No tags match.
In one month, the enterprise software industry bought its way toward agents that act rather than advise. The controls to govern them are running behind.
The Model Context Protocol has shifted from a technical curiosity to a default question in enterprise software evaluation. Following CircleCI’s June 2026 MCP server release and Databricks’ move to govern MCP services in Unity Catalog, this article explains what MCP changes for integration cost, vendor lock-in, and AI governance, and sets out the practical procurement questions technology leaders should write into their evaluations.
At WWDC 2026, Apple unveiled a rebuilt Siri powered by a custom Google Gemini model reported to carry 1.2 trillion parameters, at a reported cost of roughly $1 billion a year. This article argues the deal is a template rather than a surrender: Apple rented the frontier model but kept the customer relationship, the data path, the context layer, and the interface. It sets out the four-part playbook enterprises should copy – task-level model bake-offs, gateway architecture that keeps providers swappable, owned small models for routing economics, and treating context as the proprietary asset – alongside the dependency and concentration risks of renting without an exit.
More than $40 billion has gone into enterprise AI, and 56% of CEOs say they have nothing to show for it – the gap has surprisingly little to do with the models
On 27 May 2026 the European Commission presented its Tech Sovereignty Package, including the Cloud and AI Development Act, restricting sensitive public-sector data from US hyperscalers. The restrictions stop at the public sector, but the underlying issue, namely US CLOUD Act jurisdiction over American-incorporated providers regardless of where data is stored, applies to every regulated business. This article explains the legal mechanism, the Dutch precedent blocking the Kyndryl and Solvinity acquisition, the trap of overcorrecting toward immature European providers, and the practical steps CIOs should take around data classification, portability, hybrid and multi-cloud design, and encryption key custody.
Google now processes 3.2 quadrillion tokens a month, up from 480 trillion a year ago. While vendors compete on per-token prices, the organisations pulling ahead on AI are the ones tracking a different metric entirely: tokens per business outcome. This article explains why per-token price is the wrong number to optimise, introduces tokens per resolved outcome as the unit that makes agentic AI economics legible, and sets out the three architectural patterns – task decomposition, context discipline, and evaluation-driven model selection – that create durable cost and performance advantages in agentic AI programmes.
A techUK report has confirmed what CIOs have felt in their cost reviews for two years: cloud-first strategies are hitting sovereignty and cost walls. This article examines the egress tax problem, the implications of the UK Data Use and Access Act 2025, and what a genuine data-first architecture – one where compute moves to data, not the other way around – actually requires organisations to do differently.
AI data centres are projected to consume more energy than Germany and France combined by 2030, yet most enterprise AI strategies are still built on the assumption that compute is cheap and limitless. This article examines the energy and infrastructure constraints that are closing the brute-force compute era, and sets out the practical architecture shifts – task decomposition, context discipline, evaluation-driven model selection – that organisations need to make now.
While nearly two-thirds of enterprises have experimented with AI agents, fewer than 10% have successfully scaled them, with 80% citing data limitations as the core obstacle. This article explores why the modern pipeline-centric data stack, designed for human analysts, struggles with autonomous AI systems, and examines the critical shift toward semantic metadata, upstream quality enforcement, and machine-readable context.
dbt Labs’ 2026 State of Analytics Engineering Report found trust in data has surged to 83% as the top organisational priority — the steepest single-year rise ever recorded — driven by 71% of data professionals concerned about incorrect AI outputs reaching stakeholders. With only 7% of enterprises saying their data is completely AI-ready and Gartner finding organisations that see returns invest four times more in data foundations, this article examines the acceleration-to-governance gap and what agent-ready data infrastructure actually requires.
UiPath’s 2026 report found 78% of executives believe operating model reinvention is essential for agentic AI. McKinsey found 80% are seeing no bottom-line impact. This article examines the structural, foundational, and governance barriers creating that gap — and what the organisations pulling ahead are doing differently.
With the launch of C3 Code and Anthropic’s Project Glasswing, Agentic AI is here. Learn why modular software architecture is a non-negotiable prerequisite to safely scale autonomous AI agents in the enterprise.
