable. It transcends opinion—it embodies an enlightened perspective that captivates customers by addressing their needs in ways that others have overlooked. That’s what customers are willing to pay for.

Embracing this philosophy, we apply product sense to the Application Layer, where specialization becomes the key to unlocking power.

Product Sense and the Application Layer

Though foundation models like GPT-X will reach step-function increases in value, the long tail of specialization remains verdant ground for value capture. It’s the power of specialization that propels a company like Snap from a potential commodity to a distinctive, multi-billion-dollar public entity, far from being just another “Google Cloud Provider wrapper.”

Opinionated specialization matters.

The evolution of technology, like the journey of software engineers from the generalists of the early '90s to today's specialized roles—front-end, back-end, or full-stack—moves inevitably from the generic to the specific. Specialization is what drives progress, and so, too, must your product become specialized to meet users' needs.

The bottleneck for “ah-ha” moments (when a customer realizes the value and benefit of your service) in Gen AI is not the technology - it's in closing the gap between the user's knowledge level and the technology's abilities.

Most of today's internet users tap into only a fraction of the value that Gen AI can offer. The software hasn’t met users at their level of understanding.

The inner workings of black-box probabilistic models aren’t common knowledge. It takes familiarity with a model's nuance to go deeper into prompt engineering. For many, their initial encounter with ChatGPT in December 2022—to request a simple haiku—marked both the beginning and end of their experience. The “magic” of the technology ceased where their proficiency in its use did.

Moreover, the absence of navigational cues can be daunting. Staring at a blinking cursor on an empty text box, users grapple with a challenge that artists have faced for centuries—the intimidation of the blank canvas. Reducing the “time to trust” in new technology is a critical lever. Recognizing this issue of empathy not only frames it as an opportunity but also informs our initial focus on self-reflective questions to cultivate this very empathy.

For founders, this presents an opportunity for your solution to be focused on a workflow rather than a generic use case. Whether through guided, embedded, or subtly nuanced prompts, your design should echo with empathy, meeting users where necessity meets familiarity.

Much like Superhuman captivated its audience with low latency and intuitive keyboard shortcuts—distinct from those accustomed to a standard operating system's email client—it's essential to facilitate the most natural, zero-friction path to the “ah-ha” moment.

Differentiation comes from integrating these powerful tools into niche markets and focusing on crafting exceptional end-user experiences rather than trying to reinvent the underlying technology.

Just as Superhuman transformed how we interact with email, startups should strive for those ah-ha moments that arise from smooth, natural user experiences and an opinionated product vision.

This is at the heart of Counter-Positioning (Power Three), a strategy where a startup’s distinctive approach distinguishes itself from established companies. By concentrating on bespoke, user-centric solutions, startups can differentiate themselves and construct digital moats that legacy, generalist firms hesitate to cross. This strategic counter-positioning transforms what might be seen as a niche-market disadvantage into a compelling competitive advantage.

With an understanding of the Application Layer established, we are ready to explore the underpinnings that can elevate a startup from a participant to a leader in the Gen AI space. The 7 Powers framework offers a comprehensive lens through which to view and navigate the competitive dynamics. As we transition to this framework, it's crucial to recall the significance of product sense—a signal informed by, and informing, each strategic power. Every power, ranging from counter-positioning to branding, acts as a fulcrum, magnifying the distinctive value proposition inherent in the Application Layer.

The 7 Powers Framework

7 Powers may be the only business book worth reading.

Like a Swiss Army knife – it’s deceptively simple yet unfolds into a grounded toolset for evaluating any business.

The premise centers on guiding businesses to develop strategies ensuring profit, and hopefully, prosperity. Drawing from basic economic principles, we can understand that exceptional businesses maintain robust profit margins. However, the textbook market scenario posits that such success inevitably attracts competition, where each new entrants methodically erodes these margins until a market equilibrium is reached.

What Econ 101 didn’t cover is how, as an astute founder, you prevent competition from stealing your lunch.

The answer is the 7 Powers – a strategic manual for protecting profits against the tightening grasp of competition.

A few monomers make up the basis for understanding Helmer’s 7 Powers. The first is Strategy (big ‘S’) – the study of what determines business value, defined as enterprise value. The next is Power – the conditions creating the potential for persistent differential returns. And lastly is strategy (small ‘s’), which is the tactical path to continuing Power in significant markets.

Applying the 7 Powers to defensibility in Gen AI focuses on small ‘s’ strategies, all relative to competition within your chosen niche.

Presented here is an accessible synopsis of the 7 Powers, complete with brief explanations. It's important to recognize that not every power is pertinent to the startup phase (e.g., scale economies, network economies). Inherently, some powers are more ephemeral, their relevance fluctuating with a company's maturity. As a business transitions from its inception to a more established entity, these powers shift and realign with the changing contours of the market. For those already versed in the 7 Powers, you may wish to proceed directly to the Perplexity AI case study for applied insights.

Power One: Scale Economies

Definition:

• Gaining a cost advantage through increased production, where larger volumes spread out fixed costs, leading to lower average costs per unit. Smaller competitors have difficulty matching the price and margin profile, creating a barrier to market entry.

Example:

• Costco Wholesale operates on scale economies, using its membership model to sell products in bulk at low prices that other retailers can’t match.

Power Two: Network Economies 

Definition:

• The value of a product or service increases as more people use it. This can create a virtuous cycle where increased usage attracts more users, further entrenching the company's market position.

Example:

• Visa has leveraged network economies by creating a vast network of merchants and consumers, making it one of the most widely accepted payment methods worldwide.

Power Three: Counter-positioning

Definition:

• A newcomer adopts a new, superior business model that the incumbent does not mimic due to anticipated damage to their existing business.

Example:

• Netflix's pioneering DVD-by-mail service disrupted the video rental industry, offering a no-late-fee model that directly challenged Blockbuster's core revenue strategy, ultimately leading to the latter's decline.

Power Four: Switching Costs

Definition:

• High switching costs create a sticky customer base by making it expensive or cumbersome for customers to move to a competitor, thus reinforcing customer loyalty and retention.

Example:

• ADP, a provider of HR management software and services, benefits from high switching costs due to the complexity and disruption that changing payroll providers would cause for businesses.

Power Five: Branding

Definition:

• A strong brand can create a preference for a company's products or services, even if they are similar to those offered by competitors.

Example:

• Coca-Cola's brand is synonymous with soft drinks and has become a cultural icon, allowing it to command premium shelf space and consumer loyalty.

Power Six: Cornered Resource

Definition:

• Exclusive control over a scarce resource can provide a significant advantage.

Example:

• De Beers has historically controlled a significant portion of the world's diamond supply, influencing availability and prices in the global diamond market.

Power Seven: Process Power

Definition:

• Superior processes and operations that are hard to replicate can lead to long-term competitive advantage.

Example:

• Toyota's implementation of the Toyota Production System (TPS) or 'lean manufacturing' has been a key factor in its ability to produce high-quality vehicles efficiently.

Before this list, I made a remark that these powers can be transient. Helmer describes this as “Power Progression,” in which the powers with which one started evolve into others relative to the company's competitive landscape. Most are easily seen with mature companies:

Such powers serve to amplify the distinct value proposition identified at the Application Layer. By honing in on tailored solutions that precisely address user needs, a startup may commence with counter-positioning. Over time, it can establish formidable switching costs or cultivate a potent, niche brand presence.

Perplexity AI exemplifies the practical application of the 7 Powers, grounded in robust product sense. By focusing on delivering expert-level responses with citations, the startup not only differentiates itself but also strategically aligns with principles like Counter-Positioning and Branding, showcasing a direct path from concept to competitive edge.

Paths to Power: Perplexity AI’s Application of Counterpositioning, Cornered Resource, and Brand 

Perplexity AI's success stems from its ability to precisely address specific user needs, creating a useful and indispensable product. This focus has allowed it to effectively employ several of the 7 Powers, such as Counter-Positioning, by offering a service that goes beyond the capabilities of generic chatbots and becoming a trusted source of information.

Counter-positioning - Expertise and Citations: 

In AI-driven chatbots, the market is saturated with solutions designed for basic queries. Perplexity AI has carved out a niche by positioning itself as a chatbot that understands complex questions and provides expert-level responses with citations. Exemplifying product sense, Perplexity AI understands what users need, but more importantly, how they wish to receive that information. Perplexity AI's user-centric innovation taps into a "knowledge-thirsty" market, offering a service beyond the standard chatbot experience to become an indispensable tool for users seeking substantiated answers. It’s through this empathy in building the missing connection between delivering answers and the sources for which they are derived that Perplexity AI has established a considerable moat in this crowded environment.

Cornered Resource – RAG Pipeline and Indexing:

Perplexity AI has developed a unique asset in its Retrieval-Augmented-Generation (RAG) pipeline, which allows the system to pull from a curated database of indexed, retrievable information. This system is more than “fetching-data.” It’s an integral aspect of applying product sense. It introduces a human-in-the-loop element, peeling back the layers of Gen AI’s “black box” to reveal the origins and derivations of its responses. This technology is not just a feature; it's a cornerstone of Perplexity AI's offering, enabling the delivery of traceable responses to their sources. By cornering this resource, Perplexity AI ensures its platform is not another AI chat interface.

Brand - Trust and Authority: 

Perplexity AI’s focus on delivering cited responses and expert-level information epitomizes the essence of product sense, cultivating a brand that resonates with trust within smart search. It’s a commitment to accuracy has shaped a reputation for reliability in an industry often challenged by bias and hallucination. This brand power attracts users who prioritize quality and dependability in their information sources and who see Perplexity AI as more than a tool—it's a trusted advisor. 

Perplexity AI's adept employment of Counterpositioning, monopolization of unique resources, and Branding underscores its profound comprehension of user needs and its dedication to providing trustworthy, expert-validated information.

Conclusion

Perplexity AI's journey illuminates a path for Gen AI startups, showcasing that lasting success is not solely hinged on technological innovation but also on a deeply ingrained product sense that drives strategic distinction and meticulous execution.

Their story underscores the compelling impact of adhering to strategic frameworks like the 7 Powers, which calls on startups to embed these principles within the Application Layer thoughtfully. This acute product sense—this intrinsic grasp of and adaptation to user needs—sets the stage for the next installment. In Part II, we will dissect the Model Layer's mechanics, revealing how a steadfast commitment to product sense can cement a lasting competitive edge, elevating a startup from simply participating in the market to defining it.

Acknowledgements:

As I reflect on the journey of crafting "Dry Castles, Digital Moats: Defensibility in Generative AI – Part I," I am profoundly grateful for the collaborative spirit and intellectual rigor brought to the table by a remarkable few who answered the call for commentary. Their diverse expertise and insights have been instrumental in shaping a nuanced and comprehensive exploration of strategic frameworks in Generative AI for startups.

In-Depth Technical Insights:

• Rakesh Gidwani, CTO & Partner at Protagonist Fund, with a background at Two Sigma, generously shared his comprehensive technological and business strategy expertise across all three parts.

• Brandon Cui, a Machine Learning Engineer at Mosaic ML (acq. Databricks) with a rich background at Meta AI (FAIR), significantly enhanced Part I and Part 2 of my essay. His deep understanding of machine learning from a pragmatic perspective enriched the analysis.

• Fan-Yun Sun, a dedicated PhD Student in Computer Science at Stanford University, also played a pivotal role in shaping Part 2. His academic rigor and research acumen added a layer of theoretical depth.

Strategic and Operational Expertise:

• Camden McRae, Co-Founder at Nextwave X Partners, brought valuable entrepreneurial insights to Part 3, adding a dimension of practical wisdom to my discussions.

• Jonathan Chang, XIR at Brex, shared his expertise in scaling companies and operational strategies, significantly contributing to Part 3.

Application Layer Perspectives:

• David Salib, CEO at Minvo and with extensive experience at Google, Lyft, Playstation, and Yelp, also contributed to Part I. His industry experience offered a pragmatic view of application strategies.

• Maneesh Apte, CEO at Real Simple Labs with his experience at ThoughtSpot, helped contribute to a nuanced view on Part I, especially on considerations for the case study.

• Usman Hanif, a Master's Student in Computer Science at Stanford University, provided fresh academic insights for Part I, enriching my startup strategy discussions.

Editorial Review:

• Samuel Wheeler, a Web3 Developer, provided essential grammar and copywriting expertise, ensuring clarity and coherence throughout the essay.

• Zoe Enright, a Consultant at Clearview Healthcare Partners, provided the needed structural guidance to make this essay clear and more approachable for all audiences, even those not interested in Generative AI!

 These individuals' collective wisdom and experience have been pivotal in shaping my essay. I am deeply grateful for their invaluable contributions and the unique perspectives they brought to this work.

References:

• 7 Powers Summary

• Practical Tips for Finetuning LLMs

• Is OpenAI the Next Netscape?

• Business Report of AI 2023 – Kenny So

• Who Will Win The GTM AI Race: Startups or Incumbents (SaaStr 2023 Annual Presentation)

• An Approachable Primer to the Layers of AI

•  Introducing TUNA: A Tool for Rapidly Generating Synthetic Fine-tuning Datasets

• The State of AI in 2023: Generative AI's Breakout Year

• State of AI 2023 Report

• Personalization in AI

• Navigating the High Cost of AI Compute

• Fine-tune LLAMA-2 with Limited Resources

• Consideration on Getting Started Now with Finetuning Large Language Models

• The Battle for Digital Disruption: Startups vs Incumbents