
This is "The AI Economy," a weekly LinkedIn-first newsletter about AI's influence on business, work, society and tech and written by Ken Yeung. Sign up here.
I love speaking with entrepreneurs to understand better what they’re building and how it’ll impact our lives—it’s why I became a journalist. And some have fascinating stories that I’m glad I can tell, like with Bem. This is an early-stage data infrastructure company developing a solution many companies might have that could define a new AI category.
This week’s issue of “The AI Economy” explores what makes Bem unique and why organizations should pay attention. Then, read about the newest models and capabilities coming to Google’s Vertex AI platform to make enterprise AI app building better.
Finally, the newsletter concludes, as always, with this week’s roundup of AI headlines you may have missed.
Enjoy!
Automating the Unstructured to Structured Data Pipeline
Every organization has a treasure trove of data that can be used to create robust in-house AI systems. The problem, however, is that only some of it is in a compatible format, and getting to that point requires time and engineering staffing resources companies need. Earlier this month, I spoke with Antonio Bustamante, an entrepreneur with a startup called Bem that tackles this issue and eliminates a persistent barrier to entry.
“The big problem we’re solving is that engineers spend incredible amounts of time making sure that systems talk to each other—which seems like such a low bar. But it’s such a ‘hair on fire’ problem today,” he tells me. “The companies that I’ve built so far, it was always the first problem that we run into. Some of the systems that don’t talk to each other are entirely computer-based and some are human-based, like email, fax, texts and WhatsApp.”

Borrowing a playbook similar to Stripe, Bem utilizes a simple API that’s easy to deploy. It creates a feed in which customers send invoices, emails, videos, and unstructured data files, and the startup uses AI to convert them into “whatever data shape and form they use internally.” In its way, Bem has created a Structured Data as a Service (SDaaS) offering in which companies automate this conversion process, and the magic just happens.
Bustamante doesn’t push back on that analogy: “I never thought about it like that, but I’d say that’s a great of putting it. All of these inputs they’re sending us are incredibly messy. All our users need is a structured version of all those inputs that look like their internal data shape and schema. And so at the end of the day, yes, we’re offering them a transformation structuring service.”
However, Bem is not alone in this SDaaS category. There are others, such as Unstructured.io.
Bem isn’t a one-off service, either. It provides a feed that handles the conversion behind the scenes. “The customer sets up a pipeline and sends us thousands and thousands of pieces of data every minute, every hour. It’s a continuous stream of data.”
Structured data allows AI algorithms to process, analyze, and derive insights from information more readily. It’ll also make interpreting the data easier, running queries against it, and help train machine learning models. Organizations wanting to implement AI systems need accurate and up-to-date information, so if models lack specific information because those files aren’t in an appropriate schema, how can executives make the correct decision?

Investors are paying attention to Bem. It recently secured $3.7 million in venture funding in a round led by Uncork Capital.
“Every company I’ve ever worked with has always had some issue with messy data ingestion,” Andy McLoughlin, Uncork Capital’s managing partner, explains. “For some companies, if they’re working in insurance, logistics or agriculture, it’s even more pronounced, but every company has this. Right now, the status quo has been that docs come in, and somebody has to manually review it. Having something that automatically does it for you and it just works is such a no-brainer.”
As more companies become AI companies in the future, they will need the staffing and expertise to implement foundational AI. McLoughlin states that organizations will resort to off-the-shelf models and tooling by then. “But that’s commoditized,” he argues. “The thing that isn’t commoditized is the quality of the data they have to hand in order to get the best output. Yes, there’s going to be a lot of money made in the AI piece and AI infrastructure, but I think the data infrastructure is as important, if not more important, going forward.”
Bem is currently in a private beta and says its service best suits companies in the Series C and public stages. For Bustamante, Series C is where “you reach this critical mass where this problem becomes incredibly evident.”
▶️ Read more about Bem (VentureBeat)
Google’s Vertex AI Enterprise Scale Up
This week, Google made a few announcements to show enterprise customers that its Vertex AI platform has the features they need for any AI use case. Some are a follow-up to things the company introduced earlier this year at its I/O developer conference, such as the launch of Google’s Gemini 1.5 Flash and 1.5 Pro with a 2 million context window. The company is also making its open-source lightweight Gemma 2 model series available to more users starting next quarter. However, it surprised us with not one, but two sizes—along with a 27 billion parameter model, there’s a 9 billion one.
Organizations can also use Google’s next-gen text-to-image foundation model through Vertex AI. Like Gemini 1.5 Flash and Gemma 2, Imagen 3 first debuted in May, offering faster image generation, better prompt understanding, photo-realistic people generation, and greater text rendering control. The new model will help Google remain competitive against other similar services enterprise customers might turn to, such as Adobe Firefly, OpenAI’s DALL-E, and Midjourney.
Finally, to give businesses peace of mind that their AI models won’t hallucinate and will provide factual and accurate information, Google is partnering with four reputable third-party services to help “ground” the technology in real-world facts. It’s the first time the company has enlisted the help of external partners—previously, only Google Search was available to provide grounding through web data. Now, enterprise customers can select data from Moody’s, MSCI, Thomson Reuters and Zoominfo.
What is grounding? It’s anchoring an AI’s output or response to specific, factual information or real-world context. And for those in sensitive or regulated industries, there are significant consequences if a model spits out hallucinations, no matter if the AI is used internally or externally.
As companies turn to platforms like Vertex AI to build AI-powered apps, chatbots and agents, providers like Google must beef up their offerings so that every aspect of what developers might need is state-of-the-art.
Today’s Visual Snapshot

Gartner has published a new study exploring who implements AI within an organization. With talk about the need for Chief AI Officers, more than half of those polled reported that their AI leader is someone other than a CAIO. 55 percent say a committee leads their efforts.
“Accountability for AI is spread out. Additionally, some organizations are decentralized, siloed or unclear as to where AI initiatives should lie,” the research firm writes. Some positions tasked with leading AI include analytics and strategy leaders, heads of digital business, and those in the C-suite. CIOs top the list as being the ones most responsible for AI accountability in organizations, with 25 percent of respondents saying that role is in charge of AI orchestration.
Interestingly enough, with all the clamor about CAIOs, boards are reluctant to push companies to hire them because they don’t want to expand the C-suite. However, they do want an AI leader.
“AI and Gen AI are complex and far-reaching and touch every job, activity and strategic conversation in the organization,” Frances Karamouzis, Gartner’s distinguished VP analyst, remarks. “However, this does not mean that the people or team responsible for orchestrating AI at an organization have to have a title at the altitude of the C-suite.”
Quote This
Perplexity is not ignoring the Robot Exclusions Protocol and then lying about it. I think there is a basic misunderstanding of the way this works. We don’t just rely on our own web crawlers, we rely on third-party web crawlers as well.
— Perplexity Chief Executive Aravind Srinivas, in an interview with Fast Company, responding to allegations that the AI search startup is crawling content from websites that don’t want to be crawled.
Can’t Miss Event
VentureBeat’s flagship conference is coming up.
You’re invited to attend VB Transform 2024 in San Francisco, California, from July 9 to 11. There’s an incredible lineup of speakers and engaging topics. Plus, you’ll have plenty of top-tier networking opportunities.
Scheduled to speak are OpenAI’s Head of Product, API, Olivier Godement; Microsoft’s Corporate Vice President for AI at Work, Jared Spataro; Perplexity CEO, Aravind Srinivas; Groq CEO Jonathan Ross; Google Cloud’s Global Head of Regulated Industries, Zac Maufe; Nvidia’s Vice President of AI Models, Software and Services Kari Briski; Walmart’s Vice President of Emerging Technology Desirée Gosby; Wayfair’s Chief Technology Officer Fionna Tan; and more.
This is an excellent event for enterprise executives interested in learning about practical generative AI case studies and applications directly from industry leaders
This Week’s AI News
🏭 Industry Insights
- Mark Zuckerberg calls closed-sourced AI competitors as those trying to “create God” (TechCrunch)
- AI companies need to get society’s input as they build out the technology, OpenAI CEO Sam Altman and Airbnb CEO Brian Chesky say (Business Insider)
- Inside billionaire Alexandr Wang’s relentless rise in Silicon Valley (The Information)
- LMSYS’ new “Multimodal Arena” leaderboard compares AI models’ performance on vision-related tasks: OpenAI’s GPT-4 tops rankings, but AI fails to out-see humans (VentureBeat)
- Goldman Sachs CIO: AI tools may offer superhuman productivity but may not be “superhumanly smart” (Fortune)
🤖 Machine Learning
- Valued at $2 billion, can startup DeepL beat Big Tech on AI language translation (Semafor)
- Unababel is among the first AI startups to win millions of GPU training hours on EU supercomputers (TechCrunch)
✏️ Generative AI
- Amazon is reportedly working on a Metis chatbot to rival OpenAI’s ChatGPT designed with conversational flair and autonomous abilities (Mashable)
- Instagram to start letting some creators make AI versions of themselves through its AI Studio tool (The Verge)
- ElevenLabs launches an iOS app that turns any text into audio narration with AI (VentureBeat)
- Google is reportedly building AI chatbots based on celebrities and influencers (Engadget)
🛒 Commerce
- How Shake Shack is using Reddit and AI to drive sales (Adweek)
- Retailers like Target and Walmart are bringing generative AI into physical stores (Modern Retail)
- Shopify’s new “Sidekick” AI chatbot for merchants is launching in early access (The Verge)
☁️ Enterprise
- Box removes its enterprise AI query limits and announces new GPT-4o-powered feature (Fast Company)
- How Anthropics’ “Projects” and new sharing features are revolutionizing AI teamwork (VentureBeat)
- Orby is building AI agents for the enterprise (TechCrunch)
- Google adding Mistral Small, Large and Codestral models to Vertex AI (VentureBeat)
⚙️ Hardware
- Intel reveals first fully integrated optical compute interconnect, designed to scale up AI workloads (Silicon Angle)
- Microsoft Surface Pro Copilot+ PC review: The best Surface tablet ever made, no thanks to AI (Engadget)
- Etched is building an AI chip that only runs one type of model (TechCrunch)
- Researchers prove Rabbit AI security breach by sending email as admin (404 Media)
- ChatGPT is now available in many Volkswagen models (Engadget)
🔬 Science and Breakthroughs
- How AI revolutionized protein science, but didn’t end it (Quanta Magazine)
- Sonia’s AI chatbot steps in for therapists (TechCrunch)
- The mystery of AI gunshot-detection accuracy is finally unraveling (Wired)
💼 Business and Marketing
- Figma unveils AI-powered design tools, challenges Adobe’s dominance (VentureBeat)
- Figma’s new Slides app focuses on design, fun, and AI (Fast Company)
- Toys “R” Us Studios uses OpenAI’s Sora to generate one-minute-long brand video (CNET)
- Salesforce: AI coworkers are officially here and people are OK with it (Digiday)
📺 Media and Entertainment
- OpenAI and Time strike a multiyear deal to improve ChatGPT with journalistic content (CNBC)
- NBC to use an AI version of legendary sportscaster Al Michaels to deliver Olympic recaps on Peacock (The Verge)
- YouTube in talks with record labels over AI music deal (Financial Times)
💰 Funding
- Nvidia challenger Groq is reportedly set to raise a fresh funding round by Blackrock at a $2.5 billion valuation (Business Insider)
- Hebbia reportedly raises nearly $100 million for its AI-powered document search (TechCrunch)
- AI-focused Formation Bio secures $372 million from Andreessen Horowitz and Sanofi to license new drugs (Fierce Biotech)
- EvolutionaryScale lands $142 million to advance AI in biology (Reuters)
⚖️ Copyright and Regulatory Issues
- The Recording Industry Association of America is suing Udio and Suno, alleging the AI startups wrongfully used popular artists’ work to produce copyrighted music without consent (CNN)
- Amazon investigating Perplexity over claims of scraping abuse (Wired)
- Publisher’s perplexing AI problem (Axios)
- Reddit puts AI scrapers on notice not to ignore its instructions or risk being blocked (Engadget)
- Quora’s chatbot platform Poe allows users to download paywalled articles on demand (Wired)
💥 Disruption and Misinformation
- Profile of GetReal Labs and how this startup aims to detect all the deepfakes (Wired)
- Garbage in, garbage out: Perplexity spreads misinformation from spammy AI blog posts (Forbes)
- How a journalist turned to AI to plan a trip to Norway (The New York Times)
End Output
Thanks for reading. Be sure to subscribe so you don’t miss any future issues of this newsletter.
Did you miss any AI articles this week? Fret not; I’m curating the big stories in my Flipboard Magazine, “The AI Economy.”

Connect with me on LinkedIn and check out my blog to read more insights and thoughts on business and technology.
Do you have a story you think would be a great fit for “The AI Economy”? Awesome! Shoot me a message – I’m all ears!
Until next time, stay curious!
Subscribe to “The AI Economy”
New issues published on Fridays, exclusively on LinkedIn
Leave a Reply
You must be logged in to post a comment.