This Week in AI in Marketing & Management (15th June 26)
The week opened with Anthropic releasing its most powerful model yet and ended with the US government ordering it switched off for every non-American user. That story alone captures the tension running through this week’s news: AI is advancing faster than the governance structures around it, faster than enterprise adoption can absorb, and faster than most marketing and management teams have prepared for. Alongside that, Visa and Mastercard both placed major bets on AI agents as the next payments layer, 90% of UK businesses were found to be invisible in AI-generated search results, and Google confirmed that AI Mode ads are no longer a future prospect, they are being tested right now. Here is everything worth knowing from the last seven days.
Table of Contents
- AI News, Tech & Tools
- AI in Marketing
- AI in Management
- AI in E-commerce, Retail and Agentic Commerce
- AI for Other Sectors and Industries
- Key Takeaways
- Frequently Asked Questions
- Conclusion
AI News, Tech & Tools
Anthropic launches Claude Fable 5, then the US government switches it off
Anthropic launched Claude Fable 5 and Claude Mythos 5 on 9 June. Fable 5 is described as a Mythos-class model made safe for general use, priced at $10 per million input tokens and $50 per million output tokens, less than half the price of Claude Mythos Preview. Stripe reported during early testing that Fable 5 “compressed months of engineering into days”, completing a codebase-wide migration of a 50-million-line Ruby codebase in a single day that would otherwise have taken a full team more than two months. Mythos 5 is a fuller version with cybersecurity safeguards relaxed, initially available only to government and selected infrastructure partners through Project Glasswing.
Three days later, the US government ordered Anthropic to suspend access to both models for all foreign nationals, citing national security concerns. The directive arrived at 5:21pm on Friday 13 June. Anthropic said it received only partial information and disagreed that software used by hundreds of millions of people should be blocked on these grounds. Time reported that the decision signals the US government increasingly views frontier AI as a national security asset rather than a commercial product.
Why it matters
This is not simply a story about one model being pulled. It signals that export controls are now a live instrument for AI governance, and that any organisation running critical workflows on frontier models from US-based labs faces a new category of operational risk. If you are advising clients on AI infrastructure, the question of where a model is hosted and who controls access to it has moved from theoretical to urgent.
OpenAI files for IPO and plans ChatGPT superapp
Reuters reported that OpenAI is planning its largest ChatGPT overhaul to date, aiming to convert the product into a “superapp” with coding tools, AI agents, and integrated services, ahead of a public listing. Campaign Live confirmed OpenAI has formally filed its IPO bid and is continuing to expand ChatGPT Ads, adding grouped ad formats and moving into new markets.
Why it matters
A publicly listed OpenAI will face quarterly earnings pressure for the first time. That changes the incentive structure around product decisions, pricing, and the pace of ads expansion. Marketers should watch the ads rollout closely. ChatGPT’s advertising inventory is not yet available at scale in the UK, but the direction of travel is clear.
Google launches Gemini 3.5 Live Translate
Google announced Gemini 3.5 Live Translate, a near real-time speech-to-speech translation model covering more than 70 languages. The model is designed for natural conversational translation without the lag of traditional interpretation tools.
Why it matters
Real-time multilingual AI communication removes one of the last significant barriers to international business interaction. For agencies and clients with multilingual audiences or global customer service operations, this is worth testing now.
Agentic AI in enterprise: process first, agents second
A Pegasystems study of 500 business and IT decision-makers who have successfully introduced agentic AI found a consistent pattern: they started by reimagining business processes, not by deploying agents into existing workflows. More than half (53%) had changed their processes to a significant extent. 95% had a specific corporate-level strategy in place before deployment. 71% said their top pre-deployment objective was to make complex processes work consistently and predictably across systems. IBM CEO Arvind Krishna made the same argument at Think 2026: “the enterprises pulling ahead are not deploying more AI, they are redesigning how their workforce operates.”
Why it matters
The Pega finding that 75% of organisations cite “lack of knowledge of benefits” as the biggest barrier to agentic AI success points to a consulting and education opportunity. Organisations that try to automate broken processes simply produce broken automation faster.
EU AI Act rewritten: high-risk rules delayed to 2027 and 2028
The Parliament Magazine reported that EU trilogue negotiations finalised on 7 May 2026 have pushed implementation of high-risk AI Act rules back to December 2027, with AI integrated into regulated products delayed until August 2028. Industrial applications, including automotive and semiconductor manufacturing, have been carved out entirely. Crucially, the general-purpose AI model rules covering systems like ChatGPT and Claude remain on schedule for August 2026. The shift in tone from Commission President von der Leyen was notable: in 2023 she described the Act as transposing “European values”; in 2026 she described the same law as creating “a simple, innovation-friendly environment.”
Why it matters
For UK businesses that trade with or operate in the EU, the August 2026 GPAI rules remain live. High-risk system compliance now has more runway, but GPAI obligations do not. If you are using or building on top of foundation models, August is the relevant date.
AI in Marketing
90% of UK businesses are invisible in AI search
A study by LLM Listed, reported by Decision Marketing, analysed 500 UK business websites across e-commerce, financial services, software, healthcare, professional services and legal services. Researchers tested whether companies appeared in responses from ChatGPT, Claude, Gemini and Perplexity when users asked commercially relevant questions within their industry. The results: 91% showed no evidence of an AI visibility strategy. Only 8% appeared consistently across all four major AI platforms. 76% were outranked by competitors in AI-generated responses. 15% had content specifically designed to support AI discovery and citation.
Founder Ben Harper said: “Companies have spent the past 20 years optimising for search engines. They now need to optimise for answer engines. We are seeing businesses with strong SEO performance that are virtually absent from AI-generated recommendations.”
Why it matters
This is the UK-specific data point that makes the GEO conversation impossible to ignore. 91% no strategy, combined with traffic declines of more than 50% reported in some analyses when AI-generated answers appear above traditional results. The businesses that act on this now will take citation share while their competitors are still debating whether it matters.
LinkedIn ranks second for AI citation share behind YouTube
Research from Meltwater found that LinkedIn is the second most-cited source by AI models across major B2B categories, behind only YouTube. LinkedIn ranked in the top five cited domains across AI and data science, marketing and advertising, leadership and strategy, sales and revenue, consulting and professional services, financial services and fintech, HR and talent, and technology and SaaS.
Meltwater CPO Chris Hackney said: “In an AI-first world, the job of a brand is to be the answer. LLMs are now the first stop for decisions that used to take hours of research, and if your brand is not being cited, you are not in the consideration set.”
Why it matters
For B2B marketers, this reframes LinkedIn from a distribution channel to a citation channel. Consistent, expert-led, decision-oriented content on LinkedIn is now part of your GEO strategy, not separate from it. The Meltwater finding also reinforces the IDC “dark funnel” analysis published this week: B2B buyers are forming views and shortlisting suppliers inside AI tools before any brand interaction takes place. If you are not in the AI answer, you are not in the consideration set.
Google Marketing Live 2026: ads arrive in AI Mode
Browser Media’s GML 2026 summary confirmed that Google is now testing ads inside AI Mode, with three new formats announced. Conversational Discovery Ads use Gemini to generate ad creative in real time in response to the specific phrasing of a search query, so an ad for a fragrance brand responds to the mood and use-case described in the search, not just a keyword. Highlighted Answers place eligible ads directly within AI Mode recommendation lists. Business Agent for Leads replaces static lead forms with a Gemini-powered chat interface embedded in the ad, allowing prospective customers to ask questions and get answers drawn from the advertiser’s website.
Google’s own research states that 75% of shoppers say AI Mode helps them make faster, more confident purchasing decisions. Healthcare ads in AI Mode are already in live testing in the US. Meanwhile, Search Engine Journal confirmed that Google has extended the Dynamic Search Ads to AI Max migration deadline to February 2027, giving advertisers more time to transition. And Search Engine Land reported that Google has added structured asset experiments to Performance Max, so advertisers can now test individual asset combinations with proper controls rather than relying on Google’s automated optimisation alone.
Why it matters
The arrival of conversational ad formats changes the nature of paid search creative. Static headlines optimised for a keyword match are giving way to generative responses optimised for a conversation. Advertisers who invest in clean, structured product data and well-organised asset libraries now will give Gemini more to work with when these formats go broad.
LinkedIn reach metric and Instagram post-Reel ads
LinkedIn has rolled out a new reach metric that splits impressions into in-network and out-of-network, showing creators and marketers how far content travels beyond their direct connections. Combined with the Meltwater AI citation research above, this gives B2B marketers a clearer picture of both distribution and discoverability. Separately, Instagram has opened its post-Reel advertising format to global advertisers via Campaign Manager, the unit that appears in the organic feed between Reels, now accessible outside the US.
Why it matters
The LinkedIn out-of-network metric matters precisely because of the AI citation story: if your content is reaching beyond your network, it has a better chance of being indexed and cited by AI models. Instagram’s post-Reel format is worth testing for any brand running video creative. It sits in a high-attention placement and is now fully accessible to UK advertisers.
AI in Management
Leaders want ROI from AI but most are still stuck at proof of concept
An Avanade study of US business and government leaders found that 57% expect up to 4x ROI from AI copilots and agents, with most anticipating returns within 12 months. Yet 41% remain stuck at proof of concept, and only 30% have developed what the research describes as a “visionary AI strategy.” 75% are implementing AI in isolated functions rather than as part of a cohesive framework.
The Financial Times put it plainly this week: “Routine time savings do not automatically make organisations function better, and staff have to clean up a lot of slop.” And CIO.com drew the comparison with the cloud migration era: we are obsessing over advanced models while ignoring the messy human workflows underneath, the same mistake made when organisations rushed to cloud without redesigning the processes that ran on it.
Why it matters
The pattern here is consistent across three separate studies this week. Organisations that treat AI as a technology project rather than an operational redesign project get stuck. The PoC trap is not a technology problem, it is a strategy and change management problem. The businesses that break through it are the ones that define the outcome first and work backwards to the tool.
AI is eliminating entry-level roles, and creating a leadership crisis within five years
The World Economic Forum published a stark warning this week: AI is systematically removing the entry-level roles that have always trained the next generation of managers. Harvard research cited in the piece indicates junior employment has fallen 9%, with entry-level hiring dipping 80% per quarter since 2023 at organisations that have adopted generative AI. ZipRecruiter’s 2026 Graduate Report found the share of entry-level jobs dropped to 38.6% at the start of 2026, down from over 44% three years ago.
The argument is not simply that jobs are being lost. It is that the structured developmental pathway, the junior analyst who builds the model manually, the new hire whose memo gets marked up, is being automated away. Entry-level roles have always been learning environments. Without them, the pipeline of experienced managers empties within five years.
Why it matters
This is a significant strategic risk that most organisations have not yet put on the board agenda. The efficiency gains from AI-driven automation of entry-level work are real and immediate. The cost, a hollowed leadership pipeline, arrives five years later. Every organisation that is automating junior roles without redesigning the learning pathway is quietly borrowing against its future management capability.
CIOs and CTOs are accountable for AI systems they cannot control
An IBM Institute for Business Value study of 2,000 C-level technology executives found that two-thirds of CIOs and CTOs are accountable for AI systems they do not fully control. 70% say teams across the business are deploying AI faster than IT can track. Only 11% feel fully ready for the scale of AI agent deployment expected within the next year. By 2027, surveyed executives anticipate a 38% increase in the number of AI agents deployed. 77% say AI adoption is already outpacing current governance capabilities.
The risk is not theoretical: surveyed organisations experienced an average of 54 AI agent incidents last year requiring human correction. 17% were high-severity events taking more than four hours to contain. 37% of those resulted in data exposure or security breaches. IBM found that organisations which embed control directly into their AI systems deploy 16x more AI agents, deliver 18% higher operating margins, and spend 4x less of their AI budget than those relying on manual governance.
Why it matters
The 70% “teams deploying faster than IT can track” finding is the one that matters most practically. This is the shadow AI problem at enterprise scale. Organisations that wait for governance to catch up before acting are already behind. The IBM data suggests that building control into the architecture from the outset, rather than retrofitting governance after deployment, is the single biggest differentiator between organisations that scale AI effectively and those that create expensive liability.
AI upskilling is not delivering ROI, but the gap is fixable
BCG published research this week showing that more than 60% of organisations report little or no ROI from AI upskilling programmes. The firms that do see returns spend approximately 2x more on structured skill-building than the average, and, critically, they connect training directly to specific business outcomes rather than running generic AI literacy programmes. BCG frames the question for every CEO as: are we building skills against a clear performance gap, or are we training for its own sake?
Why it matters
AI training spend is rising fast across every sector. The BCG finding that 60%+ see no return is not an argument against training, it is an argument for designing training differently. The organisations seeing returns are the ones that start with the business problem, identify the capability gap, and build the programme around closing that gap.
AI in E-commerce, Retail and Agentic Commerce
Visa and Mastercard both place bets on AI agent payments, and the rails are diverging
Visa and OpenAI announced a formal partnership this week to enable secure AI-agent-initiated payments across OpenAI’s platform. Visa is providing its global network, tokenised credentials, real-time authorisation, and fraud monitoring. Transactions operate within user-defined permissions: spending limits, merchant category restrictions, and required approval thresholds. Within the same week, Mastercard launched Agent Pay for Machines, a platform for autonomous AI-driven machine-to-machine commerce, built with Coinbase and Ripple and supporting crypto payment rails as an alternative to traditional card networks.
Forbes characterised the situation clearly: AI agents are settling on two rails, Visa and Mastercard’s tokenised card infrastructure, or Coinbase’s x402 stablecoin protocol. Both camps are building as fast as they can.
Why it matters
The payment rail chosen for agentic commerce will shape the commercial internet for the next decade in the same way that mobile payment infrastructure shaped it in the last one. For e-commerce operators, the near-term question is whether to build agent-compatible checkout now or wait for a standard to emerge. The Visa/OpenAI partnership points to tokenised card credentials as the path of least resistance for mainstream retail. The Mastercard/Coinbase route matters more if your customers are in markets where card infrastructure is weaker.
Nearly a third of shoppers would let AI make purchases within five years
DHL’s E-Commerce Trends Report 2026, based on 29,000 consumers and 5,800 businesses across 29 countries, found that 29% of shoppers would be willing to let AI make purchasing decisions on their behalf within five years. Among Gen Z the figure is 33%; among millennials, 36%. DHL eCommerce CEO Pablo Ciano said AI is “redefining the advantage at hyperspeed.”
The report also found that 62% of consumers would abandon a purchase if their preferred payment method were unavailable, yet only 45% of businesses currently recognise this as a major cause of basket abandonment. eCommerce Times reported that as AI agents begin shopping on behalf of consumers, marketplaces are confronting new questions about trust, liability and control that existing terms of service were not designed to handle.
Why it matters
29% willingness to delegate purchasing to AI within five years, among a population that currently does almost none of that, is a significant structural shift in the making. Retailers that invest now in machine-readable product data, structured inventory feeds, and agent-compatible checkout flows will be in a better position when that 29% becomes purchasing behaviour rather than survey intent.
Amazon reveals what the new Alexa has that Rufus did not
AdWeek spoke to an Amazon executive about the new Alexa for Shopping, which combines the Rufus AI assistant with Alexa+ for a more capable agentic shopping experience. Unlike Rufus, which was built for discovery and research, the new Alexa is designed to complete multi-step purchase journeys, including comparison, selection, and checkout. Amazon is also rolling out AI-generated custom merchandise design through Alexa for Shopping, allowing customers to create on-demand printed products via text prompts.
Why it matters
Amazon is building a shopping agent, not just a product search tool. For brands that sell on Amazon, the question is no longer only “can customers find my product?” but “will the agent recommend my product when a customer asks for the best option in my category?” The factors that influence agent recommendation are different from the factors that influence organic ranking.
Stripe at London: AI-driven global markets, UK focus
At Stripe Tour London, the payments company unveiled a suite of features specifically designed to help UK businesses operate in AI-driven global markets, covering multi-currency settlement, agent-compatible payment flows, and AI-assisted fraud management. The event positioned Stripe as infrastructure for the agentic commerce era rather than simply a payment gateway.
Why it matters
UK e-commerce operators now have a clear signal from Stripe that the platform is being built for agent-initiated transactions. For any UK business planning to support AI shopping agents, either as a buyer-side tool or as a merchant, Stripe’s agent-compatible infrastructure is worth reviewing now.
AI for Other Sectors and Industries
PUBLIC SECTOR: London Tech Week and GOV.UK Chat
Prime Minister Keir Starmer’s speech at London Tech Week opened with the announcement that the former Unilever soap factory in Warrington is being converted into an AI data centre, one of several conversion projects announced across Warrington, Lanarkshire, Liverpool and Leeds. The government confirmed that UK startups have raised close to half of all European tech investment this year. The speech framed Britain as the third-largest technology economy in the world and committed to continued investment in the Global Talent Taskforce, pension fund capital deployment, and trade deals designed to open international markets.
The Guardian’s analysis welcomed the investment ambition but raised questions about the speed of the chips and data centre spending, noting that planning reform and grid connectivity remain bottlenecks even where capital is available.
Meanwhile, the government launched GOV.UK Chat, an AI chatbot available inside the GOV.UK app that draws on official HMRC guidance to answer tax queries. It is described as the first official government AI tool of its kind in the UK.
Why it matters
The UK government is no longer simply signalling AI investment intent, it is converting factories, commissioning data centres, and deploying citizen-facing AI services. For businesses that work with the public sector or rely on HMRC guidance, GOV.UK Chat is worth understanding both as a tool and as a signal of where AI in public services is heading.
HEALTHCARE: MHRA publishes landmark report on AI in healthcare
The MHRA published two reports on 11 June drawing on the National Commission into the Regulation of AI in Healthcare. The Call for Evidence received 760 formal submissions from patients, healthcare professionals, industry partners and civil society organisations. The overarching finding: broad support for AI in healthcare, provided regulation sets clear standards for safety, efficacy and accountability. Key themes included the need for ongoing monitoring of technologies after deployment, the importance of human oversight, and strong public expectations of transparency. The Commission’s formal recommendations are due later this summer.
Why it matters
The MHRA process will shape the regulatory framework for AI medical devices and clinical decision-support tools in the UK. For any technology company or healthcare organisation with AI products in development, the Commission’s recommendations, due this summer, will be the most significant UK regulatory signal of 2026 in this space.
ENERGY: How AI is transforming the UK offshore wind industry
EDP24 reported that AI is reshaping the UK offshore wind sector, moving it from a maintenance-heavy manual inspection regime to a highly connected predictive maintenance model. AI-driven monitoring of turbine components, weather pattern analysis, and autonomous inspection via drones are reducing downtime and extending asset life. The piece profiles several North Sea operators already deploying AI at scale for condition monitoring and fault prediction.
Why it matters
Offshore wind is one of the UK’s most strategically important infrastructure sectors. The shift to AI-driven predictive maintenance is not a future aspiration, it is already operational. For any business with a stake in energy, engineering, or industrial IoT, the offshore wind sector is a useful case study in what mature AI deployment looks like at industrial scale.
Key Takeaways
- The US government’s export control order on Anthropic’s Fable 5 and Mythos 5 is the most significant signal yet that frontier AI models are being treated as national security infrastructure, not commercial software.
- 91% of UK businesses have no AI visibility strategy. If your brand is not being cited by AI models, you are not in the consideration set for a growing share of buyers.
- LinkedIn is the second most-cited source by AI models across B2B categories. Publishing expert, decision-oriented content on LinkedIn is now part of your GEO strategy.
- Google’s conversational ad formats in AI Mode are in live testing now. Advertisers need clean, structured asset libraries to make the most of Gemini-generated creative.
- Visa and Mastercard are both building AI agent payment rails on diverging technical standards. The infrastructure decisions being made now will define how agentic commerce works for the next decade.
- 29% of consumers globally would delegate purchasing to AI within five years. Retailers need agent-compatible product data and checkout flows before that intent becomes behaviour.
- 41% of enterprise leaders are stuck at AI proof of concept despite expecting 4x ROI. The bottleneck is strategy and change management, not technology.
- The AI leadership pipeline crisis is five years away. Organisations automating entry-level roles without redesigning the learning pathway are borrowing against future management capability.
- 70% of CIOs say teams are deploying AI faster than IT can track. Governance embedded in architecture outperforms manual governance by every measurable metric.
- AI upskilling returns a 60%+ failure rate unless programmes are built around specific business outcomes. Training for its own sake does not move performance.
Frequently Asked Questions
What should UK businesses do about the 91% AI visibility gap?
Start with a baseline reading. Ask the four major AI models (ChatGPT, Claude, Gemini, Perplexity) the questions your customers would ask before buying from you, and see whether your brand appears in the answer. If it does not, the integrated SEO and GEO approach is the right framing: keep the technical SEO and content hygiene that already works for traditional search, then add the GEO-specific layer of citation building, brand-mention monitoring, structured FAQs and authoritative named-author content that AI models will quote back. Talk to the Anicca team if you would like a combined audit of where you sit today on both fronts.
How should I prepare for AI agent payments and the Visa/Mastercard split?
For most UK retailers, the practical answer right now is to make sure your checkout, product data and inventory feeds are clean and machine-readable, and to keep watching the standards conversation rather than committing to one rail. Visa’s tokenised card credential approach is the path of least resistance for mainstream UK retail. Mastercard’s Agent Pay for Machines and the Coinbase x402 stablecoin route matter more for cross-border, B2B and machine-to-machine commerce. Build the foundations now (structured product data, clear authorisation logic, agent-compatible APIs), and you will be ready to plug into whichever rail your buyer base settles on.
How do I make the AI ROI case to my board when most companies are stuck at proof of concept?
Reframe the conversation away from headline cost savings and towards capacity, speed and quality outcomes. Identify three measurable business outcomes (campaign briefing time, content production volume at a given quality bar, paid media response time, lead-quality scoring accuracy) and measure them before and after each deployment. Connect AI training spend to those same outcomes. The BCG finding is unambiguous: organisations that get returns from training spend it on closing a defined capability gap, not on generic AI literacy.
Conclusion
This was the week that AI moved squarely into the strategic register. The US government switching off Anthropic’s most powerful model for foreign users is not a normal tech-press story, it is an export-control story dressed up as a product story. The 91% UK AI visibility gap is not a marketing trend, it is a structural commercial risk. Visa and Mastercard placing major bets on diverging agentic payment rails sets up the infrastructure decisions that will define the commercial internet for the next decade. And the consistent message across the BCG, IBM, Avanade and World Economic Forum research is that the AI returns are real, but only for organisations that redesign their operating model rather than layer AI on top of broken processes.
Three actions for UK marketing leaders this quarter. First, audit your AI visibility across the four major models and decide whether you are content to be in the 91% or willing to do the work to be in the 8%. Second, treat the conversational ad formats in Google AI Mode as a forcing function on your asset library and product feed quality, not as a creative novelty. Third, put the leadership-pipeline question on your board agenda before the entry-level redesign decisions get made by default.
Need help adapting your AI marketing strategy? Contact the Anicca team for expert guidance.
Join the Thursday AI Club
A hands-on AI club for marketers and managers. Fortnightly 3-hour sessions: Hour 1 is an open Q&A on any AI question; Hours 2-3 split into a workshop track (for newbies) and an advanced track (live demos and deeper questions).
Led by Ann Stanley, Darren Wynn, James Allen.
Membership: £40/month or £400/year
This roundup is compiled from publicly available sources using AI-assisted research. While we review every article for accuracy, our analysis reflects our interpretation of the original reporting. We strongly encourage readers to click through to the original sources linked throughout this post for full context and detail. If you spot anything that needs correcting, please let us know.
