Welcome back. This is the first edition of WSJ Pro AI’s relaunched weekly newsletter. We start with an inside look at BMW AG’s move to ramp up deployment of artificial intelligence-powered quality-control automation. As Covid-19 lockdowns brought the auto industry to a standstill, the German auto maker began testing software algorithms in near-empty factories by simulating the process of scanning car parts for defects, WSJ Pro's David Uberti reports. Company officials say honing the system would put the German auto maker on a better footing when the economy begins to pick up steam.

BMW is one of many manufacturers turning to smart robotics and automation during pandemic, either to cope with workplace social-distancing restrictions or to take advantage of downtime to fast-track modernization efforts, and this outlay is fueling projections by Morgan Stanley and others of a sharp upturn in industrial-automation spending in the second half of the year.

How it works. BMW’s quality-control software analyzes still photos from different stages of the assembly line, using image-recognition capabilities, and compares them with the way a vehicle is meant to be assembled. It then flags any deviation, such a loose or missing screw, to a shop-floor foreman, as parts continue moving along conveyor belts.

Tighter budgets, tighter operations. The system replaces manual checks and lowers the risk of costly back-end repairs, while boosting efficiency, said Matthias Schindler, head of artificial intelligence innovation for BMW’s production systems group.

Katharina Stenholm, senior vice president and chief cycles and procurement officer, told Pro AI that smart tools enabled the company to quickly reconfigure its supply chain and keep operations running as Covid-19 spread.

Pre-crisis investments in AI helped Norsk Hydro ASA keep its factories humming when access by engineers was limited, said Jo De Vliegher, the aluminum giant’s chief information officer. A project to improve the quality of data sent to AI-enabled software bots monitoring the performance of cooling pumps, compressors and other machinery allowed engineers to oversee production without entering a plant, Mr. De Vliegher said.

Funding dollars for AI startups in the second quarter dropped 14%, from the opening quarter, while the number of deals dropped 11%, as investors fled risky bets on untested technology, according to CB Insights.

Delivery robots. As demand surges for delivery services, a handful of startups are attracting venture capital with fleets of semi-autonomous robots designed to deliver food orders and tested in recent years on college campuses. Developers at robotics startup Kiwi Campus, which recently raised $5 million in capital, say the pandemic is accelerating efforts to expand into off-campus neighborhoods and cities. (WSJ) 

Robot uprising. Robotics startups are facing looming competition from tech giants such as Amazon.com Inc. and Postmates owner Uber Technologies Inc., which are testing similar technology to capture growing demand. Amazon last week began delivering packages with its rolling robot in Atlanta and Franklin, Tenn., adding to its existing service in Washington state and Southern California. Postmates operates its own fleet of robots in Los Angeles.

Early-mover advantage fades. Roughly 75% of 2,727 executives say AI will be a key component of their enterprise applications within three years, Deloitte said in a report last week. But the pace of that growth is taking the competitive edge away from companies that were quick to adopt smart capabilities, researchers said. “Although adopters are still bullish on AI, their advantage may wane as barriers to adoption fall and usage grows." (Forbes)

Facebook settles facial-recognition lawsuit. The social media giant has agreed to a $650 million settlement in a long-standing Illinois class action suit over its use of facial recognition software to identify users in photos posted on the site. Illinois requires businesses to obtain consent before collecting biometric data, including facial recognition data. (Vox) 

Fundraising for test-prep app. South Korea-based Riiid, an AI-powered exam-studying software startup, has closed a $41.8 million fundraising round, boosting its total VC capital raised to date to more than $70 million. The company says the money will go into accelerating its expansion across the U.S., South America and the Middle East, along with an R&D lab in Silicon Valley. (VentureBeat) 

Limits of learning. An analysis of more than a thousand research papers warns that deep learning may be reaching its computational limits, weighed down by mountains of data and the growing demands for image classification, object detection, question answering, named entity recognition and machine translation, according to a group of researchers led by the Massachusetts Institute of Technology. 

The Hang Seng Tech Index, a new Hong Kong-based technology benchmark that includes Tencent Holdings Ltd. and Alibaba Group Holding Ltd., fell below the broader stock market in its trading-day debut, ending the day down 1.3%. (WSJ) 

Google said it will keep employees home until at least next July, a move affecting roughly 200,000 full-time and contract workers across parent-company Alphabet Inc. (WSJ) 

Email systems used by thousands of county and local election officials who will be administering November’s presidential election are vulnerable to online attacks, according to a report by cybersecurity firm Area 1 Security Inc. that tracked more than 12,000 accounts. (WSJ) 

Write to Enterprise Technology Editor Steven Rosenbush and newsletter author Angus Loten.

Follow us on Twitter: @wsjproai, @steve_rosenbush, @J_J_McCormick, @tomloftus, @SCastellWSJ, @JaredCouncil  and @angusloten.