Sahil Aggarwal's timing appeared to be perfect for launching his software company in 2021. He raised nearly $30 million from investors, and by the following year, his startup, Rattle, was valued at more than $100 million.
Then along came generative artificial intelligence that could easily automate the kind of administrative work Rattle specialized in. Aggarwal tried to integrate AI, but said it felt like attaching an engine onto a horse-drawn carriage. The company teetered on the brink.
"If two engineers can reproduce your entire product in a matter of a few weeks, then you deserve to be killed," he said.
Generative AI is steamrollering the once booming industry known as software-as-a-service, or SaaS, where customers access software over the internet, typically on a subscription basis. SaaS exploded in the 2010s, as companies such as Slack and Zoom raised hundreds of millions of dollars from venture-capital firms and others went public in blockbuster IPOs.
Now people in Silicon Valley are talking gloomily about the "SaaSpocalypse." AI threatens to make some software tools obsolete, particularly those built around narrow tasks such as legal drafting, research and other repetitive work.
Shares of public SaaS companies such as Workday, Salesforce and Adobe have fallen more than 30% from their peak over the past year. IBM lost $69 billion in value in a single day in July after it issued a profit warning as customer spending shifted from software to AI hardware. Earlier this week, Italian tech conglomerate Bending Spoons agreed to acquire workflow software company Airtable for $1.3 billion, well below the company's last private valuation of $11 billion.
SaaS companies are under immense pressure to reinvent themselves. Inside boardrooms, leaders have discussed the risk that entire businesses will amount to little more than features inside tools released by leading AI labs Anthropic and OpenAI. Investors, worried that some software bets will go to zero, are pushing startups to adapt.
"Waiting and seeing is no longer a strategy," said Byron Lichtenstein, chief business officer of Insight Partners, which invested in many SaaS companies and is now trying to save some of them.
The startups are faltering because demand is shifting from tools that help people get work done to AI "agents" that do the work for them. That is forcing leaders of the last chapter of venture-backed startups to lay off employees and overhaul their companies. Many are changing business models, relaunching products or, in some cases, shutting down and starting over.
Venture-capital investors are grappling with the same uncertainty. Companies that once seemed destined for lucrative IPOs now face an uncertain future. While pouring billions into AI startups they hope are outside the blast radius of major AI labs, they also are struggling to preserve the value of software companies built before the AI boom, triaging portfolio companies and preparing some businesses to be sold for parts.
Insight is one of the investors most synonymous with the SaaS era. It built a reputation over decades backing software winners such as business software company Monday.com and cybersecurity startup Wiz, which was acquired by Alphabet.
But the development of AI moved faster than Insight expected. The advice and tactics that helped build software companies for years suddenly needed to be reworked.
Last December, Insight co-founder Jeff Horing told staffers at its Manhattan headquarters that if they weren't in the boardroom telling companies they should be doubling their potential revenue opportunities with AI, they weren't thinking big enough.
"I call it scare and inspire," said Lichtenstein. "Our job is to show the power of what this could be."
Dori Yona, a former member of the Israel Defense Forces, is the founder of SimpleClosure, which helps startups wind down. Fifty-one percent of Yona's venture-backed customers in the first half of the year were software or IT-services companies, up from 44% a year earlier. As more companies fold, Yona hopes to persuade founders to make the decision while they still have cash in the bank.
"If they are not AI-first, they are having a really hard time fundraising right now," Yona said. "It's sometimes easier for a SaaS company to just shut down and just restart."
In June, his company invited founders, investors and lawyers to gather at the Pearl Box, a Manhattan cocktail bar, to swap stories on startup failures and the struggle to compete with AI giants. "Very rarely are you a first-time founder and you hit it out of the park," Yona recalls telling the crowd.
Sipping a drink at the gathering, one venture capitalist offered a blunt assessment of a company in his firm's portfolio: "Anthropic is going to eat their lunch."
Jason Lemkin, founder of a SaaS organization called SaaStr, said the industry has bifurcated into companies that have embraced AI and are growing and others who haven't and are struggling. "All the pre-AI guys are gone," he said. "They're just not growing."
That point was driven home at the group's annual conference in San Mateo, Calif., in May, when a plane appeared overhead towing a banner that read "SaaS is dead."
That stunt was the work of Manny Medina, who stepped down as chief executive of an early SaaS leader, Outreach, in 2024 and now has a startup called Paid. Outreach was valued at $4.4 billion in 2021, but has since reduced its workforce by about 30%.
Many of the software startups that launched in the early 2010s, Medina said, aren't growing fast enough to elicit any real interest from public-market investors, putting them in a holding pattern.
"Every CEO that I talk to is considering quitting and starting again," he said. His new company helps companies that make AI agents determine how much to charge customers.
Abhijit Mitra, the current CEO of Outreach, said the company has restructured. Its new business, which uses AI agents to build revenue operations systems that can perform tasks autonomously, he said, is growing quickly.
Some software-services companies that have shifted gears are reporting growth.
Lantern, a digital-health company founded 15 years ago, also pushed into AI, using it to reinvent its claims-processing operations by turning a 16-step pricing process that often took more than two weeks into one that can now be completed in about a minute. That boosted profits, said Lantern CEO John Zutter.
Gong, a decade-old sales software company, retrained its workforce and recently surpassed $500 million in annual recurring revenue -- a metric used by software companies that annualizes subscription revenue -- growing over 55% year-over-year.
Intercom, a 15-year-old customer-service software business, built an AI customer-service agent called Fin. It reached $400 million in annual recurring revenue earlier this year, and Salesforce agreed in June to acquire Fin for $3.6 billion.
Boom and bust
Aggarwal, the 34-year-old co-founder of Rattle, grew up in an industrial city in northern India and started his career in management consulting in New Delhi. He moved to San Francisco in 2019, joining marketing-software startup Mutiny. At the height of the pandemic, he decided to start his own business.
To save money, he moved back to India, going for months without seeing his co-founders, Apoorva Verma and Milan Singh. Over six weeks, the trio built a software product for salespeople.
Funding flowed easily, Aggarwal said, including from Indian subsidiaries of Sequoia Capital and Lightspeed Venture Partners. When pandemic travel restrictions eased in late 2021, the three founders gathered for planning sessions and some hiking in Kasauli, a small town in India's Himalayan foothills. On Christmas Day, they received a term sheet for a Series A funding round led by Insight.
Aggarwal moved back to San Francisco, and Rattle expanded, signing on clients such as human-resources startup Rippling, cloud-design company Figma and productivity-management startup Notion. It charged users a minimum of $20 a month, attracting dozens of customers and raised nearly $30 million from investors, he said.
By 2023, though, advancements in generative AI made it clear that Rattle needed to incorporate the new technology. Aggarwal wasn't yet convinced he had to bet the business on it, figuring Rattle would endure with just a dollop of AI.
That changed the next year, when OpenAI released its first reasoning model, o1, which showed that AI could become the core of a product, not just a feature. Rattle needed to rebuild much of its underlying software so that AI agents were at the center of the architecture, not tacked on the outside.
"You have to burn the ships and start from the ground up," Aggarwal said.
By the end of 2024, Aggarwal realized the efforts to integrate AI weren't working. The employees had spent months trying to support Rattle's original product while building a new AI "superagent" that could automatically update Salesforce records, among other tasks. Both efforts were stalling.
He told employees he was laying off the sales, marketing and customer-success teams. The startup shrank from 70 employees to 15, retaining only a small group to rebuild around an AI-focused product.
"The first thing that was very hard is letting go of the people that you sold a dream to," Aggarwal said. "How do you tell the team that, 'Hey, what we did before is not relevant anymore. We have to start from scratch.'"
Investor advice
In May, Insight held a conference-turned-survival workshop in Hollywood for about 100 executives from its portfolio of startups. Attendees crisscrossed Los Angeles, touring Universal Studios and hearing from executives navigating the AI boom. On the set of the Reba McEntire sitcom "Happy's Place," they peppered the producer with questions about how Hollywood uses AI.
Marius Buleandra, a member of Anthropic's applied AI team, gave a presentation about how the company integrated AI into its everyday processes and how other companies can do the same.
The message from Insight Partners: Adapt quickly or be left behind.
By the time Aggarwal arrived for the three-day affair, he had already laid off staff, stopped selling the company's original product and was rebuilding.
Rattle developed an operating system for sales and revenue teams that sits on top of companies' other software tools such as Salesforce. It functions like an AI teammate, independently completing tasks such as tracking customer information, identifying promising sales opportunities, spotting deals that may be slipping and forecasting future revenue.
Aggarwal began testing the new product, called Von, in February. In the spring, he announced the new direction for the company and rebranded it as Von.
The company recently reached $1 million in annualized recurring revenue. To celebrate the milestone, Aggarwal pulled three bottles of Champagne from the office fridge and passed them around to the 11 employees based in San Francisco.
"I was a horseless carriage until about 18 months ago, and I definitely saw the writing on the wall," he said. "I have made that transition...I am on the other side."
Comments