Original title: The Great Export: A New Era of Consumer Software

Source: Olivia Moore, a16z Partner
Original compilation, compilation: Leo, Deep Mind

DO YOU EVER WONDER WHY THE AI CONSUMER PRODUCTS THAT HAVE EMERGED IN THE LAST TWO YEARS HAVE GROWN FROM ZERO TO MILLIONS OF USERS IN LESS THAN TWO YEARS, WITH ANNUAL REVENUES BREAKING $100 MILLION? THIS GROWTH WAS ALMOST UNIMAGINABLE BEFORE AI. ON THE SURFACE, THIS IS DUE TO FASTER DISTRIBUTION AND HIGHER AVERAGE INCOME FOR USERS. BUT I FOUND A DEEPER CHANGE THAT MOST PEOPLE IGNORED: AI COMPLETELY CHANGED THE REVENUE RETENTION PATTERNS OF CONSUMER SOFTWARE。

Recently read an analytical article by a16z partner Olivia Moore, The Great Export: A New Era of Consumer Software, which she calls "Great Extension", I think she captures a very critical trend. After a deep reflection on this point of view, I found that it was not just an adaptation of business models, but a fundamental change in the rules of the whole consumer software industry. We are witnessing a historic turning point: consumer-level software companies no longer need to fight the loss of users, but can rely on the continued expansion of user values to achieve growth。Consumer and enterprise markets are becoming increasingly blurred in a sense。

THE IMPACT OF THIS CHANGE IS ENORMOUS. TRADITIONAL CONSUMER SOFTWARE COMPANIES SPEND A GREAT DEAL OF EFFORT AND MONEY EACH YEAR TO REPLACE LOST USERS SIMPLY TO MAINTAIN THE STATUS QUO. AND NOW, THOSE COMPANIES THAT HAVE SEIZED AI'S OPPORTUNITIES FIND THAT NOT ONLY DO THEY NOT LOSE VALUE, BUT RATHER CONTRIBUTE MORE OVER TIME. IT'S LIKE TURNING FROM A LEAKING BUCKET TO A BALLOON THAT'S SWELLING, AND IT'S GROWING IN A COMPLETELY DIFFERENT WAY。

From this perspective, I personally think that this is a huge opportunity for the offshore companies, because consumer-grade products can grow and earn with PLG, and it's perfect to avoid the fact that the Chinese team is having difficulty overseas with SLG. It's a business market, but the whole growth pattern is similar to the C-end product. And I personally feel that my own project has been on the line for a month, completely for the company's B-end Vibe coding product, but with PLG growth, I get good data feedback。

Fundamental flaws in traditional models

Let's go back to how the consumer software before AI made money. Moore mentioned two main models in her analysis, and I think her summary is accurate. The first is the advertising-driven model, which is mainly used for social applications and is directly linked to usage, so that the value per user is usually flat over time. Instagram, TikTok, Snapchat are all representatives of this pattern. The second is the single-tier subscription model, whereby all fee-paying users pay the same fixed fees per month or year to obtain access to the product. Duolingo, Calm, and YouTube Premium all follow this approach。

under both models, the income retention rate is almost always below 100 per cent. each year there is a percentage loss of users, and those who remain continue to pay the same amount. for consumer-class subscriptions, maintaining 30-40 per cent user and income retention rates at the end of the first year is considered "best practice". this number sounds desperate。

I have always felt that there is a fundamental structural flaw in this model: it creates a basic constraint that companies must constantly replace lost revenues to sustain growth, let alone expand. Imagine, if you had a bucket leaking, you would not only keep the water going to Riga to maintain the water level, but would add more than leaking to raise it. This is the dilemma faced by traditional consumer software companies: They are trapped in a never-ending cycle of customer-lose-retake。

The problem with this model is not only numerical; it also affects the corporate strategy and resource allocation. Much of the effort has been devoted to acquiring new users to compensate for the loss, rather than deepening relationships with existing users or improving product value. That is why we see that many consumer-level applications are crazy about sending notifications and using various means to make users more sticky, because they know that once users stop using, income disappears。

I believe that this model fundamentally underestimates the value potential of users. It assumes that the value of the user is fixed and that once they subscribe to the product, the income they can contribute is overestimated. But the reality is that as users become more familiar with their products, their demand tends to grow and the amount they are willing to pay increases. Traditional models do not capture opportunities for such value growth。

AL TIME GAME REWRITING

The emergence of AI has completely changed this game. Moore calls this change "Great Expansion," which I think is very appropriate. The fastest-growing consumer class AI now sees a retention rate of over 100 percent, which is almost inconceivable in traditional consumer software. This happens in two ways: first, consumer spending increases with the replacement of fixed "visiting" costs with usage-based income; and second, consumers bring tools into the workplace at an unprecedented rate, where they can be reimbursed and supported by a larger budget。

A KEY CHANGE I HAVE OBSERVED IS A FUNDAMENTAL SHIFT IN USER BEHAVIOUR PATTERNS. IN TRADITIONAL SOFTWARE, USERS EITHER USE PRODUCTS OR DO NOT USE THEM; SUBSCRIBE OR CANCEL. IN AI PRODUCTS, HOWEVER, THE PARTICIPATION AND VALUE CONTRIBUTION OF USERS IS INCREMENTAL. THEY MAY START USING BASIC FUNCTIONS ONLY OCCASIONALLY, BUT AS THEY DISCOVER THE VALUE OF AI, THEY WILL INCREASINGLY RELY ON THESE TOOLS AND DEMAND WILL GROW。

The trajectory of this difference is dramatic. Moore mentioned that under the 50 per cent retention rate, the company had to replace half of the user base each year to remain unchanged. And in more than 100 percent of the cases, every user group is expanding, and growth is added to growth. This is not just a numerical improvement; it represents a completely new engine of growth。

I THINK THERE ARE SEVERAL UNDERLYING REASONS BEHIND THIS CHANGE. AI PRODUCTS HAVE A LEARNING EFFECT, AND THEY BECOME MORE USEFUL AS THEY ARE USED. THE MORE TIME AND DATA USERS INVEST, THE GREATER THE VALUE OF THE PRODUCT TO THEM. THIS CREATES A POSITIVE FEEDBACK CYCLE: GREATER USE LEADS TO GREATER VALUE, GREATER VALUE LEADS TO GREATER USE AND GREATER WILLINGNESS TO PAY。

ANOTHER KEY FACTOR IS THE PRACTICAL NATURE OF THE AI PRODUCT. UNLIKE MANY TRADITIONAL CONSUMER-LEVEL APPLICATIONS, AI TOOLS OFTEN DIRECTLY ADDRESS THE SPECIFIC PROBLEMS OF USERS OR INCREASE THEIR PRODUCTIVITY. THIS MEANS THAT USERS CAN EASILY SEE THE DIRECT BENEFITS OF USING THESE TOOLS AND ARE MORE WILLING TO PAY FOR THIS VALUE. WHEN AN AI TOOL SAVES YOU HOURS OF WORKING TIME, IT MAKES IT VERY REASONABLE TO PAY FOR EXTRA USAGE。

Fine pricing architecture design

Let me in-depth on how the most successful consumer-class AI companies build their pricing strategies. Moore notes that these companies no longer rely on a single subscription fee, but rather use a hybrid model that includes multiple tiers of subscription plus user-based components. If users exhaust the creditits they contain, they can buy more or upgrade to higher schemes。

i think there's an important revelation from the game industry. game companies have long earned most of their income from high-consumption whale users. limiting pricing to one or two levels is likely to waste income opportunities. smart companies build tiers around variables such as generation or number of tasks, speed and priority, or access to specific models, while also offering points and upgrade options。

Let me see some concrete examples. Google AI provides $20 per month for Pro subscriptions and $249 per month for Ultra subscriptions, with additional charges for Veo3 credits when users (inevitably) exceed the amount they contain. Additional subcontracts have been extended from $25 to $200. I understand that many users may spend as much on extra Veo credits as basic subscriptions. This is a perfect example of how income increases with increased user participation。

The Krea model is also interesting, as they provide a plan of $10-60 per month, based on expected usage and training operations, to purchase additional subcontracts of $5-40 (effective 90 days) if you exceed the unit of account included. The essence of this model is that it provides both a reasonable entry price for light users and extended space for heavy users。

Grok ' s pricing pushes the strategy to the extreme: SuperGrok plans $30 per month, SuperGrok Heavy plans $300 per month, the latter unlocking new models (Grok 4 Heavy), extended access to models, longer memory and new functionality tests. This 10-fold difference in prices is almost inconceivable in traditional consumer-level software, but it became reasonable in the AI era because of the great differences in demand and value perceptions among different users。

I believe that the success of these models lies in their recognition of the diversity and dynamic nature of user values. Not all users had the same needs or capacity to pay, and the same user ' s needs would change at different times. By providing flexible pricing options, these companies are able to capture the full spectrum of user values。

Moore mentioned that some consumer companies had achieved more than 100 per cent income retention on the basis of this pricing model alone, and had not even considered any expansion to enterprises. This illustrates the strength of this strategy. It not only addresses the loss of traditional consumer-grade software, but also creates built-in growth mechanisms。

Gold Bridges from Consumer to Enterprise

Another important trend I have observed is the unprecedented speed with which consumers bring AI tools into the workplace. Moore emphasized in her analysis that consumers were actively rewarded for introducing AI tools into the workplace. In some companies, failure to become "AI-native" is now considered unacceptable. Any product with potential job applications — basically any product that is not NSFW — should assume that users would want to bring it into their team and that when they can be reimbursed, they would pay significantly more。

I AM IMPRESSED BY THE PACE OF THIS TRANSFORMATION. IN THE PAST, THE TRANSITION FROM THE CONSUMER TO THE ENTERPRISE LEVEL USUALLY TOOK SEVERAL YEARS AND REQUIRED CONSIDERABLE MARKET EDUCATION AND MARKETING EFFORTS. HOWEVER, THE USEFULNESS OF THE AI TOOLS IS SO OBVIOUS THAT USERS AUTOMATICALLY INTRODUCE THEM INTO THE WORK ENVIRONMENT. I'VE SEEN A LOT OF CASES WHERE EMPLOYEES BUY THE AI TOOLS PERSONALLY AND THEN CONVINCE COMPANIES TO BUY THE CORPORATE VERSION FOR THE ENTIRE TEAM。

THE SHIFT FROM PRICE-SENSITIVE CONSUMERS TO NON-PRICE-SENSITIVE BUSINESS BUYERS HAS CREATED TREMENDOUS OPPORTUNITIES FOR EXPANSION. BUT THIS REQUIRES BASIC SHARED AND COLLABORATIVE FUNCTIONS, SUCH AS TEAM FOLDERS, SHARED LIBRARIES, COLLABORATIVE CANVASS, IDENTIFICATION AND SECURITY. I THINK THESE FUNCTIONS ARE NOW A REQUIREMENT FOR ANY CONSUMER-LEVEL AI PRODUCT WITH THE POTENTIAL OF AN ENTERPRISE。

With these functions, price differences can be significant. ChatGPT is a good example, although it is not widely regarded as a team product, but its pricing highlights the difference: an individual subscribes $20 per month, while an enterprise plan ranges from $25 to $60 per user. This two to three-fold price difference is rare in traditional consumer-level software, but became common in the AI era。

I think some companies even price their personal plans as a balance of gain or loss or a slight loss to speed up team adoption. In 2020, Notion effectively used this method to provide unlimited free pages for individual users, while charging radical fees for collaborative functions, which contributed to its most explosive growth period. The logic of this strategy is to build a user base by subsidizing personal use, and then to make a profit through business functions。

Let me see some specific examples. Gamma's Plus plan is $8 per month to remove watermarks, which are required for most enterprises, and other functions. Users then pay for each of the collaborators added to their workspace. This model intelligently exploits business demands for professional appearance。

Refrit provides a $20 per month scheme for Core users. The team plan starts at $35 per month and includes additional credits, viewer seats, centralized charges, role-based access control, private deployment, etc. Cursor provides $20 per month for Pro and $200 per month for Ultra (a 20-fold increase in usage). Team users pay $40 per month for Pro products with an organized privacy model, access and management dashboard, centralized billing and SAML/SSO。

THESE FUNCTIONS ARE IMPORTANT BECAUSE THEY UNLOCK ENTERPRISE-LEVEL ARPU EXPANSION (AVERAGE INCOME PER USER). I THINK ANY CONSUMER-CLASS AI COMPANY NOW IS MISSING A HUGE OPPORTUNITY IF IT DOESN'T CONSIDER THE BUSINESS EXPANSION PATH. IN ADDITION TO PAYING HIGHER FEES, BUSINESS USERS ARE USUALLY MORE STABLE AND HAVE LOWER LOSS RATES。

Investing in enterprise-level capacity from day one

Moore made a seemingly anti-intuitive but actually very sensible suggestion that consumer companies should now consider hiring sales managers within one to two years of being established. I fully agree with that view, although it does run counter to traditional consumer-grade strategies。

INDIVIDUAL ADOPTION ONLY ALLOWS PRODUCTS TO REACH A CERTAIN LEVEL; ENSURING EXTENSIVE ORGANIZATIONAL USE REQUIRES PROCUREMENT BY NAVIGATION COMPANIES AND THE COMPLETION OF HIGH-VALUE CONTRACTS. THIS REQUIRES A PROFESSIONAL MARKETING CAPACITY RATHER THAN SIMPLY RELYING ON THE NATURAL TRANSMISSION OF PRODUCTS. I'VE SEEN TOO MANY GOOD CONSUMER-CLASS PRODUCTS LOSE MAJOR OPPORTUNITIES FOR LACK OF CORPORATE MARKETING。

Founded in 2013, Canva waited for almost seven years to launch its Teams products. Moore noted that in 2025 such delays were no longer feasible. The pace adopted by the Enterprise AI means that if you delay the functioning of the enterprise, competitors will take the opportunity instead. This competitive pressure was considerably accelerated in the AI era, as markets were changing faster than ever before。

I BELIEVE THAT THERE ARE SEVERAL KEY FUNCTIONS THAT OFTEN DETERMINE THE OUTCOME. IN TERMS OF SECURITY AND PRIVACY, SOC-2 COMPLIANCE, SSO/SAML SUPPORT IS REQUIRED. IN TERMS OF OPERATION AND COSTING, ROLE-BASED ACCESS CONTROL AND CENTRALIZATION ARE REQUIRED. IN TERMS OF PRODUCTS, TEAM TEMPLATES, SHARED THEMES AND COLLABORATIVE WORKFLOWS ARE REQUIRED. THESE MAY SOUND FUNDAMENTAL, BUT THEY ARE OFTEN KEY ELEMENTS OF ENTERPRISE PROCUREMENT DECISION-MAKING。

Eleven Labs is a good example: the company started using consumers extensively, but quickly built enterprise-level capacity, added HIPA compliance for its voice and dialogue agents and positioned itself to serve health care and other regulated markets. This rapid business transformation has enabled them to capture high-value business clients rather than relying solely on consumer income。

I'VE OBSERVED AN INTERESTING PHENOMENON: CONSUMER-LEVEL AI COMPANIES THAT INVEST IN ENTERPRISE CAPABILITIES AT AN EARLY STAGE TEND TO BUILD STRONGER MOATS. ONCE AN ENTERPRISE CUSTOMER HAS ADOPTED A TOOL AND INTEGRATED IT INTO THE WORKFLOW, THE SWITCH COSTS ARE HIGH. THIS HAS CREATED STRONGER CUSTOMER VISCOSITY AND MORE PREDICTABLE INCOME FLOWS。

IN ADDITION, BUSINESS CLIENTS PROVIDE VALUABLE PRODUCT FEEDBACK. THEIR NEEDS TEND TO BE MORE COMPLEX, WHICH LEADS TO HIGHER PRODUCT DEVELOPMENT. I'VE SEEN A LOT OF CONSUMER-GRADE AI PRODUCTS DISCOVER NEW PRODUCT ORIENTATIONS AND FUNCTIONAL NEEDS BY SERVING CORPORATE CLIENTS。

I think deeply about this change

After a careful analysis of Moore's views and my own observations, I think what we are witnessing is not just the adaptation of business models, but the rebuilding of the entire infrastructure of the software industry. AI changed not only the capacity of the product, but also the way value is created and captured。

WHAT I FIND INTERESTING IS THAT THIS CHANGE CHALLENGES OUR TRADITIONAL ASSUMPTIONS ABOUT CONSUMER-LEVEL SOFTWARE. CONSUMER-GRADE SOFTWARE HAS LONG BEEN CONSIDERED TO BE NATURALLY LOW, HIGHLY LOST AND DIFFICULT TO MONETIZE. HOWEVER, AI-ERA REALITIES SHOW THAT CONSUMER-LEVEL SOFTWARE CAN ACHIEVE ENTERPRISE-LEVEL INCOME SIZE AND GROWTH. THE IMPLICATIONS OF THIS TRANSFORMATION ARE FAR-REACHING。

FROM A CAPITAL ALLOCATION POINT OF VIEW, THIS MEANS THAT INVESTORS CAN NOW INVEST MORE MONEY EARLIER IN CONSUMER AI COMPANIES, WHICH ARE ABLE TO ACHIEVE MEANINGFUL INCOME SCALE FASTER. TRADITIONALLY, CONSUMER-LEVEL SOFTWARE COMPANIES HAD TO WAIT UNTIL THEY REACHED LARGE USER SIZES BEFORE EFFECTIVELY MONETIZING, BUT THEY COULD NOW ACHIEVE STRONG INCOME GROWTH BASED ON RELATIVELY SMALL USERS。

I also think about the impact of this change on entrepreneurship strategies. Moore mentioned that we believe that many of the most important companies in the AI era may have started with consumer-grade products. I think it's a very deep insight. Traditional B2B software entrepreneurship paths typically involve a large number of market research, customer interviews and distribution cycles. The path from the consumer stage allows for faster product overlay and market validation。

Another advantage of this approach is that it creates more natural product-market convergence. This is a strong product-market convergence signal when consumers voluntarily use and pay for products. Then, when these users bring their products into the workplace, their adoption becomes more organic and sustainable。

I ALSO NOTED AN INTERESTING CHANGE IN COMPETITION DYNAMICS. IN THE TRADITIONAL SOFTWARE AGE, CONSUMER- AND ENTERPRISE-LEVEL MARKETS ARE USUALLY SEGREGATED, WITH DIFFERENT PLAYERS AND STRATEGIES. BUT IN THE AI ERA, THESE BOUNDARIES BECAME BLURRED. ONE PRODUCT CAN COMPETE SIMULTANEOUSLY IN TWO MARKETS, CREATING NEW COMPETITIVE ADVANTAGES AND CHALLENGES。

FROM A TECHNICAL POINT OF VIEW, I THINK THAT THE DUAL NATURE OF THE AI PRODUCT (CONSUMPTION-LEVEL EASE PLUS ENTERPRISE-LEVEL FUNCTIONALITY) PROMOTES NEW STANDARDS FOR PRODUCT DESIGN AND DEVELOPMENT. PRODUCTS NEED TO BE SIMPLE ENOUGH TO ENABLE INDIVIDUAL USERS TO WORK EASILY, BUT ALSO SUFFICIENTLY STRONG AND SECURE TO MEET BUSINESS NEEDS. THIS BALANCE IS NOT EASY TO ACHIEVE, BUT THOSE THAT DO WELL WILL HAVE A GREAT COMPETITIVE ADVANTAGE。

I'VE BEEN THINKING ABOUT THE IMPACT OF THIS TREND ON EXISTING SOFTWARE COMPANIES. TRADITIONAL ENTERPRISE SOFTWARE COMPANIES ARE NOW FACING COMPETITION FROM CONSUMER-CLASS START-UP AI COMPANIES, WHICH OFTEN HAVE BETTER USER EXPERIENCE AND FASTER ITERATIVE SPEED. THIS MAY FORCE THE ENTIRE ENTERPRISE SOFTWARE INDUSTRY TO IMPROVE ITS PRODUCT STANDARDS AND USER EXPERIENCE。

FINALLY, I BELIEVE THAT THIS CHANGE ALSO REFLECTS A FUNDAMENTAL SHIFT IN WORKING METHODS. REMOTE WORK, INCREASED CHOICE OF PERSONAL TOOLS AND HIGHER EXPECTATIONS ABOUT PRODUCTIVITY TOOLS HAVE CONTRIBUTED TO BLURRING THE BOUNDARIES BETWEEN CONSUMER AND ENTERPRISE-LEVEL TOOLS. AI JUST SPEEDED UP THE TREND。

Opportunities and challenges for the future

While I am excited about the "Great Express" phenomenon described by Moore, I also see some challenges and opportunities that require attention。

In terms of challenges, I believe that competition will become more intense. When successful paths become clear, more companies try to follow the same strategy. Firms that are able to build strong differences and network effects will win in long-term competition。

FROM A REGULATORY PERSPECTIVE, THE RAPID ADOPTION OF AI PRODUCTS IN AN ENTERPRISE ENVIRONMENT MAY POSE NEW COMPLIANCE AND SECURITY CHALLENGES. COMPANIES NEED TO ENSURE THAT THEIR AI TOOLS MEET INDUSTRY STANDARDS AND REGULATORY REQUIREMENTS. THIS MAY INCREASE DEVELOPMENT COSTS AND COMPLEXITIES, BUT MAY ALSO CREATE NEW BARRIERS TO COMPETITION。

IN TERMS OF OPPORTUNITIES, I SEE A HUGE ROOM FOR INNOVATION. COMPANIES THAT CAN CREATIVELY COMBINE CONSUMER-LEVEL EASE AND ENTERPRISE-LEVEL FUNCTIONS WILL OPEN UP NEW MARKET CATEGORIES. I ALSO BELIEVE THERE IS A GREAT OPPORTUNITY FOR VERTICALLY-ORIENTED AI TOOLS TO BE MORE VALUABLE THAN UNIVERSAL TOOLS FOR INDUSTRY-SPECIFIC OR CASE-BASED IN-DEPTH OPTIMIZATION。

I ALSO SAW THE DATA AND THE NETWORK EFFECTS OF THE AI MODEL. AS USERS INCREASE AND USE FURTHER, AI PRODUCTS CAN BECOME MORE INTELLIGENT AND PERSONAL. SUCH DATA-DRIVEN IMPROVEMENTS CAN CREATE STRONG COMPETITIVE ADVANTAGES, AS IT IS DIFFICULT FOR NEW ENTRANTS TO REPLICATE THIS ACCUMULATED INTELLIGENCE。

From an investment perspective, I believe that this trend will continue to attract significant capital. But investors need to better identify those firms that have a real and sustainable competitive advantage, not just those that are growing rapidly in the short term. The key will be to understand which companies can build real moats and not just take advantage of early market opportunities。

In the end, I believe that Moore's description of "Great Exchange" is just the beginning of the AI revolution. We are redefining the nature of the software — from tools to smart partners, from functionality to results. Those companies that can capture this transformation and successfully implement it will build the next generation of technological giants. This is not only an innovation in business models, but also a rethinking of the human-technology relationship. We are in an exciting era, and software is becoming more intelligent, useful and indispensable。

Original Link