AI2Day Daily Brief — 3 Sept 2026

Good morning, it is Thursday the third of September, and today we are opening with a story about what actually has to happen before an AI tool can do anything useful inside a large industrial company.

Jabil is a contract manufacturer based in Florida, more than a hundred and forty thousand employees, over a hundred factories, thirty-plus countries. After roughly twenty-five years of letting each site build its own tools, its own spreadsheets and its own workarounds, the company found that data lived in separate pockets invisible to anyone outside that location. When a disruption hit one part of the supply chain, the rest of the network could not see it. The fix they landed on is standardising everything on SAP Integration Suite, a platform that connects separate software systems so they share data automatically. The company's IT director puts the principle plainly: innovation without simplification adds complexity, not value. Predictive supply-chain forecasting is on the roadmap, but only once the shared data foundation is stable. For any large organisation thinking about AI deployments, that sequencing is the real lesson.

Straightforward rule, easy to ignore when there is pressure to show results fast. Next, Adobe has made a quiet acquisition.

Adobe bought Rilo, a six-person Indian startup founded in 2025 by graduates of the Indian Institute of Technology. The price was not disclosed. Rilo built software that lets marketing and sales teams automate repetitive workflows without needing an engineer to wire each one together: pulling insights from sales calls, reformatting content for different platforms, tracking competitor activity. The team and the technology are moving to Adobe, and Rilo itself is shutting down, which means existing customers will lose access. This is Adobe's second acquisition from India, following its purchase of video platform Rephrase dot ai in twenty twenty-three. It sits alongside Adobe's one point nine billion dollar deal for SEO company Semrush last year, and together those moves signal a deliberate push into marketing intelligence.

A small deal with a clear strategic line. Now to a story with serious legal and safety dimensions.

Thirty new lawsuits were filed in a California federal court on Wednesday against OpenAI and its chief executive Sam Altman. The plaintiffs are students, teachers and a principal connected to the school shooting at Tumbler Ridge Secondary School in British Columbia, Canada. The core claim is that OpenAI's own automated safety system flagged the alleged shooter's conversations with ChatGPT about gun violence, and that the company chose not to alert Canadian authorities. The lawsuits name OpenAI's chief global affairs officer, Chris Lehane, as being involved in a decision to stay silent over concerns about reputation and finances. The suits also allege that after the account was flagged, OpenAI only deactivated it rather than issuing a full ban, which they say allowed the suspect to re-register with a new email address. OpenAI's chief strategy officer, Jason Kwon, has publicly called those claims false.

That is a case worth following closely as it moves through the courts. Our fourth story is about a structural problem that goes beyond any single company or sector.

A Harvard working paper covering sixty-five million workers at more than two hundred and eighty thousand US firms found that junior employment fell roughly nine percent within six quarters after companies adopted generative AI. A separate Stanford analysis of payroll records found the youngest workers in the most AI-exposed jobs lost ground after late twenty twenty-two while more experienced colleagues held steady. The deeper issue is what disappears with those entry-level roles: the repetitive, unglamorous work that builds the instincts experts later rely on in a crisis. You cannot skip being a junior and still become a reliable senior. Aviation ran into the same problem with autopilot making pilots rusty, and the industry's answer was deliberate: build manual practice back into training rather than remove the automation. Researchers studying AI workplaces are proposing a similar fix they call a manual gate, a designed-in moment where a human does the work without AI assistance specifically to keep a skill alive.

Deliberately preserving the hard way of doing something, that is a genuinely counterintuitive design choice but the data makes a case for it. Finally, a cloud computing story with a direct access angle.

Today's sponsor is Zero2Hero: an AI poker coach sits at your table, and you can ask it mid hand whether the player across from you is bluffing. The free tier is real: 50 hands a day against AI opponents, no card. Not a trial. Lifetime access is $99, in US dollars, until 15 September, at zero, the number two, hero, dot A I.

Reliance Jio, India's most valuable company, opened its JioPC cloud service on Wednesday to any Indian internet user, dropping the earlier requirement to be a Jio broadband subscriber. JioPC is a virtual computer that runs inside Jio's data centres and streams to whatever screen is in front of you. Plans start at roughly eleven US dollars for two months, with the highest-tier configuration running about fifty-three dollars per year. The target market is significant: IDC puts India's installed base at more than sixty-five million PCs, with replacement cycles now stretching to five or six years for consumers. An eight-year-old laptop could, in theory, run software its own processor could never handle. One open question Jio has not answered: whether the servers powering the service carry dedicated AI chips, which matters for anyone hoping to run the new generation of AI software on it.

For more depth on all of this, join us on Monday for AI Today Weekly. That is your AI briefing for today. Every story is at A-I-2-Day dot live. That is A, I, the number two, D-A-Y, dot live. We are back tomorrow morning. If this was useful, hit the thumbs up and subscribe, so the next one finds you.