Marketing teams are moving from experimenting with AI to organizing work around it. Agents now own defined jobs inside marketing operations, while search visibility is shifting from rankings to citations inside AI answers. Measurement lags adoption: 47.8% of marketing teams have replaced tasks with AI, yet only 6.5% measure ROI.

LinkedIn rewired how reach is earned this fortnight, splitting the feed, shrinking network distribution and flagging AI content, while charging for networking features that used to be free. Brands responded by moving the voice from company pages to employees and creators, a shift Microsoft formalised and Billion Dollar Boy Group backed with its first acquisition. For sellers, the common thread is that specificity, first hand proof and buyer problems now outperform volume and templates.

The edition marks a move from partner programs built around structure to partner programs built around business outcomes. Co-sell, marketplaces and AI are now judged on revenue, attribution and trust rather than on listings, tiers or tooling. Frontier AI vendors and hyperscalers are formalizing their partner networks, which raises the bar on operating discipline for everyone selling alongside them.

Agents now sit on top of the martech stack, and the vendors who own the data and permissions underneath them are repositioning around that fact. The constraint on value has moved from which agent to deploy to the quality of customer data, context and governance beneath it. In the same period, a new marketing automation entrant launched against the incumbents, and CDPs, MAPs and consultancies each had to defend their role.

GTM teams spent the period consolidating tools into agent ready stacks, with Clay, Salesforce and HubSpot competing to own the context layer beneath them. The GTM Engineer is now a defined role, though its scope, ownership and long term survival are still contested. Practitioners keep returning to buyer trust, problem led demand and clear ownership as constraints that tooling does not solve.

Field marketing is shifting from counting attendance to proving pipeline. Smaller, better targeted formats and interactive activations are replacing scale and logoed giveaways, while event tech stacks struggle to carry booth conversations through to revenue. Market data from Forrester, Norwest, and Trade Show Executive confirms that events remain a core channel, and that organizers still have limited proof of booth value to offer exhibitors.

Capital is concentrating in the largest managers while the rest of the market adapts. Mega funds across infrastructure, energy transition and buyout closed above target, yet fundraising for mid-sized managers is tightening as distributions lag and rates stay high. The response is structural: record co-investment volumes, evergreen vehicles, continuation funds and a sharper LP focus on cash returned over reported value.

Venture capital is concentrating faster than at any point this decade, with megafunds absorbing most LP capital while first time and emerging managers shrink in number. Distributions remain the binding constraint, as strong paper returns sit alongside near zero DPI for recent vintages and an exit queue that keeps growing. AI dominates both deal value and, increasingly, how funds themselves are run, while Europe and other regions confront structural gaps in speed, capital supply and early-stage pipeline.

Deal value has outpaced deal count in the UK and Germany is recovering selectively, while cybersecurity and Finland are setting activity records. Capital and strategic appetite are widely available, so the constraint has moved to execution: seller preparation, integration discipline and risk management now separate deals that close and create value from those that do not. AI has become both a rationale for acquisitions and a source of late stage deal risk.

The exit market is reopening unevenly. Capital is returning to a narrow group of managers and deals, while a large inventory of aging portfolio companies still waits for a buyer and holds back liquidity across the system. Sponsors, founders and finance leaders are responding by treating exit readiness as an operating discipline that starts years before a sale.

AI has moved professional services from experimentation to restructuring. EY's results show the scale of the shift: total revenue grew 4.7% while sales from AI services grew 49%. Across talent, pricing, firm strategy, research and SAP and Oracle delivery, judgment, trust and governed data now carry more of the value than the speed of producing a first draft.

Private equity value creation moved from financial engineering to operating execution across this period. Compressed exit multiples, longer holds and higher rates left EBITDA growth and cash conversion carrying returns, while AI adoption widened the gap between firms that deploy it and firms that can show results. Talent, ownership and execution capacity became the binding constraints on every plan.

The period brought two contradictory signals into the open at once. Capital markets priced humanoid robotics as an infrastructure category, with an IPO, a multibillion dollar compute pact and a wave of acquisitions all landing inside the same window. At the same time, operators inside the industry admitted that most of the hardware being built still cannot do paid work, that models remain too slow for real contact physics, and that Europe and healthcare are both racing to avoid being left out of a story currently being written in China and in defense

The construction technology market kept two contradictory stories running at once this period. Capital and product launches accelerated sharply, with AI agents moving from demos into live jobsite workflows and hundreds of millions of dollars flowing into robotics and AEC software. At the same time, the industry's own adoption data showed a widening gap between digital investment and delivered productivity, with skills, trust and clean field data emerging as the real constraints rather than the tools themselves.

The period's conversation split cleanly into two tracks. On one side, practitioners kept insisting that industrial AI's real constraints are governance, workforce readiness, and data foundations, not model capability. On the other, platform vendors including SAP, Siemens, Schneider Electric, and AWS pushed concrete product and partnership moves, many of them staged at a dense run of global trade events and forums.

Tools and equipment are becoming trackable, monitorable and networkable by default rather than by exception. Power tools are gaining wider cordless lineups and identity features built into the product itself, fixed assets are being tagged with RFID, QR and BLE so their history travels with them, machines are being read by cable free sensors and electrical signature analysis instead of manual rounds, and enterprise platforms are pulling all of that tool, asset and equipment data into a single visibility layer that IT, operations and safety teams now share.

Foundation models are moving from research demos to fleets, with a single video prompt now enough to teach a robot a new task and simulated training pipelines clearing eighty percent success on humanoid pick and place. Humanoid makers are racing to convert valuation into shipped units, with Chinese manufacturers currently ahead on deliveries while battery life and manufacturing consistency remain the binding constraints. Surgical robotics is consolidating around workflow and regulatory milestones rather than novelty, and capital keeps flowing into the sector at a record pace even as compute providers and industrial giants stake out the layers they intend to own.

Construction's AI conversation is splitting into two tracks that rarely meet. One track is strategic: leadership, data readiness and culture are named more often than the model itself as the real constraint on adoption. The other is operational: reality capture, BIM interoperability, generative scheduling and robotics keep compounding, each pushing more structured data into systems that AI increasingly depends on. Funding and M&A are following that second track, concentrating around whoever owns the workflow and the data behind it, while the career ladder for the people who build all of this keeps stretching further into strategy and leadership.

Quantum computing moved from proving that error correction works to proving how well it works, while the cryptographic clock for the systems it will eventually break kept running. In the same period, AI changed the pace of both attack and defense, and a critical PHP vulnerability showed how little time security teams now have between disclosure and exploitation. Across all of it, the pattern is a shift from announcements to measurable benchmarks, dated deadlines and operational accountability.

Enterprise AI has moved from experimentation to accountability. Regulation is becoming concrete, security is being rebuilt around agents that act rather than advise, and the main constraint on value is now the operating model and the quality of human judgment, not the technology. Vendors and infrastructure providers are positioning for an agentic market, while practitioners keep returning to the same message: broken processes and weak governance cap the return.

Sovereignty in cloud is moving from a compliance label to an engineering test of control and exit, and the same discipline is now being applied to AI cost and operations. Broadcom’s VMware decisions are pushing enterprises toward private and hybrid alternatives, while vendors compete to ship governed, agent-ready infrastructure for regulated and public sector buyers.

The conversation in green ICT has moved from asking how efficient a single model or facility is to asking how impact is measured, governed and decided across the whole system. Agent workflows, cooling design, grid connection and cloud procurement are now treated as places where emissions, water and cost are fixed early and are hard to reverse. Regulators and standards bodies are responding with labels, assessments and skills frameworks, while vendors package carbon and cost data into the tools engineers already use.

AI has made building cheap, and across the period the evidence from operators, surveys and conferences points to the same consequence: deciding what to build, and for whom, is now the scarce capability. Teams report faster shipping without faster customer value, SaaS vendors face rebuild pressure from their own customers, and product roles are being redrawn around judgment rather than execution.

Capital, contracts and political intent are now moving faster than the systems that turn them into fielded capability. Record European funding, contract ceilings of $500 million and privately financed billion-dollar factories coexist with unclear approval routes, small first orders and production bottlenecks. Threats are adapting on a shorter cycle than procurement, so the constraint has moved from invention to adoption and scale.