Software

AI + ML

An established AI player is in nasty trouble – in this market? What? Why?

Australia's Appen boasted of clients including Amazon and Microsoft. Then work dried up, Gen AI arrived, and Google bailed


AI is so hot right now, and assumed to be the future of everything. But troubles at an Australian AI developer show the field is not all sunshine and roses.

The firm in question is called Appen, and specializes in sourcing and preparing data for AI models. The outfit created a million-strong crowdsourced team of contributors who used their local knowledge or special skills to help classify or annotate data. One big application for its services was helping clients to optimize digital ads, but it also helped prep data for speech recognition and computer vision models.

The startup listed on the Australian Securities Exchange (ASX) in 2015, and by mid-2022 had become one of the 300 most valuable properties on the bourse as its share price topped AUD$35 ($22.75). Its leaders proudly pointed to a client roster that included Google, Amazon, Microsoft, and Salesforce, told investors it was the biggest player in a fast-growing field, and pointed to internal innovations that delivered AI models that shrank the time and cost of labelling video and speech content by 65 percent and 80 percent respectively.

Yet Appen's shares are selling at AUD$0.265 ($0.18) today.

The trouble appears to have started with the post-COVID fall in digital ad spending. August 2022 brought some nasty news: a filing [PDF] revealed "weaker digital advertising demand and a resultant slowdown in spending by some of our large customers."

Investors were told "While some of our customers are slowing the pace of their investments, their AI product development is expected to increase. We remain confident that the AI training data market will continue to grow in the longer term."

A US-based CEO was appointed to get the firm's leadership closer to its clients, and Appen tried to win more work building AI datasets.

Demand for such data did grow – but Appen wasn't able to cash in. In August 2023 it reported a 24 percent revenue drop in its half-yearly results and was chasing a substantial cost reduction target.

By then generative AI had come to dominate discussions in the field, and Appen assured investors it would get into the game.

But in January 2024, some even nastier news: without prior notice, Google decided not to renew its contract with Appen. That contract was worth AU$82.8 million for FY 2023 – and full year revenue from would be down over $100 million year on year to $273 million, with a $20.4 million loss.

On Tuesday the beleaguered biz delivered more bad news: CEO Armughan Ahmed stepped down "to pursue new challenges." COO Ryan Kolin has stepped into the hotseat.

We'll know more about the state of the business when its full year results are revealed on February 27.

The Register expects Appen won't be the last AI company to hit trouble. There are surely dozens if not hundreds of startups that won't survive – because every boom is followed by a bust.

And AI is moving so fast it's almost impossible to know what will work, or when. ®

Send us news
8 Comments

AI-hosted infomercial shifts $7.5 million worth of product in China

PLUS: India tries to untangle TXT marketing opt-ins; China’s AI crackdown succeeds; Australia and Boeing team AWACS, drones; and more!

AI agents get office tasks wrong around 70% of the time, and a lot of them aren't AI at all

More fiction than science

Fed chair Powell says AI is coming for your job

AI will make 'significant changes' to economy, labor market

AI may be after your job, but this AI agent promises to help you get a new one

'Jobright Agent' can apply for jobs on your behalf

EU businesses want a pause on AI regulations so they can cope with unregulated Big Tech players

Mistral fears continental companies may not get time to escape 'distant, behemoth corporations'

C-suite sours on AI despite rising investment, survey finds

Akkodis report suggests people skills may be helpful to bring out the best in AI

Figma files for an (A)IPO with prospectus that mentions AI 150+ times

Warns investors its codebase is harder to maintain as it bakes in brainboxes

Call center staffers explain to researchers how their AI assistants aren't very helpful

Lots of manual corrections and data entry still required

HPE customers on agentic AI: No, you go first

But like cloud computing and digital transformation, this may be a buzzword they can't ignore forever

AI scores a huge own goal if you play up and play the game

A virtual environment makes a great de-hype advisor

Cloudflare creates AI crawler tollbooth to pay publishers

The bargain between content makers and crawlers has broken down

EU rattles its purse and AI datacenter builders come running

176 expressions of interest to erect 'gigafactories' across 16 member states, with 3 million GPUs needed