Week In Review: Gartner research says half of agentic AI projects will be cancelled by 2027, Microsoft scientist claims regulation can be good for AI innovation, Pepsi teams up with Salesforce for agent deployment, and more.
Brands and Agencies
WPP Partners with Vercel for AI Expansion
WPP teamed up with cloud application company Vercel, making Vercel’s v0 and AI SDK technologies accessible to WPP employees and clients. The partnership will bring ‘vibe coding’ – a form of coding with natural language – to WPP partners. This will assist creative teams at WPP by allowing them to create digital experience design at the moment of inspiration – the models have advanced coding capabilities that allow for intricate modification of the designs. AI SDK serves as an open-source library of AI tools and agents. WPP will also be granted access to new Vercel product releases, which the company plans to integrate into WPP Open, its AI-powered marketing platform.
Verizon Integrates Google Gemini-Powered Assistant
Telecoms company Verizon announced the integration of AI assistants into its customer service experience. The firm is relaunching their My Verizon app with the addition of an AI assistant. The chatbot will assist with customer service issues, including answering billing questions, managing upgrades, and adding new lines. Customers will be redirected to a human agent if the AI assistant is unable to fulfill the request. The chatbot operates on Google Gemini technology. Verizon said that it worked with Google to ensure the assistant makes as few mistakes as possible. The telecoms company is now opening its customer service to 24-hour live support because of the AI assistants.
PepsiCo Incorporates Agentforce into Workflow
PepsiCo announced it will be deploying Agentforce, Salesforce’s agentic AI network, into its workflows. The drinks corporation plans to use Agentforce for a variety of work tasks, such as improving customer support and encouraging sales growth and engagement. PepsiCo will have access to Salesforce’s Data Cloud, Service Cloud, Consumer Goods Cloud, and Marketing Cloud. It will use these data and services to gather insights about consumer behaviour and expedite the customer service process. “AI is reshaping our business in ways that were once unimaginable,” said Ramon Laguarta, chairman and CEO of PepsiCo. “This collaboration with Salesforce is another step toward a more connected and adaptive PepsiCo – deploying AI to unlock smarter and faster decision-making, fuel innovation and power sustainable growth.”
Media
AI Training on Books OK, Says US Judge in Anthropic Case
AI company Anthropic won the ruling in a copyright lawsuit filed by book authors. The authors who filed the lawsuit include Andrea Bartz (We Were Never Here, The Last Ferry Out), Charles Graeber (The Good Nurse: A True Story of Medicine), and Kirk Wallace Johnson (The Feather Thief). All three accused Anthropic of using their work to train its AI model Claude without permission. The ruling in Anthropic’s favour indicates the use of books to train AI models is not a violation of copyright laws, particularly because of notable changes made in model outputs. Judge William Alsup explained that Anthropic’s utilisation of the published works was “exceedingly transformative,” therefore permitted by law. The judge, however, ruled that Anthropic would still have to stand trial for its use of pirated books, including a ‘central library’ of over 7 million titles.
Perplexity Calls the BBC “Manipulative & Opportunistic” in Copyright Row
In a letter, the BBC threatened AI company Perplexity with legal action if it continues to scrape BBC content to train its AI models without permission or payment. The broadcaster also asked the AI firm to delete copies of its materials and provide adequate compensation for content already used. In response, Perplexity described the BBC’s demands as “manipulative and opportunistic”, saying the broadcaster’s efforts are linked to Google’s domination. “The BBC’s claims are just one more part of the overwhelming evidence that the BBC will do anything to preserve Google’s illegal monopoly,” Perplexity said in a statement.
BBC Introduces Gen AI Tools for Journalists
The BBC announced it will let journalists create new “‘At a glance’ summaries” of longer news pieces using Gen AI, making news more accessible to audiences. The shorter, bullet-pointed story summary will appear in a box within selected articles. To generate the summary, BBC journalists give the AI a single approved prompt. The summary is continuously reviewed and edited before and after publication to ensure editorial standards are met. Bite-sized content, the BBC says, is increasingly favourable among news readers, especially younger audiences. The company will disclose when and where AI has been used to uphold its transparency principles.
Tech
OpenAI Faces IO Name Controversy
Earlier this year, OpenAI made headlines for its planned acquisition of IO, an AI startup founded by former Apple design exec Jony Ive. The news signaled a trajectory for the AI firm towards wearable AI products. However, this week, these plans were called into question after a lawsuit was filed against the giant from a hardware startup called Iyo, claiming name trademark infringement. OpenAI chief exec Sam Altman described the case as “silly, disappointing and wrong.” OpenAI removed the blog post announcing the acquisition from its website following the court order.
Regulation Can Boost AI, Says Microsoft Chief Scientist
Regulation if “done properly” can boost AI innovation, said Microsoft’s Chief Scientist Dr Eric Horvitz. These comments were in response to President Donald Trump’s proposal of a ten-year ban on AI regulations, saying that limiting laws restricting AI development could stunt growth in the industry. Trump’s sentiment is fuelled by growing AI competition between the US and China. Contrary to this idea, however, Horvitz, who served as tech advisor to former US president Joe Biden, said there needs to be better communication between scientists and governments around regulation. At a meeting of the Association for the Advancement of Artificial Intelligence, he said: “Guidance, regulation… reliability, controls, are part of advancing the field, making the field go faster, in many ways. We need to be very cautious about jargon and terms like regulation or bumper stickers that say no regulation because it’s going to slow us down. It can speed us up if done properly. We should be cautious and care and communicate to governments about that.”
Salesforce Introduces Agentforce 3
Salesforce released an upgrade to its agentic AI platform, Agentforce, to encourage enterprise AI adoption. Agentforce 3 features a command centre to manage AI agents; from monitoring agent interactions to tracking performance. The command centre also allows users to build and test agents. Agentforce 3 includes Model Context Protocol (MCP) support, allowing agents to access and operate enterprise tools. Salesforce expanded their AgentExchange, allowing customers to be connected to Agentforce through partners such as AWS, Google Cloud, IBM, and PayPal.
Meta Considered Acquisition of Runway AI
Meta reportedly had its sights on Gen AI company Runway AI before its investment into Scale AI. Discussions on acquiring the startup were held by Meta chief exec Mark Zuckerberg. The deal, however, was never finalised. Runway is a multimedia AI platform that is best known for its video creation tools. Instead, Meta invested $14bn in data annotation firm Scale AI, utilising the company’s technology to accelerate Meta’s AI capabilities. Meta also reportedly considered buying AI companies Perplexity AI and Safe Superintelligence.
BearingPoint Publishes Guide for AI Agent Use in Marketing
Technology consultancy firm BearingPoint published a report titled “The AI Sales & Marketing Revolution: a guide towards 2028.” The report’s purpose is to give companies a framework for utilising agentic AI in marketing and sales practices. It emphasises the need for a concrete strategy before integrating AI into the enterprise. The report also stresses the importance of scaling AI with direction and coordination – that all sectors of the enterprise are in agreement on their AI strategy. BearingPoint provides examples of AI use cases in sales and marketing, such as using Gen AI tools to produce marketing content.
Nearly Half of Agentic AI Projects Will Fail by 2027, Gartner Predicts
Gartner released a report stating that over 40 percent of agentic AI projects will be cancelled by 2027. The cancellations are attributed to rising costs, inadequate risk prevention, and unclear business value. The company explains that “agent washing” – rebranding AI products without actual agentic abilities – is becoming increasingly common. The research also estimates that only around 130 agentic AI vendors are authentic. “Most agentic AI projects right now are early stage experiments or proof of concepts that are mostly driven by hype and are often misapplied,” said Anushree Verma, Senior Director Analyst, Gartner. “This can blind organisations to the real cost and complexity of deploying AI agents at scale, stalling projects from moving into production. They need to cut through the hype to make careful, strategic decisions about where and how they apply this emerging technology.”
Google DeepMind Reveals Internet-Less Robots
Google DeepMind, the AI division of Google, revealed a new AI robot that can function without internet connection. Unlike typical robots which need access to a faraway server, DeepMind’s Gemini Robotics On-Device model, can operate anywhere on earth without access to the web. In a blog post, Google DeepMind’s robotics lead, Caroline Parada, said the robot is useful for “latency sensitive applications” and “environments with intermittent or zero connectivity.”
Apple Eyes Perplexity Acquisition
Apple is weighing up acquiring major AI player Perplexity, Bloomberg reported. Perplexity AI is currently at a $14 billion valuation, and would appear attractive to Apple, the one tech giant notably lagging in their AI developments. All the major tech companies, including Meta, Microsoft, Amazon, OpenAI and Google have sought to remain competitive in the midst of tense competition. Meta reportedly had its eyes on Perplexity earlier this year before investing $14 billion in AI annotation firm Scale AI. Other potential deals for Apple may include French AI startup Mistral, Sierra AI, Safe Superintelligence, or Thinking Machine Labs.
Number of the Week
$3.77 trillion. That’s the market cap of AI chip giant Nvidia. Share prices reached a record high this week, surpassing a previous record set in January. Shares closed on Wednesday at $154.31 after a 4.3 percent hike, increasing the tech giant’s market cap to $3.77 trillion – continuing its reign as the richest business in the world. The news comes after an uncertain start to the year, following US sanctions on AI chips going to China. This stopped Nvidia from shipping its own China-specific H20 chip, which was made in response to previous export restrictions. As a result, Nvidia projected an over $8 billion loss.
