Article

From Automation to Innovation: How AI in SaaS is Changing the Tech Landscape

by Vanessa Larco and Juan MogollonApr 24, 2023

Generative AI.

It’s arguably the buzziest tech term of the past year. But as we’ve watched AI technology evolve and become more sophisticated, it has become clear that what we’re experiencing is more than a passing fad. Powered by foundation models, Generative AI is the latest era of AI/ML that is unlocking new opportunities and tackling previously unaddressable challenges.

With the introduction of every AI era comes new opportunities for iconic companies to emerge—we believe this current wave of Artificial Intelligence is no exception. But it's not without nuance. When considering the opportunities unlocked by this technology wave, founders must also consider the impact of market dynamics as well as the risks and value propositions of what they build. It’s the companies that get this combination right that will emerge as winners.

Throughout the early 2000s and even before, Neural Networks enabled technology that could recognize handwriting and classify basic images and other unstructured data. In the early 2010s, Deep Neural Networks enabled face and speech recognition, driver assist technologies (aka self-driving), and more accurate predictions for scenarios ranging from weather to customer churn.

In the current era of artificial intelligence, foundation models will play a crucial role in shaping the future of enterprises and human-computer communications. These models have the ability to understand and generate human-like language and images, and can be fine-tuned to perform a wide range of tasks with higher accuracy and efficiency.

Infographic showing major AI trends from 2000-2022, including Neural Networks, Deep Neural Networks and Foundation Models.

What’s happened in the past year?

While this Generative AI era has been in the making for a few years, its popularity exploded during the summer of 2022 with the release of DALL-E 2 and Midjourney and was further catalyzed by the release of live internet plug-ins, which set the foundation for a new platform of applications with enhanced human-computer interactions.

These elements have kept Generative AI at the forefront of the media and industry conversation through the (semi)-recent release of ChatGPT. The fascinating trait for this era is that technical expertise is no longer a requirement for leveraging AI/ML. Ease of use (no coding knowledge required), facilitated distribution methods (meme and photo sharing via social media), and high-quality results have contributed to the mainstream exposure (and partial adoption) of this generative technology.

For consumers, products like ChatGPT and Lensa AI have taken Instagram and Twitter feeds by storm, further increasing exposure to large language model (LLM) performance. For developers, OpenAI’s GPT APIs and Stable Diffusion’s open-source project are offered as easy-to-use developer tools that unlock AI capabilities for engineers without AI expertise. With the release of GPT-4, the improvements over the last iteration are outstanding and the subsequent decrease in price for older models continues to fuel the adoption of these types of products.

The growing accuracy and accessibility of the technology allows creators and entrepreneurs the opportunity to run leaner teams and optimize capital as more business needs become programmable.

The Next Era: AI in SaaS

Compared to previous major tech platform shifts (On-prem to Cloud,Web-based to Mobile native apps), the shift to Generative AI has unique qualities that make it just as exciting — yet potentially less disruptive to incumbents and less beneficial to prospective startups. Unlike with the prior shifts, incumbents do not need to re-architect their entire products to adopt this new platform shift. It also takes relatively little effort to implement Generative AI features into existing products and architectures, allowing incumbents to quickly build and launch features based on this technology. In addition, this shift favors companies with bigger, proprietary data sets which can give an edge to more established companies.

Current industry leaders with existing distribution networks can leverage established channels, customer base, and market reach to accelerate the adoption and distribution of products and features. The speed with which legacy companies have been able to embrace this shift is remarkable, as evidenced by recent announcements.

Incumbents

Source: Information and associated dates are from each respective company's websites or press releases as of the date listed above.

Later stage startups have also been quick to launch Generative AI-based features and tools to support them. Notion recently announced the private alpha for its copy editing and generation feature. The company can leverage its existing network of users (and core product integration) as it takes on the dozens of startups building companies around this use case.

Later Stage SaaS

Source: Information and associated dates are from each company's respective website or press release as of the date listed above.

Opportunities for AI SaaS Startups

Although incumbents might have an initial edge with this current platform shift, we believe it may also present an innovator’s dilemma for some. Legacy players may be forced into strategic changes that could jeopardize their core business in the short term.

Even Google's primary business, search, is now in jeopardy as they balance the need to maintain dominance with the potential revenue disruption posed by Chatbot-based search engines like Perplexity AI.

Most recently, ChatGPT’s integration with applications also introduced a potential risk for app-store operators, given its ability to execute commands without the need of traditional mobile apps. Concerns around reputational risk, cannibalization of existing products, and potential offensive/inaccurate responses create challenges for incumbents as they navigate the changing landscape of Gen AI technologies.

Still, we believe there are big opportunities on the application layer. Products that can successfully integrate into enterprise workflows or leverage high-value datasets while solving a real unmet need have the potential to succeed.

Here are the six biggest opportunities we see on the near horizon:

1. Addressing Real Customer Needs: Companies should have a clear and compelling use case for their AI-powered technology that addresses a real-world problem or need. The technology shouldn’t be in itself the solution, it should enable a workflow or feature set that wasn’t possible with previous technologies. The availability of Generative AI can sufficiently answer the “Why now?”, but it cannot be the answer to “Why this?” Companies will need a sustained advantage beyond just using AI.

2. Leveraging Strong Datasets / Management: As model performance begins to converge (and particularly as RLHF task-specific feedback data seems increasingly important for useful applications of LLMs), companies can seek to differentiate themselves through the development of specialized models that address specific business needs and provide higher-quality outcomes. Companies’ acquisition and data management should have robust governance practices in place to ensure that the data being used to train and evaluate their models is high quality, accurate, and unbiased. Users of the technology should also be able to explain how they are addressing isssues such as data privacy and security.

3. Employing Thoughtful Product Design: Elevated UX design is expected for SaaS companies today. This is no exception for GenAI, but there needs to be thoughtful consideration for how the GenAI features get surfaced, validated and improved (if user training is enabled). Currently, we are seeing GenAI features bolted onto the UI as a floating text box / chat interface. This will need to evolve for this functionality to be core to the user experience.

4. Accepting a New Definition of Quality: Historically, quality has been associated with uptime/availability/bugs, where the presence of an application error or issue is clear to the end user. With GenAI functionality, the presence of errors is hard to decipher. Hallucinations are more common and yet difficult to detect. The output from this functionality can seem correct and yet an erroneous output can have significant consequences for the user or company.

Having a well thought out quality/validation component to your application is a key consideration as we evaluate companies building with this feature set at its core.

5. Experimenting with Compelling Business Models: While we expect the cost of this technology to go down over time, it’s still not negligible. Companies should continue to optimize their operating structure and find an optimal balance between cost, performance, accuracy, and ease of use to build a sustainable business.

Companies will also have to clearly articulate what their sustaining advantage will be over time as competition catches on. Integrations are key to stickiness, not only because they facilitate workflows, but because they also give companies access to valuable data to improve their GenAI’s utility.

Lastly, we look for companies where existing incumbents cannot just replicate the functionality. We believe some of the most compelling opportunities we are seeing are products/functions that did not previously make sense to integrate, but now do.

6. Hiring a Phenomenal Team: As with all early startup opportunities, the team is a key determinant in the success of the product and company. With the latest AI developments, it is no longer required to be an expert to build SaaS products rooted in AI. Still, teams need to understand the technological landscape to stay ahead in terms of product development and they also need a strong understanding of the customer and user experience.

-----

The current wave of AI progress and increased understanding of foundation models presents an opportunity for new companies to emerge and make an impact in the industry. However, this newfound accessibility in building AI-powered products will create an abundance of noise in the market.

Incrementalism won’t prevail.

We believe the SaaS companies that prove to be foundational in this era of AI will need to go beyond having the latest technology and great user experience—they will need to demonstrate novel approaches that displace previous applications and create entirely new AI in SaaS categories and markets.

If you're building an AI SaaS application or want to learn more, reach us at: SaaS-Team@nea.com

----------

Snapshot of NEA's AI SaaS Portfolio

About NEA

About the authors

Vanessa Larco

Vanessa joined NEA as a Partner in 2016 and focuses on enterprise and consumer investing. She has led investments in Assembled, Kindred, Rewind AI, Cleo, Evernow, Rocket.Chat, and Mejuri, among others. She is also a board observer at Forethought, SafeBase, Orby AI, Granica, Modyfi, and HEAVY.AI. She was a board observer at Robinhood until its IPO in 2021. Prior to Venture, she led product teams at Box, Twilio, Disney, and Xbox.
Vanessa joined NEA as a Partner in 2016 and focuses on enterprise and consumer investing. She has led investments in Assembled, Kindred, Rewind AI, Cleo, Evernow, Rocket.Chat, and Mejuri, among others. She is also a board observer at Forethought, SafeBase, Orby AI, Granica, Modyfi, and HEAVY.AI. She was a board observer at Robinhood until its IPO in 2021. Prior to Venture, she led product teams at Box, Twilio, Disney, and Xbox.

Juan Mogollon

Juan joined NEA in 2022 as an investor on the Technology team, where he focuses on early-stage enterprise investments. Prior to NEA, Juan was an investment banker at Evercore, in New York, where he worked on M&A transactions and capital raises for consumer and enterprise technology clients globally. Previously, he spent time at Bank of America and Marbella Interests. Juan graduated from the University of Texas at Austin with a bachelor’s of business administration in finance. He is proudly Colombian.
Juan joined NEA in 2022 as an investor on the Technology team, where he focuses on early-stage enterprise investments. Prior to NEA, Juan was an investment banker at Evercore, in New York, where he worked on M&A transactions and capital raises for consumer and enterprise technology clients globally. Previously, he spent time at Bank of America and Marbella Interests. Juan graduated from the University of Texas at Austin with a bachelor’s of business administration in finance. He is proudly Colombian.