Ann: Customer update and FY23 preliminary results, page-387

This Company Helped Make AI a Reality. Now the Machines Are Crushing It. -- WSJ22/03/2024 1:00AMBy Stuart CondieSYDNEY -- Australia's Appen helped to train the AI systems of many of the world's biggest tech companies, including Alphabet's Google. Now it is at risk of being overtaken by those machines.The company, once worth $3.3 billion, is closing offices in North America and has changed its chief executive in an effort to stem bleeding that worsened recently when Google ended a relationship that had contributed 30% of its revenue. A week ago, Appen said a potential all-stock takeover by Nasdaq-listed Innodata for $101 million had collapsed.Appen's struggles illustrate how quickly AI is transforming global business -- including the tech companies that fed its development. Appen has long relied on its so-called global crowd: More than a million gig workers who manually analyze data, usually at home and often to supplement their income from elsewhere.AI-powered machines are now doing much of that work. Appen has tried to diversify in recent years and has gained some traction with Chinese automakers and retailers, but that business doesn't fill the gap left by tech clients. Appen's contract with Google finished at the end of last week."If we do see another one leave, I think that could be a pretty fatal blow," Jefferies analyst Wei Sim said. Appen's shares have fallen about 98% since August 2020.At its peak, Appen's stock valuation was underpinned by deals with large U.S. tech customers that generated about 80% of its revenue. Appen has said its cohort of humans can help to train machines to recognize any language, even those not widely spoken.The challenge for Appen is that customers such as Google are plowing cash into heavily automated generative AI, which uses neural networks to identify patterns within data and generate new content. This process can be less reliant on the human input on which Appen is built, and competition is fierce for what training is required.Google said its decision to end the contract with Appen was part of a broader review of deals with suppliers."When they just decided to spend less, our revenue turned in the other direction," said Ryan Kolln, a longtime employee who became chief executive officer of Appen in February when predecessor Armughan Ahmad stepped down.Two weeks before he was promoted to CEO, Kolln had joined a weekend call with Google when it surprised Appen by announcing it wouldn't provide any more project work. Kolln said Google didn't fully explain why it was walking away.The first sign of cracks in Appen's business came in August 2020 when it missed expectations for its half-year earnings, prompting its stock to fall 11%.More downgrades, the scrapping of its dividend and multiple equity raises followed. The stock lost its spot on Australia's benchmark S&P/ASX 200 index, putting it out of view for many investors and funds. Appen kept trying to focus on top-line growth, but customer orders often failed to meet expectations, partly due to the way its contracts were written, Sim said."There weren't really any minimum volume requirements put in place or any kind of penalties for not reaching those targets discussed at the start of the year," although Appen did eventually overhaul its contracts, he said.Appen had tried to meet customers' desire for increased automation with the 2019 acquisition of Figure Eight, but Sim said the machine-learning software platform wasn't popular with its intended users.In May 2022, a Canadian rival tried to buy Appen for about $800 million, but the talks collapsed around 24 hours after they were made public. Appen had warned of declines in sales and earnings when it revealed the deal talks with Telus International.The uncertainty facing Appen is such that Kolln doesn't yet know where he will be based as CEO. As chief operating officer, Kolln had been working out of Appen's Toronto office, which is one of two North American sites that the company intends to close as part of its cost cutting.Announcing a larger-than-expected $118.1 million net loss for 2023, Kolln in February said that Appen is working with 22 large language model builders globally and aiming to restore its core business to profitability. With revenue down some 30% on 2022, that forecast looks tough."They do not have enough revenue to cover their costs," Sim said.Write to Stuart Condie at [email protected] (END) Dow Jones NewswiresMarch 21, 2024 10:00 ET (14:00 GMT)Copyright (c) 2024 Dow Jones & Company, Inc.