David Yue, CEO (left) and Finsam Samson, CTO (right) started Accordance last year to help accountants with its AI software.<\/p>\n
Accordance. <\/p>\n
After months of providing AI consulting services to different businesses, David Yue discovered a “quiet but dire\u201d problem in the field of accounting: a massive shortage of professionals. There are 340,000<\/a> fewer accountants in the U.S. than there were five years ago and 75% of CPAs are retiring in the next decade.<\/p>\n Realizing that artificial intelligence could help fill the gaps, Yue, a Stanford dropout and Y Combinator alum, cofounded Accordance, which has developed AI software to help tax and accounting professionals do their jobs. Tax problems are especially gnarly, he said, requiring deep expertise around an ever-changing web of regulations as well as specific edge cases.<\/p>\n Accordance\u2019s software is trained on public tax data, including up-to-date code and court rulings and also learns from customers\u2019 own financial histories. Tax advisors are widely expensive, with large companies splurging thousands if not millions to hire them. Yue said his company\u2019s tools can bring those costs down\u2013for one of its customers, it helped save $50,000 a month. That has appealed to companies like Figma and Quora, who are among the startup\u2019s customers. Design behemoth Figma used the startup\u2019s software to help with complicated tax issues as it prepared to go public earlier this year. \u201cYou give [the company\u2019s software] to a junior, then they become an expert, you give it to an expert, they become a champion,\u201d Yue told Forbes.<\/p>\n Accordance announced today that it raised $13 million in seed funding from Silicon Valley heavyweights including Khosla Ventures, General Catalyst, and Anthropic, as well as NEA and Sequoia.<\/p>\n With the field quickly evolving, Yue is also partnering with universities like University of San Francisco to help the next generation of accountants adopt AI in their practices. Starting this fall, he claims tens of thousands of students will use the AI tool in their tax research and tax analysis classes. Khosla Ventures Partner Kanu Gulati said it\u2019s a convenient inroad into the startup\u2019s future customer base and students would \u201ctake the tool with them when they graduate.\u201d<\/p>\n Yue doesn\u2019t believe that artificial intelligence will ever fully replace tax professionals, but he said it could help them to do more with less. For the young founder, that\u2019s a satisfying way to make a lasting impact. \u201cEmpires rise and fall on the back of taxes,\u201d he said.<\/p>\n Let\u2019s get into the headlines.<\/p>\n BIG PLAYS<\/p>\n OpenAI is facing both major roadblocks and public pushback in its plans to restructure from a nonprofit-controlled entity to a for-profit corporation\u2014 and has even discussed moving out of California as a last resort, the Wall Street Journal<\/a> reported. A group of opposing nonprofits have banded together and asked the state attorney general to investigate if the restructure violates California\u2019s nonprofit law. A lot is at stake for OpenAI, which has $19 billion in future financing on the line from Softbank that\u2019s contingent upon it becoming a for-profit entity.<\/p>\n ETHICS+ LAW <\/p>\n Anthropic<\/a> has agreed to pay $1.5 billion to a group of 500,000 authors and publishers to settle a massive copyright infringement lawsuit after a judge ruled it had used pirate websites to steal and store millions of copyrighted works, though the AI giant said it had not used those books in its model training. The landmark settlement is the largest in U.S. copyright history. However, there is one win for the AI company here: the judge ruled that Anthropic\u2019s use of legally obtained texts for model training fell under the \u201cfair use\u201d clause of the law.<\/p>\n Elsewhere, data labelling company Scale AI, which provides training data to top AI companies, is suing a former employee Eugene Ling for alleged corporate espionage. The lawsuit accused Ling, who now works at data labelling rival Mercor, of stealing 100 confidential documents containing trade secrets. It also alleges he tried to convince a customer to hire Mercor and fellow employees to jump for the rival company while he was still at Scale.<\/p>\n AI DEALS OF THE WEEK <\/p>\n Sierra, a startup that sells AI customer service software, raised $350 million at a $10 billion valuation. Cofounded by OpenAI chairman and former co-CEO of Salesforce, Bret Taylor, Sierra develops AI agents \u2014systems that can autonomously carry out entire tasks\u2014 used by hundreds of millions of people for tasks like roadside assistance, delivering furniture and disputing charges, the company claimed in a recent blog<\/a> post.<\/p>\n Basten, which provides inference services that let AI systems generate outputs based on previously learned data, has raised $150 million at a $2.15 billion valuation led by BOND. The company counts startups like OpenEvidence, Writer and Abridge as customers.<\/p>\n Databricks, which makes data-focused software for AI training, has raised $1 billion, nabbing a valuation of more than $100 billion. The startup, cofounded by seven University of California, Berkeley academics, claims its annualized revenue in Q2 was $4 billion.<\/p>\n French AI startup Mistral<\/a> has raised a 1.7B\u20ac ($1.9 billion) Series C funding round at a 11.7B\u20ac ($13.6 billion) valuation.<\/p>\n DEEP DIVE<\/p>\n Sitting in a conference room at Mercor\u2019s office in San Francisco\u2019s South of Market district, CEO Brendan Foody recalls the day that changed everything. Meta had announced it was buying almost half of data labeling giant Scale AI for $14 billion and poaching its star CEO Alexandr Wang. Mercor, a smaller rival that recruits PhDs and other experts to train models for AI labs, saw an immediate opening.<\/p>\n \u201cI was initially surprised,\u201d Foody tells Forbes. \u201cThen it slowly transitioned from surprise to excitement around the future.\u201d His cofounder and CTO Adarsh Hiremath chimes in: \u201cIt just doesn’t happen too often in startups where your biggest competitor gets torpedoed overnight.\u201d That\u2019s because Scale\u2019s tie-up with Meta means many of the big AI labs, worried about a loss of neutrality, no longer want to work with them, the founders argue.<\/p>\n In a brief period of time Mercor has become one of the AI era\u2019s poster children for Silicon Valley\u2019s twenty-something founders<\/a>. At age 22, Mercor\u2019s leaders Foody, Hiremath and Surya Midha are all Thiel Fellows, members of conservative billionaire investor Peter Thiel\u2019s program to dole out $100,000 grants every year to young people in exchange for foregoing college. The three longtime friends, who met on their Bay Area suburb\u2019s high school debate team, started Mercor in 2023 before their fellowships.<\/p>\n The buzzy startup, backed by heavy hitters including blueblood venture firm Benchmark, Twitter cofounder Jack Dorsey and former U.S. Treasury Secretary Larry Summers, is a company in transition. Last month, it hired former Uber product chief Sundeep Jain as the startup\u2019s first president, the company told Forbes, injecting tech-giant experience into the neophyte operation. Mercor will soon move to Instagram\u2019s old offices in downtown San Francisco, a bigger space for its fast growing team. And this year the company made its debut at No. 89 on Forbes\u2019 Cloud 100 list, our ranking of the world\u2019s top private cloud computing companies.<\/p>\n Read the full story on Forbes<\/a>.<\/p>\n WEEKLY DEMO<\/p>\n Hundreds of user conversations with Anthropic\u2019s Claude chatbot inadvertently showed up on Google search results, Forbes<\/a> reported. The chat transcripts, which identified the names and personal details of some users, were visible due to a bug that published conversations on the web when a user clicked the \u201cshare\u201d button. Anthropic said the chats, which are now removed, were visible on Google and Bing because users had posted links to those conversations on social media, but Forbes spoke to one such user who said they hadn\u2019t shared their chatbot conversations on social media. It isn\u2019t the first time this has happened: in recent months, conversations with xAI\u2019s Grok and OpenAI\u2019s ChatGPT have also previously been searchable online, posing major privacy concerns.<\/p>\n MODEL BEHAVIOR<\/p>\n