Add source Login
8 billion tokens a day forced AT&T to rethink AI orchestration — and cut costs by 90%

8 billion tokens a day forced AT&T to rethink AI orchestration — and cut costs by 90%

When your average daily token usage is 8 billion a day, you have a massive scale problem. This was the case at AT&T, and chief data officer Andy Markus and his team recognized that it simply wasn’t feasible (or economical) to push everything through large reasoning models. So, when building out an internal Ask AT&T personal assistant, they reconstructed the orchestration layer. The result: A multi-agent stack built on LangChain where large language model “super agents” direct smaller, underlying “worker” agents performing more concise, purpose-driven work. This flexible orchestration layer has dramatically improved latency, speed and response times, Markus told VentureBeat. Most notably, his team has seen up to 90% cost savings. “I believe the future of agentic AI is many, many, many small language models (SLMs),” he said. “We find small language models to be just about as accurate, if not as accurate, as a large language model on a given domain area.” Most recently, Markus and his team used this re-architected stack along with Microsoft Azure to build and deploy Ask AT&T Workflows, a graphical drag-and-drop agent builder for employees to automate tasks. The agents pull from a suite of proprietary AT&T tools that handle document processing, natural language-to-SQL conversion, and image analysis. “As the workflow is executed, it's AT&T’s data that's really driving the decisions,” Markus said. Rather than asking general questions, “we're asking questions of our data, and we bring our data to bear to make sure it focuses on our information as it makes decisions.” Still, a human always oversees the “chain reaction” of agents. All agent actions are logged, data is isolated throughout the process, and role-based access is enforced when agents pass workloads off to one another. “Things do happen autonomously, but the human on the loop still provides a check and balance of the entire process,” Markus said. Not overbuilding, using ‘interchangeable and selectable’ models AT&T doesn’t take a "build everything from scratch" mindset, Markus noted; it’s more relying on models that are “interchangeable and selectable” and “never rebuilding a commodity.” As functionality matures across the industry, they’ll deprecate homegrown tools in lieu of off the shelf options, he explained. “Because in this space, things change every week, if we're lucky, sometimes multiple times a week,” he said. “We need to be able to pilot, plug in and plug out different components.” They do “really rigorous” evaluations of available options as well as their own; for instance, their Ask Data with Relational Knowledge Graph has topped the Spider 2.0 text to SQL accuracy leaderboard, and other tools have scored highly on the BERT SQL benchmark. In the case of homegrown agentic tools, his team uses LangChain as a core framework, fine-tunes models with standard retrieval-augmented generation (RAG) and other in-house algorithms, and partners closely with Microsoft, using the tech giant’s search functionality for their vector store. Ultimately, though, it’s important not to just fuse agentic AI or other advanced tools into everything for the sake of it, Markus advised. “Sometimes we over complicate things,” he said. “Sometimes I've seen a solution over engineered.” Instead, builders should ask themselves whether a given tool actually needs to be agentic. This could include questions like: What accuracy level could be achieved if it was a simpler, single-turn generative solution? How could they break it down into smaller pieces where each piece could be delivered “way more accurately”?, as Markus put it. Accuracy, cost and tool responsiveness should be core principles. “Even as the solutions have gotten more complicated, those three pretty basic principles still give us a lot of direction,” he said. How 100,000 employees are actually using it Ask AT&T Workflows has been rolled out to 100,000-plus employees. More than half say they use it every day, and active adopters report productivity gains as high as 90%, Markus said. “We're looking at, are they using the system repeatedly? Because stickiness is a good indicator of success,” he said. The agent builder offers “two journeys” for employees. One is pro-code, where users can program Python behind the scenes, dictating rules for how agents should work. The other is no-code, featuring a drag-and-drop visual interface for a “pretty light user experience,” Markus said. Interestingly, even proficient users are gravitating toward the latter option. At a recent hackathon geared to a technical audience, participants were given a choice of both, and more than half chose low code. “This was a surprise to us, because these people were all very competent in the programming aspect,” Markus said. Employees are using agents across a variety of functions; for instance, a network engineer may build a series of them to address alerts and reconnect customers when they lose connectivity. In this scenario, one agent can correlate telemetry to identify the network issue and its location, pull change logs and check for known issues. Then, it can open a trouble ticket. Another agent could then come up with ways to solve the issue and even write new code to patch it. Once the problem is resolved, a third agent can then write up a summary with preventative measures for the future. “The [human] engineer would watch over all of it, making sure the agents are performing as expected and taking the right actions,” Markus said. AI-fueled coding is the future That same engineering discipline — breaking work into smaller, purpose-built pieces — is now reshaping how AT&T writes code itself, through what Markus calls "AI-fueled coding." He compared the process to RAG; devs use agile coding methods in an integrated development environment (IDE) along with “function-specific” build archetypes that dictates how code should interact. The output is not loose code; the code is “very close to production grade,” and could reach that quality in one turn. “We've all worked with vibe coding, where we have an agentic kind of code editor,” Markus noted. But AI-fueled coding “eliminates a lot of the back and forth iterations that you might see in vibe coding.” He sees this coding technique as “tangibly redefining” the software development cycle, ultimately shortening development timelines and increasing output of production-grade code. Non-technical teams can also get in on the action, using plain language prompts to build software prototypes. His team, for instance, has used the technique to build an internal curated data product in 20 minutes; without AI, building it would have taken six weeks. “We develop software with it, modify software with it, do data science with it, do data analytics with it, do data engineering with it,” Markus said. “So it's a game changer.”

Nvidia reports Q4 revenue up 73% YoY to $68.13B, vs. $66.21B est., Data Center revenue up 75% to $62.3B, and forecasts Q1 revenue above estimates (Ian King/Bloomberg)

Nvidia reports Q4 revenue up 73% YoY to $68.13B, vs. $66.21B est., Data Center revenue up 75% to $62.3B, and forecasts Q1 revenue above estimates (Ian King/Bloomberg)

Ian King / Bloomberg : Nvidia reports Q4 revenue up 73% YoY to $68.13B, vs. $66.21B est., Data Center revenue up 75% to $62.3B, and forecasts Q1 revenue above estimates —  Nvidia Corp., the world's most valuable company, gave another bullish quarterly revenue forecast, signaling that the massive build-out of AI computing remains on track.

Everything announced at Samsung Unpacked: The Galaxy S26 Ultra, Galaxy Buds 4 and more

Everything announced at Samsung Unpacked: The Galaxy S26 Ultra, Galaxy Buds 4 and more

Mobile World Congress is right around the corner, but Samsung got out ahead of many rivals that will be showing off new handsets at that event by running the latest edition of Unpacked on Wednesday. At its event in San Francisco’s Palace of Fine Arts, the company revealed the Galaxy S26 lineup, which includes the base S26, the S26+ and the S26 Ultra. We've got some hands-on time with all three handsets as well, and you can read about our in-person experience with the Galaxy S26 Ultra , as well as our S26 and S26+ impressions in those articles. In addition to those, Samsung announced the Galaxy Buds 4 along with (you guessed it) some AI updates. All the devices unveiled today are already available for pre-order , should you already be dying to get your hands on them. Here's a look at everything Samsung announced at the latest Unpacked: Galaxy S26 and S26+ Sam Rutherford for Engadget New-ish year, new Samsung phones. Let's deal with the out-and-out bad news first. The S26 and S26+ are each $100 more expensive than their predecessors (the RAM shortage isn't exactly helping to keep prices down). They start at $900 and $1,100 , respectively, for variants with 256GB of storage. Samsung has tweaked the design a bit this time by rounding the corners to align them more with the S26 Ultra's look. The base model has a slightly larger display than the S25 at 6.3 inches, though the S26+ still  has a 6.7-inch screen (albeit with a higher resolution than the S26 can handle). The S26 has a larger battery capacity than the S25 too at 4,300mAh. In North America, China and Japan, Samsung is sticking with Qualcomm chips rather than using its own Exynos 2600. If you pick up an S26 or S26+ in those markets, it will run on the Snapdragon 8 Elite Gen 5 chipset. The camera modules are the same as last year, but Samsung is aiming to supercharge them with upgrades elsewhere , such as ProScaler image upscaling and an MDNIe chip that's said to greatly improve color precision. There's also a video stabilization feature that tries to keep the horizon level while you're following a moving person or pet, which sounds useful for action shots. The new Object Aware Engine is said to better render skin tones and hair textures to make your selfies look better. Samsung has reworked some AI features too, such as making Now Brief and Auto Eraser compatible with more apps. Pre-orders for the S26 and S26+ are open today, and they'll be available on March 11. The phones will be available in purple, blue, black, white, silver and rose gold, though the latter two are online exclusives. Galaxy S26 Ultra The Galaxy S26 Ultra will be available in the same colorways and on the same date as its smaller siblings. It starts at $1,300, so there’s no price increase from the S25 Ultra. Preorders open today. The S26 Ultra has a 6.9-inch AMOLED display with a QHD+ resolution of 3120 x 1440 and a 120Hz refresh rate. That's all well and good, but the display is hiding (that being the key word) what's perhaps the Galaxy S26 Ultra's most interesting feature. The device has a Privacy Display that’s said to be the first of its kind on a smartphone. The idea here is to prevent people around from seeing what’s on the screen from acute angles. There's a small decrease in brightness when Privacy Display is active, and there are lots of customization options. You can set up Privacy Display to activate when you're asked for a password or PIN, or when you get a notification or open certain apps. So if (for instance) you tend to look at your banking apps when you’re on public transit and don’t want other passengers to see how much moolah you have, Privacy Display seems like a very handy feature. Elsewhere, the S26 Ultra runs on the same chipset as its smaller siblings. It comes with 12 or 16GB of RAM and 256GB, 512GB or 1TB of storage. The battery is larger than the ones in the other S26 models, as the Ultra has a 5,000 mAh capacity. There's support for Super Fast Charging 3.0 as well. Alas, Samsung still hasn't seen fit to offer built-in Qi2 charging magnets in the S26 lineup, which seems like a wild oversight in the year 2026. The selfie camera is the same as on the S26 and S26+. The S26 Ultra has 50MP ultrawide and 200MP wide lenses, along with dual 10MP 3x and 50MP 5x telephoto sensors. The resolutions of those cameras are the same as on the S25 Ultra, but the main 200M and 5x telephoto sensors now have wider apertures to let in more light. The S26 Ultra of course has the camera software features (and other AI features) found in the S26 and S26+. We'll have a review of the devices soon. In the meantime, head on through to our hands-on story for our initial impressions of the S26 Ultra . Galaxy Buds 4 and Buds 4 Pro Sam Rutherford for Engadget While the S26 phones are more iterative updates this year, Samsung has given its Galaxy Buds a proper refresh. It revamped the design and shape of the Galaxy Buds 4 and Galaxy Buds 4 Pro to do away with the angular look of the stems and remove the lights from them. The earbuds have a "more refined, computationally designed fit" too, according to Samsung. The company claims the latest earbuds have smaller earbud heads that allow for a better, more secure fit and a more "comfortable experience during all-day wear." The Galaxy Buds 4 remain in an open-fit format while the Buds Pro 4 have a canal-fit design. The latest earbuds are said to offer improved audio quality and active noise cancellation (ANC), with an ambient sound mode, adaptive EQ and adaptive ANC. On Buds 4 Pro, there's a siren detection feature that enables ambient sound to let you hear things like alarms or emergency vehicle warnings. The Buds 4 Pro have a wide woofer that increases the effective speaker area by nearly 20 percent compared with the previous gen earbuds, Samsung said. They support 24-bit/96kHz audio. If you're using Galaxy Buds 4 or Buds 4 Pro with a Galaxy device, you'll be able to use Bixby, Google Gemini and Perplexity with hands-free voice controls (though the "hey, Plex" command for the latter might be a tad confusing for folks who use a certain media server app). The Buds 4 Pro support head gesture controls for managing calls and Bixby interactions as well. As with the S26 phones, pre-orders for the earbuds open today and they'll hit shelves on March 11. The Galaxy Buds 4 cost $180 and Galaxy Buds 4 Pro will run you $250. Both models are available in white and black with a matte finish. There's an online-exclusive pink option for Buds 4 Pro as well. Android AI features Ahead of Unpacked, Samsung confirmed that it would offer Perplexity as an AI agent option in Galaxy AI on the S26 lineup. As part of that update, it shared that the S26 series would respond to the “Hey Plex” wake phrase, and that Perplexity’s features would also be embedded in the Samsung Browser app. The company also recently updated Bixby to make its own virtual assistant more conversational. On top of that news, Google had announcements of its own to make at Unpacked regarding new Android AI features , which will of course be available on S26 devices. On those handsets and the Pixel 10 lineup, the Gemini app will soon have a feature (in beta) that enables you to offload multi-step tasks, such as booking a ride or putting a grocery order together, to AI. It sure sounds like an attempt to build out agentic AI features on mobile devices. Launching soon as a beta feature in the Gemini app for #Pixel10 , Pixel 10 Pro, and Samsung Galaxy S26 series, you can offload multi-step tasks directly to Gemini. Simply long-press the power button and ask Gemini to help book you a ride home or reorder your last meal. Gemini… https://t.co/GjfXTnGg0k pic.twitter.com/YGIvqBkbu3 — Google Gemini (@GeminiApp) February 25, 2026 Starting this week on Pixel 10 devices (and soon on S26 phones), Circle to Search will offer the ability to find details about multiple objects at once , such as entire outfits instead of single pieces. Moreover, Gemini-powered, on-device Scam Detection for phone calls will be available for S26 devices in English in the US. Update, February 25 2026, 4:35PM ET: This story has been updated to include more details on the Perplexity AI integration, as well as include mentions in the intro of our hands-on and pre-order articles. This article originally appeared on Engadget at https://www.engadget.com/mobile/everything-announced-at-samsung-unpacked-the-galaxy-s26-ultra-galaxy-buds-4-and-more-180000530.html?src=rss

Salesforce reports Q4 revenue up 12% YoY to $11.2B, vs. $11.18B est., forecasts Q1 revenue above estimates, and announces a $50B share repurchase program (Zaheer Kachwala/Reuters)

Salesforce reports Q4 revenue up 12% YoY to $11.2B, vs. $11.18B est., forecasts Q1 revenue above estimates, and announces a $50B share repurchase program (Zaheer Kachwala/Reuters)

Zaheer Kachwala / Reuters : Salesforce reports Q4 revenue up 12% YoY to $11.2B, vs. $11.18B est., forecasts Q1 revenue above estimates, and announces a $50B share repurchase program —  Salesforce (CRM.N) forecast first-quarter revenue above Wall Street expectations on Wednesday, betting on strong demand …