AI News Archive: September 2, 2026 — Part 4
Sourced from 500+ daily AI sources, scored by relevance.
- Google Pics Rolls Out to Workspace With Default-On AI Image Tools
Google Pics is rolling out to eligible Workspace customers with AI image generation, editing tools, and admin controls enabled by default. The post Google Pics Rolls Out to Workspace With Default-On AI Image Tools appeared first on TechRepublic .
- Zuckerberg, Musk make plea at G20 for more AI data centers
Mark Zuckerberg and Elon Musk on Tuesday stressed the need for more AI data centers and the electric power needed to support them at a meeting of G20 technology ministers focused on the disruptions and opportunities brought by artificial intelligence.
- Conveo raises $50M to scale its AI-powered consumer intelligence platform
Consumerresearch technology company Conveo has raised $50 million in Series A fundingto expand in the US and develop its platform into a broader consumerintelligence infrastructure for enterprises. Th...
Score: 48💰 MoneySep 2, 2026https://tech.eu/2026/09/02/conveo-raises-50m-to-scale-its-ai-powered-consumer-intelligence-platform/ - MrBeast partners with Gemini to turn impossibly big ideas into reality
Mr. Beast
Score: 48🌐 MovesSep 2, 2026https://blog.google/company-news/inside-google/company-announcements/mrbeast-gemini-google-health/ - Continuous identity becomes the new front line for AI agents: theCUBE’s Fal.Con 2026 day two keynote analysis
Identity security has always worked the same way: log in once, get trusted until you log out. That model breaks down the moment an AI agent does the logging in, since an agent can call a dozen tools in the time it takes a human to read one email. The industry’s fix is continuous AI […] The post Continuous identity becomes the new front line for AI agents: theCUBE’s Fal.Con 2026 day two keynote analysis appeared first on SiliconANGLE .
- When AI starts doing the work: The new economics of software pricing
AI calls for a new commercial lens as enterprise software moves from enhancing capabilities to creating new outcomes and performing work. Ashish Nayyar examines how this shift is changing the economics of software and the principles that should guide pricing, packaging and value capture.
- The Modern CUDA Toolbox in Practice: A Step-by-Step Optimization Walkthrough
NVIDIA CUDA remains the foundation of GPU-accelerated computing, powering everything from scientific simulations to large-scale AI training. But writing...
- Agentic AI could improve corporate cash forecast accuracy to 90%: EY
EY India says treasury teams spend up to 70% of their time on manual work, while agentic AI could improve cash forecast accuracy to as much as 90%
- Why Elon Musk is making turbine blades for AI data centres
Why Elon Musk is making turbine blades for AI data centres YourStory.com
Score: 48🌐 MovesSep 2, 2026https://yourstory.com/ai-story/spacex-turbine-blades-ai-data-centres-power - AI infrastructure cost: Where is America’s $1.4 trillion going?
AI infrastructure cost: Where is America’s $1.4 trillion going? YourStory.com
Score: 48🌐 MovesSep 2, 2026https://yourstory.com/ai-story/ai-infrastructure-cost-america-14-trillion - Former Google applied AI expert launches AI startup Guickly, raises $4.2M in seed funding
Former Google applied AI expert launches AI startup Guickly, raises $4.2M in seed funding YourStory.com
Score: 48💰 MoneySep 2, 2026https://yourstory.com/ai-story/guickly-raises-4m-to-bring-tighter-control-to-enterprise-ai-spending - With Catalyst 3.0, Zoho focuses on what happens after AI writes the code
Catalyst 3.0 is Zoho’s attempt to close that gap by connecting AI coding assistants directly with the cloud infrastructure needed to run an application.
- Spotify draws its AI line at deception, not how the music was created
Spotify will not ban AI-generated music but will focus on preventing impersonation. The company introduced AI Credits for artists to disclose AI usage in songs. New AI Persona badges will identify profiles that appear to be AI-generated. This approach aims to inform listeners and protect genuine artists from deception.
- Hey EU, your new rules for ChatGPT don’t cover chat
The world's most popular chatbot is now under Europe’s stringent regulatory purview. But classifying it as a search engine leaves many of its core functionalities uncovered.
- Betting on AI and Robots to Automate Superconductor Discovery
Startups seek out new materials that can be made into usable wire
- China cracks down on AI ‘slop’, clearing out clutter from WeChat, RedNote, Douyin
China is cracking down on artificial intelligence-generated “slop” – low-grade, mass-produced synthetic content cluttering social media and news feeds – as part of a sweeping campaign against the abuse of the generative technology, the country’s top internet watchdog said on Wednesday. In a statement posted to its website, the Cyberspace Administration of China (CAC) said it had scrubbed more than 5.61 million pieces of harmful or illegal content and removed some 49,000 accounts during its...
- Warren Buffett piled into Alphabet to bet big on AI, successor Greg Abel says
Warren Buffett piled into Alphabet to bet big on AI, successor Greg Abel says Business Insider
- Palantir CEO Alex Karp backs ousted Ukraine defense minister's new defense tech startup
Fedorov is launching a defense tech company backed by Palantir CEO Alex Karp, aiming to turn Ukraine’s wartime experience into new technologies.
Score: 48🌐 MovesSep 2, 2026https://www.cnbc.com/2026/09/02/palantir-alex-karp-ukraine-federov-defense-startup.html - Tencent’s Hy4 model gains in open-source AI rankings after ecosystem-driven training
Tencent Holdings’ use of its vast product ecosystem to train its new Hy4 preview model gives it an edge in developing AI agents and brings its flagship model suite back into the top tier of open-source offerings, according to analysts. The Chinese tech giant’s “differentiated product-plus-model strategy”, where preview models were first deployed across Tencent’s suite of products, enabled it to collect user data before feeding the information back into subsequent rounds of training, Goldman...
- NSF launches a new AI research operations center
The new center will help support the National Artificial Intelligence Research Resource’s transition from a pilot into a permanent part of the agency.
- Fable 5.1 on Frontier Coding Tasks: Efficient Successes, Distinct Failure Modes
We evaluated Fable 5.1 on a series of frontier coding tasks from our proprietary Terminal-Bench+ dataset and compared the results against Opus 5. Fable remained competitive across most categories and was materially more efficient on successful runs, while its gap was concentrated in a small set of terminal-heavy and build/dependency tasks. Because category sizes are small and uneven, we treat... The post Fable 5.1 on Frontier Coding Tasks: Efficient Successes, Distinct Failure Modes appeared first on Snorkel AI .
- Meta prices Muse Voice Transcribe at $0.18 an hour, with real-time diarization for 20+ speakers: a steal for enterprises?
Meta is entering the increasingly competitive real-time speech-to-text market with Muse Voice Transcribe, a new audio perception model that combines streaming transcription, endpoint detection and speaker diarization for more than 20 speakers — at a public API price of just $0.18 per hour of processed audio. Developed by Meta Superintelligence Labs, Muse is designed to process speech while it happens rather than waiting for a recording to finish. Meta’s launch post for Muse Voice Transcribe says the model supports long audio exceeding an hour, seamless multilingual code-switching, language and keyword biasing, and diarization without a separate post-processing pipeline. The model was trained across more than 70 languages, with 25 extensively validated for the initial release. The 20-plus-speaker figure is substantial, but it is not a world record. A review of current vendor documentation turns up systems with higher published ceilings. Speechmatics' real-time transcription service says it can identify 50 speakers by default and up to 100 when the limit is increased, while Amazon Transcribe's diarization documentation specifies a maximum of 30 unique speakers, including for streaming transcription. ( Speechmatics ) Muse nevertheless lands toward the high end of the market, and Meta's broader proposition is arguably more important than the raw maximum: high-capacity real-time diarization combined with low-latency transcription, endpointing, multilingual code-switching and aggressive API pricing in the same model. For enterprise developers building meeting systems, call analytics, live assistants or ambient AI, that combination could matter more than who holds the speaker-count record. Diarization is becoming part of the core voice stack Traditional speech recognition answers a relatively simple question: What was said? Diarization adds another: Who said it? That distinction becomes critical as transcripts feed downstream AI systems. A meeting assistant can correctly transcribe every sentence and still create an unreliable corporate record if it attributes an approval, commitment or objection to the wrong participant. The same issue affects customer-service analytics, compliance workflows and AI agents operating in rooms where several people can speak. Muse incorporates speaker attribution directly into its autoregressive multimodal architecture. Meta says audio arrives in 80-millisecond chunks, or 12.5 chunks per second, with each transformed into a soft token. At each step, the model decides whether to consume more audio or emit text. Meta calls this mechanism adaptive delay: rather than applying one latency budget to every word, Muse can wait longer when speech is ambiguous and commit earlier when it has enough context. Meta says reinforcement learning combines word-error-rate and delay rewards to train that behavior. Meta's technical explanation of Muse details the architecture. ( Meta AI Research ) Speaker attribution and endpointing then become part of the same token sequence. A <|start_of_turn|> token marks a potential new speaker turn, tokens such as <|speaker_A|> identify the speaker, and separate onset and endpoint tokens identify speech boundaries. Meta says it trains ASR, diarization and endpointing together rather than running speaker clustering as an unrelated downstream process. Meta's Model API speech-to-text documentation also exposes diarization as a first-class operating mode alongside push-to-talk and endpointing. Speaker labels such as A and B are scoped to a session rather than verified identities, and the API provides turn-level rather than word-level timestamps. 20+ speakers is high, but Speechmatics goes considerably higher Speaker-count comparisons require care because vendors implement diarization differently and do not all publish a maximum. Speechmatics currently makes the strongest explicit real-time capacity claim found in this review. Its real-time STT documentation says speaker diarization is available live, while its real-time FAQ says the system supports 50 speakers by default and can be increased to 100. AWS likewise exceeds Meta's stated figure: Amazon Transcribe can differentiate a maximum of 30 unique speakers, and AWS provides explicit instructions for speaker partitioning in a streaming transcription. Soniox supports diarization in both real-time and asynchronous processing, but documents a maximum of 15 speakers per session. AssemblyAI's streaming diarization system lets developers set max_speakers between one and 10. Both companies caution that live speaker attribution is more difficult because streaming systems must make decisions with less future audio context than offline models. xAI's current Speech-to-Text API also supports speaker diarization in streaming mode, but its documentation reviewed for this story does not publish a maximum diarized-speaker count, so a direct ceiling comparison with Muse is not possible. ( X.ai Docs ) That means it would be inaccurate to describe Muse's 20-plus capability as a new global record. The highest explicitly documented real-time number identified in this survey is Speechmatics' configurable 100-speaker ceiling. Meta also does not demonstrate 20-plus simultaneous participants in its launch material. Its principal live demonstration uses eight speakers, while its long-form recording contains 11 labeled participants. The 20-plus number is a stated model capability rather than the participant count in the public demos. At $0.18 per hour, Muse competes aggressively on price Meta's pricing makes the competitive picture more interesting. According to its Muse Voice Transcribe developer page , Muse costs $3 per 1,000 minutes, or $0.18 per hour. Streaming and non-streaming transcription cost the same, and Meta says zero-data-retention processing is priced at parity with standard processing. Billing applies to audio actually processed and is rounded down to whole seconds. Standardizing publicly posted rates to one hour of streaming audio gives the following rough comparison: Streaming speech-to-text service Approx. public cost/hour Real-time diarization Soniox stt-rt-v5 $0.12 Included; up to 15 speakers Meta Muse Voice Transcribe $0.18 Included; 20+ speakers xAI Speech to Text $0.20 Supported; maximum not stated Speechmatics Real-time Standard $0.24 Included; 50 default, configurable to 100 Qwen3 ASR Flash Realtime ~$0.324 international No comparable maximum documented in source reviewed Deepgram Nova-3 Multilingual ~$0.35 base / ~$0.47 with diarization $0.12/hour diarization add-on ElevenLabs Scribe v2 Realtime $0.39 PAYG Not supported in real time AssemblyAI Universal-3.5 Pro Realtime $0.45 base / $0.57 with diarization $0.12/hour add-on; up to 10 speakers Gemini 3.5 Transcribe Live ~$0.54 blended Not supported in live mode Amazon Transcribe Streaming ~$0.60 in AWS's N. Virginia streaming example Included; up to 30 speakers OpenAI GPT Live Transcribe $1.02 Diarization not listed as a model capability The comparison is necessarily imperfect. Qwen's price varies by deployment geography; its international real-time rate of $0.00009 per second works out to about $0.324 per hour. Google's Gemini figure is an estimated blended token cost rather than a flat hourly tariff. AWS prices vary by region and usage tier. ElevenLabs lists $0.39 per hour on its API pricing page but advertises $0.28 per hour or lower on annual Business plans. Deepgram's pricing particularly illustrates why feature-level comparisons matter: its current Nova-3 Multilingual streaming rate is about $0.35 per hour, but speaker diarization costs another $0.002 per minute , bringing the comparable total to roughly $0.47 per hour. AssemblyAI similarly lists $0.45 per hour for Universal-3.5 Pro Realtime and another $0.12 per hour for streaming diarization. Cartesia is harder to normalize because Ink-2 is packaged through monthly credit plans rather than a simple metered PAYG hourly rate. Its $5 Pro plan includes roughly nine hours and 16 minutes of Ink-2 transcription, which works out to about $0.54 per transcription hour if every credit is consumed exclusively on STT. That should not be treated as equivalent to a standalone $0.54 hourly API tariff. Even with those caveats, Muse's positioning is clear. It is not the absolute cheapest streaming transcription service — Soniox currently publishes a lower equivalent rate — but $0.18 per hour with diarization included puts Meta toward the low end of the market, especially against providers that charge separately for speaker attribution. At 1,000 hours of processed audio, Meta's public rate implies roughly $180 in transcription charges. Meta also leads its launch accuracy benchmarks Price matters less if it comes with a large accuracy penalty. Meta's benchmark material argues the opposite. On the Artificial Analysis AA-WER Streaming Index supplied with the launch, Muse records a 3.1% final-transcription word error rate, ahead of Cartesia Ink-2 at 3.4%, ElevenLabs Scribe v2 Realtime at 3.6%, Qwen3 ASR Flash Realtime at 3.7%, GPT Live Transcribe and Grok Speech to Text Streaming at 3.9%, and Gemini 3.5 Transcribe Live and AssemblyAI U3.5 Realtime Pro at 4.0%. Meta points out that Muse took the number one spot on third-party independent AI benchmarking firm Artificial Analysis' streaming speech-to-text evaluation as of September 1. Meta published the following benchmark charts in its launch post . Its diarization result may be even more relevant to the product's positioning. Meta reports an average 17.5% diarization error rate across AMI-IHM, AMI-SDM and VoxConverse, lower than the competing systems shown in its chart. Speaker capacity and diarization error rate should not be conflated. A platform capable of representing 100 people is not automatically better at correctly attributing speech than one supporting 20, and Meta's benchmark does not test every competitor operating at its advertised maximum speaker count. There are deployment tradeoffs as well. Meta's API currently provides turn-level but not word-level timestamps, and it does not expose word-level confidence scores, sound-event detection or emotion detection. The documentation also specifies eight concurrent streams per tenant by default and real-time sessions of up to 60 minutes before an application must reconnect. Still, Muse's launch creates an unusually sharp price-performance proposition. Its 20-plus-speaker diarization does not establish a world record, but the record may be the less important metric. For enterprise developers, the larger question is whether a service can preserve speaker attribution, accurate text and usable turn boundaries while a complicated real-world conversation is still unfolding. At $0.18 per hour, with 20-plus-speaker diarization inside the same real-time model that currently leads Meta's supplied streaming accuracy benchmarks, Muse Voice Transcribe gives enterprise teams a serious new option for meeting intelligence, live transcription and voice-agent infrastructure — while putting additional pressure on competitors to compete on speaker-aware accuracy and total operating cost, not merely raw speech recognition.
- Google Gemini's new agent-based video analysis cuts token usage by up to 88 percent
Google is adding agent-based video analysis to Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite. Instead of scanning videos frame by frame at a fixed rate, the model decides on its own which segments to examine and at what resolution. Google says this cuts token usage by up to 88 percent while improving accuracy, especially for multi-hour footage. The article Google Gemini's new agent-based video analysis cuts token usage by up to 88 percent appeared first on The Decoder .
Score: 47🌐 MovesSep 2, 2026https://the-decoder.com/google-geminis-new-agent-based-video-analysis-cuts-token-usage-by-up-to-88-percent/ - Amazon’s AI assistant can now spot fake emails from the company
Alexa for Shopping can compare the message you received ‘against a record of every message Amazon has sent.’
Score: 47🌐 MovesSep 2, 2026https://www.theverge.com/tech/988518/amazon-alexa-for-shopping-verify-emails - Frontier AI research moves into cyber defense as attackers gain speed
Frontier artificial intelligence research is moving into security operations, and the arrival of a cyber superintelligence lab at one of the industry’s largest platform companies marks how far that shift has traveled. The question is no longer whether models can spot threats, but whether they can absorb a decade of human defender knowledge and act […] The post Frontier AI research moves into cyber defense as attackers gain speed appeared first on SiliconANGLE .
Score: 46🌐 MovesSep 2, 2026https://siliconangle.com/2026/09/02/cyber-superintelligence-lab-aims-tilt-ai-edge-defenders-falcon/ - AI is getting closer to being able to exploit OT, and that's very bad news for critical infrastructure
The situation is not disastrous just yet, but it's definitely time to start paying attention, Forescout hints.
- Nvidia boss presses G20 ministers to build AI infrastructure
Nvidia chief Jensen Huang urged G20 ministers on Wednesday to build more data centers to support the artificial intelligence revolution and warned that leaders who talk up the technology's dangers risk leaving their countries behind.
- The Republican Nominee for New York Governor Made a Creepy, AI-Generated Video of Mamdani and Hochul
The video makes Zohran Mamdani and Kathy Hochul look like they're hanging out in a prescription medication commercial.
- Meta says AI glasses fix stops users from secretly recording others
Meta upgrades its AI glasses' privacy features to block device users from filming people without their consent.
- How We Rebuilt Playbook Review as a Multi-Agent System
Explores redesigning playbook review using multi-agent architecture for improved legal workflow.
Score: 46🌐 MovesSep 2, 2026https://www.harvey.ai/en-US/blog/rebuilding-playbook-review-as-a-multi-agent-system - EU antitrust regulators quiz publishers on Google's AI search opt-out
The publishers' feedback could determine the outcome of an ongoing EU investigation that could result in yet another hefty fine for Google if the proposal fails to address competition concerns and publishers' worries about unfair use of their content.
- Anthropic Leader Praises President Trump’s Data Center Stance at G20 Meeting
Anthropic Leader Praises President Trump’s Data Center Stance at G20 Meeting theinformation.com
- Hugging Face's new duck robot is selling fast. A Chinese chip powers it
The colorful "Microduck" robot from HuggingFace's French subsidiary Pollen Robotics has sold more than 10,000 units since launching on Thursday.
Score: 46🌐 MovesSep 2, 2026https://www.cnbc.com/2026/09/01/hugging-faces-new-duck-robot-is-selling-fast-a-chinese-chip-powers-it.html - Google's September Android Drop brings Motion Assist and more Gemini tricks
Google's September Android Drop adds new Gemini features, Motion Assist, Guided vision, Google Keep integration in Messages and new chat customisation options
- How the Unacademy-upGrad deal unfolded; Why AI could push India’s ER&D revenue past $100B
How the Unacademy-upGrad deal unfolded; Why AI could push India’s ER&D revenue past $100B YourStory.com
Score: 45🌐 MovesSep 2, 2026https://yourstory.com/2026/09/unacademy-upgrad-deal-ai-could-push-india-erd-revenue - India could be a big winner in AI data centre boom, but power demand may be the catch
India is set to gain significantly from the global AI data centre expansion. Demand for local cloud infrastructure and sovereign AI is driving this growth. Significant investments are flowing into India's data centre development projects. However, power and grid infrastructure expansion pose major challenges. Addressing these will be crucial for India to become an AI hub.
- Spreadsheets still eat up 70% of treasury teams’ time as agentic AI promises sharper forecasts
Treasury teams spend much time on manual tasks, hindering strategic work. Spreadsheet use causes forecast variances, impacting liquidity management significantly. Agentic AI can improve cash forecast accuracy to ninety percent. Workflow automation and AI offer scope for transformation and efficiency gains. Companies need strong data foundations and governance for successful AI adoption.
- The EU is ready to bet billions on data centers. Not all countries want in.
An EU industrial policy plan to build AI infrastructure is forcing European capitals to free up investment — or else fall further behind.
- National data centre projects are consolidating America’s AI lead
Countries may host facilities but dispersion of hardware does not equal decentralisation of power
- Lutnick: Anthropic is "back on the right side" with Trump administration
CHAPEL HILL, N.C. — Commerce Secretary Howard Lutnick said the Trump administration now trusts Anthropic following months of clashes with the AI company. Why it matters: The statement marks a remarkable turnaround in a relationship strained by a bitter fight over national security and AI safeguards . What they're saying: Asked if he trusts Anthropic CEO Dario Amodei, Lutnick told Axios' Mike Allen: "We trust Anthropic." "They've done what we asked. They're back on the right side. So the answer is: Yes," Lutnick said in an interview on Tuesday. The big picture: The declaration comes as Anthropic co-founder Tom Brown takes a more prominent role in the company's relationship with the White House, including a headlining spot at this week's G20 Innovation Ministerial . "Really excited for our conversation, so I'd like to introduce you all to Tom Brown, one of the founders of Anthropic," Lutnick said on Wednesday to a room full of G20 ministers. Lutnick kicked off the discussion offering Brown an opportunity to introduce himself to an international audience: "I don't think they know you so well. Anthropic, of course, is famous, but a little story helps." What they're saying: Brown praised President Trump's pro-data center post on Truth Social this week. "I really love Trump's post from earlier this week... where he was pointing out that the data centers are just an enormous source of prosperity," Brown said. "They produce a ton of jobs. They produce taxes. Now the way that we design them, we actually bring on more power to the grid," he said. Brown also highlighted Anthropic's support for the administration's ratepayer pledge. Between the lines: Industry knows aligning with the Trump administration goes a long way for policy priorities. Catch up quick: Personality differences added fuel to the fire in a clash between the Trump administration and Anthropic earlier this year that resulted in sweeping export controls on the company's most advanced models. Brown was instrumental in patching up the relationship, holding multiple conversations with Lutnick and National Cyber Director Sean Cairncross. A federal judge in August, meanwhile, struck down the Pentagon's blacklisting of Anthropic as a supply-chain risk , ruling that the government's actions violated Anthropic's constitutional rights. The company is also fighting a separate Pentagon designation under a different statute in the D.C. Circuit. The bottom line: Brown offers a reset for a company once on the outs with the White House.
- Exclusive: Meta Tells Engineers AI Token Usage Won’t Be Part Of Performance Reviews
Exclusive: Meta Tells Engineers AI Token Usage Won’t Be Part Of Performance Reviews theinformation.com
- This Sneaky ‘Jedi Mind Trick’ Hack Can Hijack a Business’s AI Chatbot and Cost It Millions
Prompt injections can lead to devastating losses. Here’s how to protect yourself.
- The AI arms race is throwing fuel on the global bond sell-off and pushing yields even higher
The AI arms race is throwing fuel on the global bond sell-off and pushing yields even higher Business Insider
- UKG's CIO says HR tech is AI's next big bet. The software firm's employees have already launched 387 AI tools and over 12,000 agents
UKG's CIO says HR tech is AI's next big bet. The software firm's employees have already launched 387 AI tools and over 12,000 agents Fortune
- India’s richest man now wants to turn aging computers into AI-ready PCs
Jio is betting it can turn an aging computer into an AI-ready PC for as little as about $11 for two months.
Score: 45🌐 MovesSep 2, 2026https://techcrunch.com/2026/09/02/indias-richest-man-now-wants-to-turn-aging-computers-into-ai-ready-pcs/ - Snowflake spikes 22% on healthy results and AI coding momentum
Snowflake business is growing quickly, in part thanks to adoption of an artificial intelligence agent for writing code.
Score: 45🌐 MovesSep 2, 2026https://www.cnbc.com/2026/09/02/snowflake-snow-q2-earnings-report-2027.html - Fable 5.1 🧠, World Labs Atlas 🌍, Cognition $47B 💰
Fable 5.1 🧠, World Labs Atlas 🌍, Cognition $47B 💰
- Advertising is coming to AI. Regulators should prepare now
OpenAI’s move to bring ads to ChatGPT in India marks a shift in how AI companies could monetise free users. But chatbot advertising raises questions that traditional digital advertising does not. The post Advertising is coming to AI. Regulators should prepare now appeared first on MEDIANAMA .
Score: 45🌐 MovesSep 2, 2026https://www.medianama.com/2026/09/223-chatgpt-ads-india-ai-advertising-regulation/ - The UK must exploit narrowing window of opportunity to build resilient AI future
The UK faces a narrowing window of opportunity to strengthen its AI future, unless...
Score: 45🌐 MovesSep 2, 2026https://www.turing.ac.uk/news/uk-must-exploit-narrowing-window-opportunity-build-resilient-ai-future - How Caterpillar is using AI to augment human intelligence—and optimize its supply chain
Jamie Engstrom has a winning personal motto: “We reserve the right to work smarter tomorrow”—shorthand for her IT organization’s commitment to keep learning as AI advances and Caterpillar’s business evolves. As CIO and SVP, Engstrom leads a global IT team of more than 2,200 professionals across 28 countries. Though Caterpillar has used AI in some capacity for three decades, Engstrom now drives the company’s use of AI for things like advanced analytics, automation, and supply chain optimization. She approaches these key functions in a holistic way. “We think about a digital thread from our demand planning at the enterprise level, and supply planning through to manufacturing and operations, and being able to apply AI to that,” she says. “We’re seeing some really awesome early benefits,” she adds, citing improved quality on the factory floor, better labor planning leading to a reduction in variable labor costs, better management of material shortages, and more. Engstrom created Caterpillar’s IT AI Center of Excellence and partnered with Accenture and Snowflake to extend that work into manufacturing quality, financial reporting, and supply chain planning. “That is a major part of our strategic initiative and, frankly, where I spend a significant amount of my time,” she says. Under her direction, Caterpillar built an internal data assistant on a Snowflake-based enterprise data foundation, letting employees query trusted company data and get answers in seconds. In June 2026, that work earned Caterpillar Snowflake’s AI Innovator Award for the second consecutive year. Engstrom has championed an additive, rather than a reductive, approach to leveraging artificial intelligence. AI, she says, is best used for automating mundane tasks and augmenting human intelligence so that workers can do more valuable work. “I think creating the culture [of] safe experimentation, and the education of what the ‘art of the possible’ is, is where employees are going to embrace it more, versus feeling like AI is going to be a competition to their job,” she notes. Getting workers to buy into the idea of reimagining processes using AI is a question of trust and, Engstrom explains, it’s not just trust in the accuracy and relevance of AI models’ outputs. “Technology adoption is going to accelerate when our employees trust both the processes, the guardrails, the access, and the governance that we have in place—when they have confidence in the underpinning data to be able to know that the output that they’re getting is productive and meaningful for them.”