- With its coding capabilities, generative AI is making it easier to develop software.
- This could disrupt the way software is created, distributed, and used, VCs and startup founders say.
- However, the death of the traditional SaaS company still seems a long way off.
While ChatGPT has been wowing the public, behind the scenes investors and technologists are beginning to talk about a deeper disruption to the inner workings of the established software industry.
A new potential framework for software, whose earlier iteration was coined "malleable software" by researcher Philip Tchernavskij, describes a future where generative AI and humans work together to customize tooling and even create entire applications.
This outcome would flip the traditional software industry on its head, calling into question the value of SaaS companies in a world where everyday people can build software themselves.
"No-code was the first step," said Matt Turck, a partner at venture capital firm FirstMark. "This is the final chapter of software eating the world, where a bunch of people can create enterprise software within the enterprise."
This would represent quite a reversal for the industry. Software-as-a-Service companies have been the disruptors for a decade, not the disruptees. They have sky-high valuations because investors are betting their subscription revenue will continue steadily rising for many years to come. If generative AI really catches on, though, that future may look very different.
Democratizing tech creation
Venture capitalists and startup founders have been obsessed with the idea of democratizing tech creation for years, as seen by the rise of low-code and no-code startups like Airtable, last valued at $11 billion, and Webflow, which landed a $4 billion price tag last year.
Some technical knowledge was still required to build most software. Now, though, the emergence of generative AI tools like GitHub Copilot has opened up the ability to generate code using just natural language, Ethan Kurzweil, a partner at Bessemer Venture Partners, told Insider.
For Jake Saper, a general partner at Emergence Capital, the use cases that stand to be disrupted first are simple, low-risk tasks and applications in small and midsize businesses. These instances offer the lowest chance of business disruption and require the least cross-company coordination, he said.
Vertical software companies taking existing technologies and making them easier to use in antiquated industries could also be under threat of replacement if their value-add is more around convenience versus actual product differentiation, Fika Ventures senior associate James Shecter said.
Already, technologists have begun to use generative AI tools like Copilot to build simple apps, including a trivia game and a site for discounted Amazon items.
Some later-stage tech startups are trying to get ahead of the curve by sharing the power of creation with their customers. One example can be found in knowledge base startup Guru's AI writing assistant, which lets customers create their own custom tones of voice using generative AI. This challenges the traditional idea of software as a rigid tool with a fixed set of available actions for users, Guru cofounder and CEO Rick Nucci told Insider.
"We've talked about 'platforms' in the SaaS world for a long time, the idea that someone can create a set of foundational building blocks that customers can configure and shape to be what they want," he said. "This is a step change that's actually happening."
A new era for software
Some VCs and founders believe that generative AI could not only transform the way we create technology but also the way we interact with it through ultra-personalization.
For instance, new generative AI technology could help startups create user interfaces customized to each person's exact preferences, Bessemer partner Talia Goldberg said. Already, ChatGPT is showing early signs of this by choosing to provide certain responses in data table format, even when users don't specifically ask for that, she explained.
In more extreme cases, entire tools could be generated by AI on the fly to replace common actions a user takes, CRV principal Brittany Walker said.
In the long term, VCs like NEA partner Vanessa Larco and investor Elad Gil believe that autonomous AI agents, rather than humans, will be the main entities interacting with software. One potential scenario could be a world where individuals have a primary AI agent that coordinates and manages a number of "micro-agents" capable of doing everything from text messaging to scheduling dinner reservations, Larco told Insider.
These types of connections and interactions — the technical plumbing that currently makes different software programs work together — is the bread and butter business of many SaaS companies. If generative AI models can do this work automatically, what will happen to these SaaS businesses?
A 'healthy pressure' for traditional SaaS providers
To be sure, the death of the traditional software company still seems a long way off.
First, the choice between building software yourself or buying from a third party brings with it a substantial opportunity cost.
"I don't necessarily want to sit on my computer for 10, 12, 15 hours developing this when I can go and find something that's ready out-of-the-box," CRV's Walker said. "The barrier would need to drop very low for a critical mass of people to start creating their own bespoke software."
Additionally, paying an outside software vendor helps people put the burden of safety, maintenance, and accountability onto a third party, Emergence Capital's Saper said.
However, even skeptics admit that the threat of generative AI to traditional SaaS will push established software companies to prove their worth.
"It'll probably be healthy pressure because the 'build' decision may be more tempting because it'll be theoretically easier to do," Saper said. "It's going to put pressure on software vendors to really deliver value."
Got it a tip? Contact Stephanie Palazzolo using a non-work device on encrypted messaging app Signal (+1 979-599-8091), email ([email protected]), or Twitter DM @steph_palazzolo.