Airtable CEO Howie Liu calls his recent AI tool Cobuilder the corporate’s biggest release in five years.
Leigh Nile
Airtable’s recent product “Cobuilder” puts AI-generated apps within the hands of each worker – and the $11.7 billion startup is now considering an IPO again after painful layoffs.
For greater than a decade, Airtable CEO Howie Liu has been on a mission to simplify software development. His tools turn data tables and document collections into apps large and small to save lots of employees time. But while Airtable’s low-code tools have proven popular—500,000 firms have built 50 million such apps so far—they’ve mostly required hours of time and technical expertise.
Now Liu believes Airtable has found the answer to its biggest bottleneck: generative AI. And with a brand new tool launched on Wednesday, Airtable Cobuilder, his startup is using a mix of leading AI models and its own recent tools to make constructing an app as easy as typing out an easy text prompt.
“What would have taken many hours can be done in literally 30 seconds,” said Liu Retrieved 2018-08-18. “We see this as a big step that will enable the development of many more applications.”
Cobuilder is not Airtable’s first foray into AI—the corporate began integrating AI tools into its app constructing process last yr and added more features in March—but it surely’s the primary to depend on AI to do all of the heavy lifting, not only speed up a few of it. Airtable generates the app by analyzing the user’s role, company, and query and making a series of API calls to a large-scale language model—currently primarily OpenAI’s GPT-4 family of models—with its own improved, more specific prompts.
For example, a visible effects manager at Netflix could tell Cobuilder their job title, and Cobuilder could suggest an app that might keep the planning and all relevant files for an upcoming release in a single place, pulling information from calendars, Dropbox, and Google Sheets. Users also can ask for more specific tools: an e-commerce seller could request an app that might track all deliveries of a selected product, taking shipping costs and duration into consideration. The result: something that usually effectively looks like a dashboard or interactive spreadsheet, with the relevant variables (orders, due dates, customs wait times) displayed in a simple interface that might be used to tweak or toggle on and off.
Eventually, the corporate will offer businesses the power to feed their very own data into the AI-generating process, but for now, users must accomplish that manually after Airtable returns its app attempt. (Non-customers also can try launching apps on the corporate’s website.) Airtable’s software doesn’t train these generated apps itself — it uses metadata, comparable to whether the client asked it to try again — and it doesn’t share customer data with OpenAI, either, Liu said.
While the output is not way more effort for a user, the means of returning a working app is more complex behind the scenes than, say, asking an issue to OpenAI’s ChatGPT. “We have to do a lot of work behind the scenes because the models aren’t all-powerful,” Liu said. And unlike with fact-based queries, Cobuilder’s answers aren’t as easily right or incorrect. “There might be 50 different viable ways to implement and structure an app, and some work better than others.” Airtable’s AI tools absorb user feedback—not the generated outputs themselves—and proper themselves over time to raised match an app to a question type.
Testers have built 1000’s of apps with Cobuilder in recent weeks, but the corporate declined to call specific users. It has previously named Amazon Web Services as an early adopter of Airtable AI and claims to be working with a significant player in each the legal and media industries. Aetna, Nike, and Walmart are other notable Airtable customers. For regulated industries like banking and healthcare, Airtable can also be developing tools for managers like chief information officers to simply label what data and workflows needs to be available for worker app use, and to “bless sources of truth” for, say, any app that desires to make use of an organization’s product roadmap or upcoming marketing materials.
Airtable’s foray into AI marks a brand new phase for an organization that has gone from hot startup darling to cautionary tale lately. Airtable was founded in 2012 by Liu and two others and launched the next yr. Like popular design software startup Figma, Airtable took years to launch a product but succeeded because people bought the product on their very own—a beacon of, in startup parlance, “product-led growth.”
By 2018, enterprise capital firms comparable to CRV, Benchmark, Thrive Capital and Coatue had backed Airtable; because the then 30-year-old head of a $1.1 billion unicorn, Liu became the topic of glossy profiles. By December 2021, investors valued Airtable at $11.7 billion, giving it certainly one of the best prices within the technology industry. Just a few months later, the corporate landed at number 6 on the Forbes Cloud 100 list in 2022.
This is what an AI-generated app looks like in Airtable.
Air table
But when free-flowing capital from a enterprise capital market fueled partly by low rates of interest suddenly dried up this yr, Liu was caught flat-footed. Employees paid the worth for bloated hiring and unrealistic growth visions when Airtable laid off 20% of its workforce in December, then one other 27% in September 2023 — nearly 500 employees in total. “It’s a bad feeling,” Liu said. said back then. “I made the decisions that brought us here.”
A yr later, Liu acknowledged that Airtable had “lost its bearings” on account of its own momentum. “The exuberance obscured the true nature and quality of virtually all companies that were in hypergrowth and hyperfunding mode,” Liu reflected. “Nobody had to put their cards on the table and show how good their company really was.”
During that point, Liu had dinner with Coatue investor David Schneider, the previous chief revenue officer and president of ServiceNow who was raised within the classical school of enterprise sales. Schneider asked Liu to call his top 25 customers and the way they used Airtable; Liu couldn’t. The message was clear: Get on the market. “He urged me to meet with them, and I did,” Liu said on a recent trip to New York to just do that. “Here, different business models require different CEO styles.”
These customer meetings gave Liu and his team cause for hope, at the same time as morale suffered with the departure of colleagues and several other executives. “Howie was like the architect of a house who never saw the finished house,” Schneider said. “So he went and looked at the beautiful houses that were being built based on his blueprints.”
Airtable wasn’t near running out of cash – it still has about $1 billion within the bank, including the complete $735 million Series F it raised in 2021 and the majority of its $270 million Series E from a couple of months ago, which it raised to present it “financial freedom,” in line with Liu said on the time – however the tightening stabilized the corporate, Liu said, with positive money flows. After that, the corporate could invest again in its product – particularly Airtable AI. “It’s not just about forging your way and optimizing,” said Vince Hankes, a partner at Thrive Capital who has also worked closely with Liu. “At the end of the day, we’re in the technology business, a growth business, and you have to find a way to get back to it.”
For Liu, Airtable’s opportunity to expand its market through AI – each when it comes to how it could actually be used and who uses it – is the reply to that query. “We’re going to be a winner in this new world order. If I didn’t believe that, I probably would have explored options to sell,” he said.
Instead, Liu made acquisitions of his own that proved crucial to Cobuilder’s development: Airplane, which offered custom workflows for developers, and Balsa, which developed a tracking tool for software projects. Airtable bought each primarily for his or her employees, who became core members of the Cobuilder offensive. “We have a very clear idea of what layer of the AI stack we want to play in,” Liu said.
That means Airtable won’t construct its own models or provide resources to customize existing ones. Cobuilder uses OpenAI today, Liu said, but the corporate is “agnostic” to any single model provider going forward. And while Airtable is eyeing open-source models, comparable to Meta’s recently launched Llama 3.1 model, Liu said he stays skeptical concerning the must move away from OpenAI within the near future, given the pressure the corporate faces to supply more models at competitive prices: “These things are getting cheaper, better and faster.”
Investors, meanwhile, are glad he hasn’t called private equity firms yet. More than half of the highest 500 U.S. firms by revenue are paying customers, Airtable said; greater than 25 customers spend at the least $1 million annually, and several other spend at the least $5 million. The company has revenue within the “hundreds of millions” and meets the “rule of 40,” a commonly used benchmark for software firms that expects revenue growth and profit margins of at the least 40 percent, said Liu, who added that Airtable’s numbers would place the corporate in the highest decile of public cloud competitors within the BVP Nasdaq Emerging Cloud Index.
An IPO is back on the table, although Airtable’s CEO said that like lots of his peers, he’s “waiting to see how the markets play out.” Airtable already reports its quarterly results internally, he added, so the corporate is “ready for an IPO at any time.”
