Airtable CEO Howie Liu talks big AI push and future IPO

Airtable's new product, Cobuilder, puts AI-generated apps in the hands of every employee, and after painful layoffs, the $11.7 billion-valued startup is once again eyeing an IPO.

---

For more than a decade, Airtable CEO Howie Liu has been on a mission to make building software easier. The company's tools turn data tables and piles of documents into apps, big and small, that save people time. But while Airtable's low-code tools are popular and have helped 500,000 organizations create 50 million apps to date, they require a lot of time and technical proficiency to develop.

Now, Liu thinks Airtable has found a solution to its biggest bottleneck: generative AI. And with a new tool released Wednesday, Airtable Cobuilder, his startup is leveraging leading AI models in tandem with its own new tools to make it possible to launch an app by simply entering a simple text prompt.

“What would take hours can literally be done in 30 seconds,” Liu said. Forbes. “We think this is going to be a big enabler for building even more applications.”

Cobuilder isn't Airtable's first foray into AI: the company began integrating AI tools into its app-building process last year and added more features in March. But this is the first time it's let AI do all the heavy lifting rather than just speeding up some parts. Airtable analyzes the user's role, company, and query, then makes a series of API calls to large-scale language models (currently primarily OpenAI's GPT-4 family of models) to generate an app with its own improved and more specific prompts.

For example, if a visual effects manager at Netflix told Cobuilder her job title, it might suggest an app that would pull information from Calendar, Dropbox, and Google Sheets to schedule upcoming releases and compile related files all in one place. Users can also request more specific tools: an e-commerce seller could ask for an app that would help them track all deliveries of a particular product, taking into account shipping costs and times. The result is more of what looks like a dashboard or interactive table, where relevant variables (orders, deadlines, customs wait times) are presented in a simple interface that can be tweaked or toggled on or off.

Eventually, the company plans to offer companies the option to include their own data in the AI-generation process, but for now, users must manually connect after Airtable returns an app attempt. (Non-customers can try launching the app on the company's website.) Airtable's software doesn't train the generated app itself, but instead uses metadata, such as whether the customer asked for a retry. Liu also said that the company won't share any customer data with OpenAI.

While the output doesn't require much effort for users, the process of returning a functioning app is internally more complicated than, say, asking OpenAI's ChatGPT. “The model is not one-size-fits-all, so we have to do a lot of work behind the scenes,” Liu says. And unlike fact-based requests, Cobuilder's responses aren't simply right or wrong: “There are 50 viable ways to implement or structure your app, and some of them will be better than others.” Airtable's AI tools absorb feedback from users, not the output they generate, and self-correct over time to figure out better ways to map apps to types of requests.

Testers have built thousands of apps using Cobuilder over the past few weeks, but the company did not reveal the names of specific users. The company has previously cited Amazon Web Services as an early adopter of Airtable AI and claims to be working with major companies in both the legal and media sectors. Aetna, Nike, and Walmart are also notable Airtable customers. Airtable is also developing tools for regulated industries like banking and healthcare that allow chief information officers and other administrators to easily label the data and workflows that are available to employee apps. This can be a “source of truth” for apps that use the company's product roadmap or upcoming marketing materials, for example.

---

Airtable's AI push represents a new phase for the company, which in recent years has fallen from buzzworthy startup darling to cautionary tale. Founded by Liu and two others in 2012 and launched the following year, Airtable, like the popular design-software startup Figma, took years to ship a product and then found success by people flocking to it organically—a precursor to what startup jargon calls “product-led growth.”

By 2018, venture capital firms including CRV, Benchmark, Thrive Capital and Coatue had backed Airtable, and Liu, then 30 years old, had become the subject of a glamorous profile as the leader of a unicorn company with a market cap of $1.1 billion. By December 2021, investors valued Airtable at $11.7 billion, making it one of the most expensive companies in the tech industry. A few months later, the company was ranked 6th on the World Economic Forum's stock rankings. Forbes The 2022 Cloud 100 list.

But the free-flowing capital of the venture bull market, fueled in part by low interest rates, suddenly stopped flowing that year, catching Mr. Liu off guard. Airtable laid off 20% of its staff in December that year, and another 27% in September 2023, totaling about 500 people, and employees paid the price for overhiring and an unrealistic vision for growth. “It was nauseating,” Mr. Liu said. Said “That's when I made the decision to come here.”

Looking back a year later, Liu acknowledged that Airtable had become “disoriented” under its own momentum. “The bubble obscured the essence and quality of almost every fast-growing, well-funded company,” Liu said. “Nobody had to lay down their cards and show how good their business was.”

Around that time, Liu had dinner with Coatue investor David Schneider, the former chief revenue officer and president of ServiceNow, who grew up in the classic school of enterprise sales. Schneider asked Liu for the names of his top 25 customers and how they were using Airtable, but Liu couldn't answer. The message was clear: stay steadfast. “He encouraged me to meet with them, and I did,” Liu said during a recent trip to New York. “Different business models require different styles of CEOs.”

These client meetings gave Liu and his team hope, even as morale slumped after colleagues and several executives left. “Howie was like an architect who'd never seen a finished house,” Schneider says. “So he went to see a beautiful house that could be built based on his blueprints.”

---

Airtable wasn't about to run out of cash. The company still has about $1 billion in the bank, including the full amount of its $735 million Series F raised in 2021 and most of the $270 million it raised a few months earlier to give the company “financial freedom,” Liu said at the time. But the tightening has put it on positive cash flow and on firmer footing, Liu said. Once that's done, the company can reinvest again in its products, particularly Airtable AI. “It's not just about opening up and optimizing,” said Vince Hankes, a partner at Thrive Capital who has worked closely with Liu. “At the end of the day, we're in a technology, growth business, and we need to find a way back into that.”

For Liu, Airtable's opportunity to expand its market through AI — both in terms of how it's used and who uses it — provided the answer. “We're going to be winners in this new world order. If we didn't believe that, we probably would have explored options for sale,” Liu said.

Instead, Liu made the acquisitions that would have a defining impact on Cobuilder's development: Airplane, which offers custom workflows for developers, and Balsa, which makes tracking tools for software projects. Airtable acquired both companies primarily for their employees, who became core members of Cobuilder's push. “We have a very clear view of where we want to play in the AI ​​stack,” Liu said.

That means Airtable won't be building its own models or dedicating resources to adapting existing models. Liu said that while Cobuilder currently leverages OpenAI, the company will be “agnostic” to any particular model provider going forward. Airtable is keeping an eye on open source models, such as Meta's recently announced Llama 3.1 models, but Liu said he's skeptical about the need to move away from OpenAI anytime soon, given the pressure to ship more models at competitive prices. “These things are going to be cheaper, better, and faster.”

Meanwhile, investors are happy he hasn't called private equity firms yet. According to Airtable, more than half of the top 500 companies in the U.S. by revenue are paying customers, with more than 25 customers spending more than $1 million a year, and several spending more than $5 million. The company boasts “hundreds of millions of dollars” in revenue, and meets the “Rule of 40,” a standard often used by software companies, which aims for combined revenue growth and profit margins of more than 40 percent, Liu said, adding that Airtable's numbers would put it in the top 10 percent of public cloud companies on the BVP Nasdaq Emerging Cloud Index.

With an IPO back on the table, Airtable's CEO said that like many other companies, they're “waiting to see how the market changes.” He added that Airtable already reports quarterly earnings internally and is “open to an IPO at any time.”