Want Smarter Spights in your Inbox? Sign up for our weekly newsletters to get what items on business leaders, data, and security leaders. Subscribe now
Chatter around artificial general intelligence (through) can rule headings from Silicon Valley companies Openi,, Meta and XaiBut for land leaders in the soil, the focus of practical applications and measured results. Venturebeat recently Change 2025 The event of San Francisco, a clear picture has emerged: The weather of real, drilled agent AI is here, hastily and it resorts to businesses moving.
Firms like ChinaG and brass,, Capital,, LinkedIn,, Stanford University and Health Health AI agents are silent in AI production agents, facing concrete problems, and to see visible returns. Here are four biggest takeaways from the activity for technical workers.
1. The AI agents that transition to production, faster than anyone realized
Businesses now set up AI agents in the applications facing the customer, and the shape can easily accelerate a speed. A recent venturebeat survey of 2,000 industrial professionals held before vb change reveals that 68% of business companies (with 1,000+ employees) AI agent has been adopted – a number as high as time. .
However new data proves this fast transfer. art KPMG Survey released on June 26, one day after our activity, shown that 33% of organizations now send AI agentsA striking threefold increase from 11% only past two quarters. This market transfersifies a couple of weeks recognized in weeks ago in its pre-transformed survey.
This acceleration is set to results to see. Ashan willyCEO SA New relicfound a shocking 30% quarter on top of quarter growth To monitor AI applications through its customers, especially due to its customers transfer to adoption agents. Company activates AI agents to help customers automate workflows they need help. ChinaG and brassFor example, deployment of invoice agents and agents to remind software in QuickBooks. The result? Businesses with part Five-free-fashioned pays and 10% more willing to pay fully.
Even non-developers feel the transition. Scott Whitethe product leading the Antropic’s Claude Ai Product, describes how he is, yet not a professional programmers, is now the software prepared to build on building has its own. “It can’t be six months ago,” he explained, highlighting the power of tools like Claude Code. Similarly, Openia’s Product head for its API platform, Olivier GodementDetail how customers want drawing and cardboard THE Using SDK agents to strengthen multi-agent systems.
2. The Hyperscaller’s race is no obvious winner as multi-cloud, dominated multi-model
The days of betting on a large language model (LLM) trovider is over. A steady theme across 2025 is step toward a multi-cloud multi-model and strategy. The businesses want flexibility to choose the best tool for work, if it is a strong proprietary model or a good open open source.
As Armand RuizVP on AI platform to IBM explainedThe company’s development of a model gate – applications of applications to use whatever LLM is the most efficient and performer for the specific case – even a direct answer to customer’s needs. IBM started by offering business customers with own open-source models, then added open-source support, and finally knowingly need to support all models. This desire for flexibility echoes XD Huang, CTO in Zoom, describing his company Three years model method: Support proprietary models, offering their own good model fluently and allow customers to make their own good-tute versions.
This trend creates a strong but suppressed ecosystem, which GPUs and the power needed to create tokens are in limited supply. As Dylan Patel on Semiianalysis and fellow panelists Jonathan Ross on Gle and Sean lied on BRAINS deceivedIt puts the profit pressure to many companies simply buy a lot of tokens if it is available, instead of locking tokens continue to fall. Businesses moved about how they use different models for different tasks to optimize for NVIDIA costs – a vb transform session is also led by Sex around the Attach usual memory and savings solution for AI.
3. Businesses are intended to solve real problems, not pursuing through
While tech leaders like Elon Musk, Mark Zuckerberg and Sam Allman talked about the Dawn of Superintellence, business practices surrounded their arms and resolved business challenges. Conversations in change is easily based on reality.
arrested Health Health, The third largest integrated health insurance company and health manager. The main data officer Richard Clarke said this using LLMS for Protical Applications such as multilingual communications that are better service to their customer base, and streamlion to medical claims. In other words, the firm delivery technology is currently better today. Similarly, Capital is Tims of agents of agents who change the company’s functionsWith specific agents for tasks such as risk assessing and auditing, including helping their car marketing clients connect customers.
The journey industry has also seen a pragmatic shift. CTOs from Burden and Kayak mentioned how they adapt to new paradigs to search-moved llms. Users can search for a hotel with an “infinity pool” on chatgpt platforms, and travel platforms must be attached to the level of natural language language to remain competitive. Customer focuses, not the technology for one’s own good.
4. The future of AI teams is small, nimble, and empowered
AI agents age also changes how teams are structured. Agree is small, nimble “squads” in three to four engineers the most effective. Varun MohanCEO SA Windsurfinga rapidly growing agent agent, kicked the activity by dispute That this small team structure allows for easy-to-try product hypothes and avoid slowing strikes in larger groups.
This transition means “all is a builder,” and increasingly, “all is a manager” to AI agents. As Kamahir and Atassian Noticed, the engineers Now learn to handle the fleets of agents. The skills needed is change, with a greater emphasis on clear communication and strategic thinking to guide these autonomous systems.
This nimbleness is supported by a growing acceptance of sandboxed development. Andrew NgA leading voice of AI, advised attendees Leave safe, handling, and observe at the end of the development cycle. While it may not be counted for many businesses, the idea is to arouse rapidly changing within a controlled environment to quickly verify the value. This greeting shows our survey, which finds that 10% of AI adopt organizations do not have dedicated AI safety teamsuggests readiness to prioritize the speed at early stages.
Mag-uban, kini nga mga takeaway nagpintal sa usa ka tin-aw nga litrato sa usa ka talan-awon sa negosyo nga AI nga paspas, pagbalhin gikan sa halapad nga pag-eksperimento aron mapunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipunting nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipatuman nga gipunting, gipahamtang nga gipunting, gipahamtang nga gipunting, gipabug-atan nga gipunting. Changing conversations in 2025 show that companies move AI agents now, even if they need to know the hard lectures on the road. Many have been through one or two large pivots since the first trial of geneative ai one or two years ago – so good to start early.
For more conversation with these themes and further analysis from the incident, you can listen to the entire discussion I have in the independent AI developer in our recent podcast. We also uploaded key talks to VB Transform HERE. And our full range of articles from activity is HERE.
Listen to VB Transform Podcast Podcast with Matt Marshall and Sam Witteveeen here:
Editor’s note: As an appreciation to our readers, we open early bird registration for VB change in 2026 – $ 200 only. Here where AI ambition meets the reality of surgery, and you want to be in the room. Reserve your place now.
