Hello and welcome to Protocol Enterprise! Today: how AI is being used to spot manufacturing defects, why Salesforce is focused on integration and Elastic CEO Shay Banon steps down.
Spin up
Morgan Stanley’s quarterly survey of U.S. CIOs is a closely watched barometer for enterprise tech demand, and Altimeter Capital’s Jamin Ball was kind enough to tweet a few highlights. CIOs reported running 25% of their workloads in the cloud today, but expect that to jump to 44% by the end of 2024.
Small data for big problems
To the uninitiated, a tiny stain on several yards of car seat upholstery or a minuscule gas bubble on the surface of an industrial oil pipe might seem like an insignificant imperfection. But factory inspectors are always on the lookout for these sorts of defects, because they can create serious slowdowns in time-sensitive manufacturing production schedules.
Companies including Landing AI and Mariner are helping take defect detection to the next level with AI software, betting that manufacturers want highly customized algorithmic models to monitor for product defects. And they have another selling point that flies in the face of what we know about most big data-hungry AI systems: Their models work using very small datasets.
- Manufacturers like Stanley Black and Decker have relied on humans to scrutinize X-ray images on dark room monitors, said Amir Kashani, director of AI and digital products for the company’s industrial division.
- “You’re looking at tea leaves of cloudy shapes,” Kashani said. AI can spot those sorts of obscured defects more readily, he said.
Pre-built machine-learning models offered by companies like IBM, AWS and Microsoft don’t provide the level of control or customization manufacturers want, according to the newcomers.
- Landing AI CEO Andrew Ng said he expects the manufacturing industry will need “tens of thousands” of unique AI models for each and every unique type of product and imperfection.
- Machine-learning models that are not built to specific company needs could cause new problems, said David Dewhirst, vice president of marketing at Mariner, which sells visual inspection hardware and image processing software that attaches to factory camera systems.
- He said he’s seen systems built from models that were not trained with company-specific visual data mistakenly flag a harmless speck of lint when looking for real defects like stains, for example, and false positives can trigger unnecessary production line slowdowns, which can cost time and money.
Product anomalies don’t happen very often when the manufacturing process is going smoothly, and that actually creates a dilemma for manufacturers that want to use AI to spot them.
- Most manufacturers don’t have millions, thousands or even hundreds of examples of a particular type of flaw they need to watch out for. In some cases, they might only have 20 or 30 photos of a windshield chip or small pipe fracture, for example.
- Providing thousands of images to train a model to detect a break in a piece of equipment or surface coating imperfection is “not feasible,” said Marcos Fernandez, director of Engineering Consulting at Capgemini.
- “Data is food for AI,” said Ng. “It’s not just about massive numbers of calories; it’s about good food.”
- As an influential AI engineer, Ng recognizes that he has come 180 degrees from where he was a decade ago as an evangelist for pumping giant datasets into neural networks.
And there’s another tricky problem: For AI, data quality goes hand-in-hand with data labeling or annotation.
- While one subject matter expert in one country might refer to a steel welding defect as a “slag,” a colleague in a different locale or even just from another company might call it by a different name, or argue over how to define it or even whether a defect is present at all.
- “There’s no ground truth,” said Kashani. “Two experts might not agree.”
A MESSAGE FROM CLARI
Think success is predictable? Join the Club. Watch the season premiere of Clari's Club Revenue on Nasdaq as Pilar Schenk, COO of Cisco Collaboration, unpacks the rapid growth trends driving the future of revenue.
In other news
In enterprise software, ongoing success still requires integration — especially at Salesforce.
During a talk at the Jefferies Software Summit Wednesday, Bill Patterson, executive vice president for CRM applications at Salesforce, said the company’s acquisition of Slack was about providing a “uniform layer to work across applications,” which helps bring continuity to workflows. Earlier in the day Slack CEO Stewart Butterfield also pointed out the need for greater overall integration across the SaaS industry.
At a time when companies are struggling to manage dozens of applications and disparate workflows, interoperability between applications — even between competitors — is an imperative. Butterfield noted how customers want better connectivity between Slack and Teams, which was part of the reason Slack sued Microsoft in Europe last year citing unfair competition.
The strategic shift for Salesforce will be to focus on making work as a whole easier, not just thinking about it “app by app or API by API,” Patterson said. But MuleSoft, the integration platform Salesforce acquired in 2018 to do just that, underperformed last quarter. And investors at the conference expressed concern about Salesforce’s ability to provide a cohesive integrated service.
It goes to show that providing a truly integrated experience — no matter where the application functionality sits — is much easier said than done.
Startup Wednesday
AI hype is seasonal; proponents always worry about the dawn of another “AI winter,” in which even enthusiasts start to lose faith in its potential to impact the average person. But AI is making a clear impact in niches like manufacturing, and another place where AI technologies are showing real promise is IT operations.
BigPanda just raised $190 million for its AIOps technology, valuing the tools startup at $1.2 billion. Think of BigPanda as AI-meets-observability; its software aggregates data from across IT to help companies understand how their apps are performing and resolve problems by automating responses to common issues.
IT automation has been on a steady march for over a decade with the rise of infrastructure-as-code tools and serverless computing. While our self-driving cars might not arrive for years to come, our data centers are well on their way to allowing machine learning to take the wheel. Stay tuned for an in-depth Protocol Enterprise look at AIOps later this year.
Around the enterprise
Elastic co-founder and CEO Shay Banon stepped down but will remain with the company as CTO. Ash Kulkarni is the new CEO.
Microsoft’s server processor plans might be taking shape now that the company hired former Arm designer Mike Filippo away from Apple.
Christofer Dutz, an open-source maintainer working on the Apache PLC4X project, explained why he’s no longer providing free support for the project amid a wider debate about the true costs of open-source software.
Protocol Enterprise always appreciates a straightforward outage postmortem. Shopify shared the tale of how an expired certificate nearly took down its entire operation.
A MESSAGE FROM CLARI
Club Revenue breaks down the tactics of the sharpest minds shaping the digital transformation of sales.
Hosted by Clari's Cornelius Willis and presented by Nasdaq, each episode features a revenue virtuoso and their creative strategies for high performance and explosive growth. Watch now to learn how to transform your sales teams into revenue machines.
Thanks for reading — see you tomorrow!