As the world’s largest companies swarm around artificial intelligence, corporate credit card and expense management startup Ramp is leveraging the technology to dive deeper into customer savings—joining a small but growing wave of fintech firms hopping on the trend for customer service, automation and more.
Launched on Thursday, Ramp Intelligence features a suite of services using OpenAI’s GPT-4 technology to analyze software prices, scan email receipts, audit expense reports, and respond to customer queries—not unlike the popular ChatGPT service itself.
“For many people going to [software-as-a-service] vendors, it’s very hard these days to know that you’re getting a great price,” Ramp cofounder and CEO Eric Glyman, 33, tells Forbes of the company’s new AI-powered contract analysis tool. “Often your only resort is to go to a pricing page that may not tell you anything, or ask around.”
With GPT-4, Ramp can take vendor contracts that businesses upload and automatically pull out details such as pricing and software seats. The service then compares the pricing against anonymized data from the close to $1 billion some 15,000 companies spend on Ramp each month to show how the quoted price compares to market averages on a graph—and say whether it’s a good deal. If it’s not, Ramp can then go back to vendors and negotiate prices, using their data as leverage. Much of this is because GPT-4 has been trained to understand and interpret large amounts of text, so Ramp Intelligence can use it to scan through a contract’s fine print and understand the context—whereas older AI tech hasn’t been as reliable with messy, unstructured text. Though some companies are already using artificial intelligence (and GPT-4) to help lawyers analyze contracts, Glyman says he believes Ramp is the first company bringing large-scale pricing data to consumers.
Coinciding with the announcement, Ramp is disclosing Microsoft
MSFT
CEO Satya Nadella, who led the storied tech giant through its $10 billion investment in OpenAI this January, has come on board as a Ramp investor and advisor, in addition to Quora CEO Adam D’Angelo, Instacart chief Fidji Simo and Stanford AI Lab professor Chris Ré. Ramp wouldn’t disclose the size of their investments, but the four entrepreneurs join a slate of venture investors including Founders Fund, D1 Capital Partners and Coatue Management.
Other new Ramp AI offerings announced Thursday include a chatbot with cost-cutting advice (using expense data to answer questions such as, ‘How can I cut operating costs?’) and a Gmail and Microsoft Outlook integration that automatically identifies and scans email receipts to process expenses in real time—while flagging unusual transactions (like a late-night weekend Uber
UBER
).
Ramp’s AI launch follows a slew of corporate giants diving into the buzzy technology, many through their own search engine chatbots. However, a small wave of fintech startups have also jumped on the bandwagon. Fintech giant Stripe, one of Ramp’s investors, launched a tool in March using GPT-4 to help developers search company documents and get summarized responses—effectively allowing them to cut down time spent reading. One month earlier, Navan, the travel and expense management startup formerly called TripActions, debuted a customer service chatbot named Ava using OpenAI to field queries from travelers and company finance teams.
Others have been using artificial intelligence since launch, such as Upstart
UPST
to dole out loans, and Sentilink to catch identity fraud. Founded in 2019, Ramp, which was valued at $8.1 billion last March, has also used artificial intelligence since day one, analyzing transaction data to flag wasteful spending on duplicate subscriptions or sudden increases in spending. So far, Ramp claims the technology has helped its average customer shave 3% off their expenses each year, potential savings of some $300 million based on annualized volume of $10 billion.
With the latest batch of tools, that could increase “very meaningfully” as Ramp analyzes spending in the coming months, says Glyman.
Read the full article here
