SaaS in the age of Generative AI
As the ChatGPT-fuelled AI zeitgeist sweeps through the world, how will it impact the world of SaaS? How will Indian SaaS startups adapt to this new era?
Where are the emergent markets? While this technology platform shift will throw up enormous opportunities, the traditional Indian SaaS playbook has to be adapted to the new AI order. Founders need to capitalize on this generational shift by adopting a first-principles foundational mindset coupled with rapid and bold execution.
If there is one term that has captured the imagination of the entire world over the recent past, that term is undoubtedly “Generative AI”. While artificial intelligence and machine learning have been topics of tech research for decades, it is generative AI that has heralded the “Age of AI” — a transformational moment that is radically changing the way that software is created and consumed.
And nothing captures this “The Age of Artificial Intelligence” zeitgeist better than the rise of ChatGPT.
It’s been about six months since ChatGPT was released as a consumer product. A lot has happened in a frighteningly short span of time. A hundred million people have used the product, it has made its way onto every software product roadmap, reignited the long-dormant search wars and arguably destabilized the most dominant company of the internet era, Google unleashed a flurry of investment dollars in an otherwise bearish funding environment and left pretty much every company grappling with how it will alter the fundamentals of their business and the future of work.
We are at the beginning of a tectonic platform shift in technology. If Apple’s iPhone launch was the epochal moment of the mobile revolution, the launch of ChatGPT is AI’s iPhone moment. There has never been anything to parallel the pace and momentum with which it is capturing the world’s imagination.
ChatGPT took 5 days from launch to get to 1 million users and is the quickest app in history to reach 100 million users. It is changing the way the world works, creates content and searches for online information. ChatGPT’s actual impact could possibly surpass even the biggest technology paradigm shifts such as the cloud and mobile in terms of scale, scope and speed. A recent report by Goldman Sachs predicted that generative AI could raise global GDP by as much as 7% over the next few years. To appreciate how huge that impact is, consider that the personal computer’s contribution to GDP was 0.006%. AI could potentially have the type of global impact that electricity and the steam engine had in terms of productivity changes.
To understand why this hype is justified, we need to pay attention to three non-obvious facets of ChatGPT’s rise.
First, ChatGPT’s main innovation is the user interface or rather the lack/simplicity of it. That might seem incongruous given the fact that there is arguably nothing new or innovative about a chat messaging UI as such and chatbots, both text-based as well as voice-based (Siri, Alexa etc) have been around for a while. But as we now see retrospectively, the combination of a simple universally-familiar user interface abstracting away the power of GPT contributed to ChatGPT’s massive adoption. Furthermore, it has set a benchmark for the ease of use and simplicity that users will increasingly expect from all their applications. It is changing end-user expectations around automation and app user experience
Secondly, the biggest indicator that AI, in general, is here to stay is the universal excitement and enthusiasm with which the world, in general, has adopted ChatGPT — not only is social media abuzz with interesting uses and applications of ChatGPT and the broad ecosystem that is rapidly coalescing around the core foundational models, every company from small startups to behemoths are actively exploring ways to incorporate AI into their offerings. Lock in step, the amount of funding that is looking to back generative AI companies has gone to stratospheric levels. Analysts at research firm PitchBook predict that venture investment in generative AI companies will easily be several times last year’s level of $4.5 billion.
Finally, there is no doubt that the success of ChatGPT establishes AI as an idea whose “time has come” — a perfect storm of market dynamics and technological progress. As the chart below illustrates, the underlying technology foundations have improved steadily for many years but have reached a tipping point in the last two years with a sharp increase of capabilities powered by progress in both algorithmic as well as compute models.
The tipping point of AI performance
And to think, we are still just at the beginning of the age of AI — the rate of growth from this point could well be exponential over the next few years completely changing every facet of work and life. This seismic shift is a golden opportunity for nimble and ambitious companies to create iconic companies.
“There are decades when nothing happens, and there are weeks when decades happen.”
This famous quote is particularly applicable to the way in which AI is changing the world at the moment. While every technology wave opens up opportunities to create disruptive companies, the AI age is fundamentally different from previous waves like the mobile and the cloud.
How so?
While previous advances such as mobile and cloud computing followed an adoption curve that was limited to enthusiastic tinkerers and “nerds” in the early days, generative AI was useful right out of the gate for a much broader audience of “ordinary” people. The fact that the value of the answers from ChatGPT is instant and obvious and in many cases, astonishing enough to meet the high threshold for “advanced science that is indistinguishable from magic”. People across industries and job functions are tinkering with ChatGTP in myriad ways and sharing ways in which the tool has massively improved a specific job to be done. Everyone from school kids looking for homework help to grandmothers seeking cooking recipes is relishing the value of ChatGPT. There is possibly no precedent in terms of a technology advancement that has had so many relevant use cases for over a billion people within a few months of its release and equally importantly, for free.
While the first wave of generative AI applications resembles the mobile application landscape when the iPhone first came out in terms of being gimmicky and thin with unclear competitive differentiation and business models, the value of many of these applications is obvious and immediate. From writing creative marketing copy to generating homework to conjuring up stunning images from just a text prompt. Even though it is early days, these applications provide an intriguing glimpse into what the future may hold. Once you see a system produce detailed blog posts or complex code faster and with less effort than you could believe was possible, it’s hard to imagine going back to the “old” ways of how we used to work and live.
Additionally, there was a long-standing belief that AI will first automate manual and repetitive tasks such as data entry and other relatively simple tasks but robotics turned out to be harder than some parts of cognitive knowledge work. The spread of ChatGPT and generative AI has demonstrated that it will fundamentally transform every form of knowledge work starting top-down from “high-value” jobs such as software development, product management and marketing. The approximate nature of generative models makes them better than expected at creative work. Generative AI will fundamentally change every business activity from management consulting to movie production to customer support. PitchBook estimates the market for such AI applications in enterprise technology alone will rise to $98 billion in 2026. The remarkable advancements AI is going to make in the coming years with advancements in this field not growing linearly henceforth, but hockey-stick curves that will wipe out entire professions and careers and create entirely new categories of businesses and jobs. Early evidence of this shift is already visible. IBM announced that it will cut 7,800 jobs that will be automated by AI. Coca-Cola just released an advertisement that was created entirely by AI. So while there is hype around AI, the real-world impact and implications are already visible.
It is therefore essential to adopt a “first principles” approach towards parsing and understanding the opportunities ahead. The ability to adapt nimbly and take advantage of emergent opportunities will be key. Far-fetched scenarios that seemed like science fiction just a few months back have not merely entered the realm of the possible, they seem all but inevitable now. AI will play a key role in shaping not just how enterprises work but also how humans work with the software itself.
This shared epiphany has invigorated tech ecosystems all over the globe from the Bay Area to Bangalore. From startups to big tech, everyone is sprinting.
So where are the opportunities? Who has a good chance to win the race? Incumbents or startups?
To answer these questions, we need to first parse the overall landscape.
There are two broad layers in the AI ecosystem.
The first is the “Model” layer. This includes all the infrastructure needed to build the foundational layer of AI — it includes databases, networking, & compute.
The second is the “Application” layer. These applications learn and generate content, work, and emulate actions for many tasks.
This landscape map by Sequoia Capital provides a succinct overview of the categories and dimensions of the model and application layers.
One aspect that stands out in this map is that the model layer pits companies against all the major tech behemoths — Microsoft (OpenAI and Azure), Google (Bard. PaLM), Amazon (Sagemaker), and Facebook. This is not surprising given the nature and size of the opportunity before us — it is essentially an opportunity to become to the AI ecosystem what AWS (Amazon Web Services) became to the cloud. Occupying a central role in the foundational layer will enable the winner to basically extract an “AI tax” for every transaction built through these engines. This “AI Tax” will decrease over time as the costs around foundation models reduce with scale and improvement in base metal GPU performance — the likes of OpenAI have already dropped costs significantly in the last 6 months. This dynamic makes it even more difficult to establish a competitive moat in this layer — playing at this table requires billions of dollars in funding and decades-long gestation periods and is therefore likely to be outside the purview of all but a few well-funded startups (the likes of OpenAI, Cohere, Anthropic, Stability.ai) that have raised hundreds of millions of dollars in funding with the knowledge that the base models and hence the business model can potentially be a commoditization play with a race to the bottom.
The immediate opportunity for startups is therefore in the application layer where companies like Copy.ai, Jasper, Midjourney and Runway have established early leads and built meaningful businesses and brands. While it is tempting to believe that these companies are going to remain leaders in the future as well, it bears notice that many, if not most, of these companies are equivalent to the early “toy” apps like flashlights and cat animations that were popular in the early days of the mobile era. The real opportunity is in front of us.
The only question is how a startup should think about competing in this age of AI — how should they pick a market and domain, how can they compete against incumbents and other startups, and how can they build a competitive advantage or moat?
With every major technology platform shift, many legacy companies get disrupted as they get caught flat-footed and are slow to respond to changes around them. Traditionally, startups had one key advantage against incumbents — speed — the ability to move quickly and nimbly. However, in the case of gen AI, speed doesn’t seem to be an advantage for startups.
Why so?
Previous technology platform shifts such as mobile and cloud computing were innovations that required incumbents to make large-scale changes, not just to their technology stack but also to their business model. Re-architecting and transforming a desktop application to a web application or mobile app was a resource-heavy and time-intensive effort. However, generative AI can be added to an existing product in a matter of days by integrating a simple API call without changing its tech architecture. As a result, it comes as no surprise that Microsoft was able to integrate generative AI into Microsoft Office and their Edge browser within months of the release of ChatGPT. Other behemoths such as Google and Adobe have made similar integrations within their apps just as quickly. As have large startups such as Salesforce, Notion and Airtable. The bottom line is that incumbents are able to move just as fast as startups in the AI race.
Also unlike previous platform shifts, large companies are not ignoring or underestimating the disruptive potential of AI. Every large technology company is actively incorporating AI across their product and even in places where they may have been previously short-sighted.
Beyond this, incumbents have the advantage of distribution and data — two key weapons of the AI battle. Incumbents can leverage existing data on which they can train their models or build industry-specific models or even customer-specific models. Unlike the cloud or mobile revolutions where new channels had to be built from scratch, incumbents can leverage their existing GTM and distribution strengths to reach customers. For instance, Google has as many as nine products with more than a billion users each — they don’t need to reinvent the wheel to expose these vast audiences to AI capabilities within their products.
If this was a boxing match, it wouldn’t be wrong to say that the first round has definitely been won by incumbents.
But the battle is not yet lost for startups.
While incumbents have an advantage in the first round, their victories might have come with strings attached.
Take Google for instance. In a world where content can be generated instantly and infinitely by AI, how will SEO change? When users can find answers to their questions by chatting with AI, will they even click on online advertisements — if not, how will Google’s formidable SEM advertising business change? While the answers to these questions are moot, there is no doubt that Google will have to radically transform and reassess their business model to defend its legacy revenue streams. It is a tricky balance as making the wrong move can cannibalize existing businesses while not making any change at all is a sure-shot recipe for disaster.
Similarly, take Microsoft. In a world where a single person can create as much content as ten people would previously, does a seat-based pricing model make sense? Also, while layering AI features on top of existing products is an easy win, will it be enough for them to protect their turf? It is one thing to have AI as a feature within a larger product but it is entirely another to rethink and reimagine products in an AI-native form to fully capitalize on the benefits of generative AI. Will the likes of Salesforce, SAP, Workday and other incumbents be bold enough to take such radical steps to completely overhaul their legacy products?
Given this context, there are numerous opportunities for SaaS startups to innovate and win.
Let’s explore some of these opportunities.
Early AI-native applications like Jasper.ai and Copy.ai were widely perceived as being niche apps, mere “thin wrappers” around OpenAI’s GPT models, that catered to a small audience of freelance marketers. It was just a matter of time before these apps would stagnate or be killed by larger apps that incorporate these marketing copy creation capabilities within a broader product. But contrary to these perceptions, these startups have grown and prospered reaching tens of millions of dollars in ARR and even unicorn-level billion-dollar valuations. They arguably achieved this the old-fashioned SaaS way — by offering a well-crafted product with a clear and valuable value proposition to a targeted audience. AI might have changed the landscape of SaaS in many ways but the core ingredients for success don’t seem to be too different from the old ways.
Other AI-native apps such as Midjourney and Runway have also carved out clear winning positions for themselves. Midjourney is an 11-member bootstrapped startup that defies many conventions — its image creation capabilities are available only through a single Discord channel but it has managed to attract millions of paying customers and is said to be well over $100 million in ARR. Similarly, Runway is a video generation app that has attracted attention and audiences — including Hollywood studios that have used their technology to win Oscars. Both these startups also demonstrate the value of building a loyal community beyond building a clearly-positioned brand.
Workflow automation is not a new idea but the generative AI boom has reinvigorated the concept by breathing new life into the idea. AI-driven automation will enable entirely new capabilities in areas which were not possible with previous-generation technology. They will disrupt other workflows and make possible new ones that can potentially increase productivity by 10x or more. The advent of AutoGPT and ChatGPT plugins has enabled the creation of a new ecosystem of workflows and integrations that can now power cross-application communication and end-to-end business use cases.
Autonomous agents are programs, powered by AI, that when given an objective are able to create tasks for themselves, complete tasks, eliminate bottlenecks, reprioritize their task lists and loop by themselves until their objective is reached. They can range from specialized functions such as lead generation workflow or can take up a much broader role similar to all the tasks an SDR (sales development representative) would perform. AutoGPT has surpassed 100K stars on GitHub, making it the fastest-growing open source repository ever and several startups such as Inflection AI and Adept have emerged as early leaders in this space, attracting tens of millions of dollars in funding with ambitions to offer “intelligent omnipresent companions” to millions of users over the next few years. Langchain, a popular open-source framework for building with LLMs, has built integrations to agent projects like BabyAGI, CAMEL and AutoGPT and all these are available as loosely-coupled frameworks that startups can innovate on top of. While these systems are already powerful, the ultimate aim is to allow anyone to merely issue a simple instruction such as “Book me a flight from Bangalore to San Francisco and a hotel room for my stay” and have an army of personal autonomous agents perform the task for you. Imagine how many apps and platforms such a workflow will disrupt. While LLMs made us rethink everything from marketing to medical research to software development, and even what it means to be creative, agents will multiplex this impact, redefining how we interact with technology and with each other, opening up new vectors of possibilities and opportunities for nimble startups to leap-frog previous generations.
One of the industries that are likely to see the biggest impact from AI is the world of software development and programming itself. It might seem counterintuitive at first glance but in hindsight, it should not be surprising at all. A large corpus of open-source software code has been publicly available for LLMs like GPT-4 to ingest and index. This is why AI coding assistants like Github’s Copilot have proven themselves to be excellent augmentation tools for developers. By some estimates, Copilot is already automating 40–50% of code for many developers by abstracting away boilerplate code from manual development. We are facing a future where the output of software engineers is likely to increase ten-fold within the next decade — imagine a world where every developer is a 10X engineer! OpenAI’s ChatGPT can already pass Google’s exam for a high-level software engineer and the scores are continuously increasing.
In this brave new world, every layer of the dev tools and infrastructure stack is up for grabs. New AI layers will be added to the stack ranging from LLMOps and ML workflows to AI-driven software testing and automation. Given the impact of LLM costs on gross margins, every SaaS firm is likely to want to adopt frameworks and tools that will help control and optimize these new backend costs. ChatGPT’s Code Interpreter and ChatGPT plugin’s drop-in API architecture will add new layers around the capabilities of software solutions and the surrounding ecosystem.
The world of programming itself is likely to see major changes over the next decade with a lower bar to entry and a large set of hobby developers emerging — folks who don’t necessarily have extensive experience in coding but can leverage coding assistant tools to whip up single-use or temporary software solutions. Roles around data analysis and visualization are also likely to be democratized by AI.
How will the emergence of generative AI impact the world of IT services, an area that India has been traditionally strong at? At one end of the spectrum, there is a belief that outsourcing will be negatively impacted by AI with automation replacing manual workers. While that may well be true to a limited extent, most large enterprise customers have concerns and considerations around security, privacy and reliability which might not be achievable with a hallucination-prone AI solution. On the contrary, at the other end of the spectrum, there are a number of new opportunities around consulting and digital transformation that will emerge around enterprises seeking to adopt AI.
These enterprise-level use cases are potentially a large business opportunity for services-led startups that can help customers to leverage existing data sets and develop specialized LLM or fine-tuned models. Many Indian companies, small and large, are already working on transformation projects and developing solutions in the AI and Intelligent Automation (IA) space, assisting enterprises to meet demand, improve efficiency, and implement smarter business models.
Vertical SaaS is a nascent area that has seen tremendous growth over the past few years — these startups focus exclusively on a single industry or domain and build deep solutions tailored to their specific needs. Vertical SaaS startups such as Veeva have built large businesses adopting this approach. AI is likely to have a major impact on vertical SaaS — areas such as healthcare and industrial manufacturing will see major gains. Healthcare is seeing the emergence of vertical LLM service providers such as Hippocratic while manufacturing companies will benefit from generative AI’s strengths around supply chain planning and simulations.
In fact, there is not going to be any vertical Saas domain that is unlikely to be unaffected by the emergence of generative AI. Startups such as Harvey.ai and EvenUp are revolutionizing the legal industry by automating contract analysis, due diligence, litigation and regulatory compliance and augmenting analysis and synthesis that previously required manual labour. Similarly, models such as BloombergGPT are emerging in the financial services sector. Vertical saas players have an inbuilt advantage by virtue of having access to proprietary data that can be used to build fine-tuned models that offer better results compared to other competitors.
The end game for vertical SaaS is also likely to be very interesting — when software development costs trend towards zero, innovative startups can target and build for $10m TAMs instead of $100 million or $1 billion ones. SaaS solutions that were traditionally perceived as being subscale or VC-unfundable can deliver positive unit economics and efficient business models with very small teams.
While all of the above are specific opportunities, there is possibly a broader opportunity around UI (user interface) and UX (user experience) that could potentially change every SaaS application that exists today. Many, if not most, SaaS applications use the familiar paradigm of “forms” that help users view, interact with and update data from some backend data source. This common UX pattern can be seen across every major SaaS domain from CRM to support and from project management software to ERP.
These form-based interfaces might have reached their expiry dates.
ChatGPT is an early indicator of how AI will revolutionize UX. A chatbox where the user can just ask the system what he wants using natural language, textual or verbal, is a major step up from clunky forms. These affordances are much simpler and more powerful than manually entering data into a form. Imagine a world where every SaaS application is front-ended by a chat interface where you can just tell the application what you want it to do.
But this is just the beginning.
Generative AI will usher in a world of hyper-personalized UX and possibly even screenless UI where users interact with other modalities. In fact, LLMs can themselves self-generate the interface at run time as apps such as Perplexity.ai are demonstrating already. Others such as Humane are facilitating effortless navigation in unfamiliar environments along with providing personalized recommendations and seamless communication across all languages aided by the power of AI.
AI is fundamentally changing the mental models that we have built around SaaS UI and UX over the last two decades. Startups that win by innovating on UX in this new era of invisible interfaces and AI-enhanced experiences can emerge as big winners in the future.
AI will fundamentally transform every form of work. Every company will be an AI company. Or risk getting disrupted by one.
The new world will create breakthrough opportunities for startups to build completely new AI-native solutions, new form factors, new business models and even entirely new software categories. Traditionally, Indian SaaS startups have adopted the “fast follower” approach — they benchmark against a successful US SaaS startup and offer a comparable solution at a lower cost but higher value. In the age of AI, this approach is not likely to work. The pace of change is exponential and entire markets could be created and captured by early pioneers in a matter of weeks if not months. These pioneers could even be incumbents who have recognized this generational opportunity and have bet the farm on AI being a critical component of every aspect of work and play in the near future. While there are going to be countless opportunities for new players, the key for Indian SaaS startups is going to be in terms of thinking from first principles to leverage AI to innovate in fundamentally different ways and execute quickly and boldly.
There has never been a better time to build a SaaS startup and especially so, from India.
Let’s go and win the age of AI.
By Manav Garg
