SaaS 4.0: Say Hello to The Era of Cognition

It’s already cliché to say that AI is changing everything, faster than we could possibly have imagined even a year ago. Some of its most profound impacts will be felt in enterprise software, making this a unique time to be building.

Generative AI tools like ChatGPT, Google Bard, Bing Chat, Midjourney, Stable Diffusion, and others are already changing the face of software. This is causing a seismic upheaval in how we think about the tools we use each day to perform our jobs and generating a lot of excitement for what lies ahead.

At Lightspeed, we believe every software company will integrate generative AI into their stack. The distinction will eventually go away as every SaaS application becomes AI native. As a firm we have been backing entrepreneurs at the forefront of AI for more than a decade, and we’re excited about the opportunities this new paradigm shift will enable. Generative AI will change the types of tasks we ask software to perform and how SaaS applications get built. We’ll see new business models, massive gains in productivity and efficiency, greater personalization, and an explosion of new creative content.

But these tectonic shifts will also raise serious questions around ethical and legal implications, data sovereignty and privacy, creativity, intellectual property, and the nature of work itself; issues that we are only just starting to grapple with.

When Salesforce first emerged in 1999, it was a revolution. The idea that line-of-business software could be delivered via the web and provisioned in a multi-tenant solution offered businesses the ability to quickly scale as needed. Salesforce pioneered both a new technology model (cloud-based computing) and a new business model (recurring licenses vs. one-time perpetual software). SaaS applications offered lower upfront costs, easier set-up and deployment, and more frequent software upgrades. Over time, this movement spawned an entire generation of market-defining companies, including Hubspot, Marketo, Workday, and Zendesk.

Once systems of record were cloud-based, the next obvious frontier was activating data stored in these systems to engage customers directly for customer service, sales, marketing, and other front-office use cases. The emergence of smartphones during this period opened up new channels such as mobile and chat. Internal collaboration apps emerged that made knowledge workers that much more productive. Companies like MuleSoft, Slack, and Twilio embodied these systems of engagement.

As enterprises generated increasing volumes of data and began applying analytics, SaaS applications morphed into early systems of intelligence. Vendors began embedding AI capabilities into their applications, enabling greater productivity and automation. Suddenly, enterprises were able to scale processes and drive new insights at a rate previously not possible. Companies like Gong, Intercom, Monday, and Moveworks leveraged automation to drive greater value for customers.

Thanks to the astonishing growth in the capabilities of generative AI, we believe SaaS is now entering its fourth generation: a system of cognition. Generative AI will permeate every SaaS application. Incumbents and startups alike are sprinting to incorporate these capabilities into their products. In this new era, upwards of 50% of a knowledge worker’s job will be exposed to AI that can augment/supplement their existing workflows, depending on the task.

These are very early days. As we look ahead, we recognize that generative AI will fundamentally transform software in every dimension. We are already seeing a democratization of creativity, with generative AI able to draft poetry, compose songs, and even write jokes. This will likely unlock a new era of personalization that can be 100x more specific to each user.

TikTok already uses data to determine which reels to load next to maximize engagement. What if it could conjure a bespoke AI-generated video to evoke a particular emotional response in each user?

Pricing models in SaaS will also evolve. They will no longer be based on seats but on the value of the service, the level of personalization, the depth of engagement, and the amount of work generated. Database and application workflows will continue to converge, further blurring the lines between the stack. New business models will arise as a result of this innovation.

The truth is, no one yet knows how all of this will play out. But we do know one thing: deeply entrenched incumbents are aware of the changes underway, and they are not standing still. Microsoft’s integration of large language models into its cloud office suite, Microsoft 365 Copilot, is just one example. Salesforce’s Einstein GPT is another. And Adobe’s introduction of Firefly is yet another.

Having met with hundreds of SaaS entrepreneurs recently, here are a few criteria we are thinking through at Lightspeed when evaluating investments. We’re sharing this to generate more discussion and to make our conversations with founders that much more engaging.

• Proprietary data. Large language models (LLMs) today are trained on public data sources. Startups that are able to tap into proprietary datasets will have an opportunity to create differentiation. Importantly, helping companies leverage their own first-party data will also become increasingly valuable.
• Frequency of interaction. Historically, successful software companies built dedicated workflows via scalable products and leveraged go-to-market expertise to get those products in the hands of customers. Gen AI will supercharge this process in terms of what can be automated. Frequency of interaction will matter a lot. We see great value in companies that are able to tap into workflows that recur multiple times a day or week, versus workflows that are more episodic or infrequent.
• Depth of UI. We are seeing a lot of startups where their application is a thin wrapper around an existing application. Our working hypothesis right now is that there is more value to be created from startups that are offering greater heft or gravitational pull in their own applications, while integrating into other systems where needed. A customer front end to ChatGPT alone will not be defensible long-term
• 10x value proposition. IT decision makers already have too many applications to choose from. Given the number of new companies being built around generative AI, that problem is about to get a whole lot worse. Founders able to deliver a 10x better value prop — whether around price, customer experience, or productizing an existing manual/cumbersome workflow — will see greater success in breaking through the noise. In our experience, defensibility will also come from a significantly better product experience, a narrow focus on a big problem, or a feedback loop that is somewhat proprietary.
• Low-level vs. high-level intelligence. Early experience with generative AI shows us that LLMs will be able to automate substantial parts of our work. High-level intelligence — defined as repeatable work that drives significant value to a company — will be more attractive to customers long-term than automating simpler tasks that don’t have immediate or proven ROI.

We are actively engaged in conversations about SaaS 4.0 and hope you’ll consider becoming a part of them. If you’d like to offer your own perspective on this, please reach out on email, Twitter, or LinkedIn.