26 de junio de 2026 · [[El Abismo de Máquina/Ecos|¿qué es un eco?]] # Eco: Inteligencia + Confianza, la tesis de Microsoft para acertar con la IA > [!entradilla] > Judson Althoff (Microsoft) resume su tesis para adoptar IA con cabeza, Inteligencia + Confianza, y cómo se traduce en diversidad de modelos, una capa que gobierna costes y agentes, y un IQ propio que compone desde dentro.  > [!tip]- Por qué me interesa > > Lo firma Judson Althoff, el CEO de la parte comercial de Microsoft, y es la versión ordenada de la conversación que tiene con cada cliente que se plantea adoptar IA. Su tesis cabe en dos palabras: Inteligencia + Confianza. > > Por debajo hay tres preocupaciones que se repiten: si la IA va a amplificar la inteligencia de mi empresa o a aprovecharse de ella, si puedo confiar en un retorno que aguante y respete mi gobierno y mi seguridad, y cómo controlo los costes. La respuesta de Microsoft: no dependas de un solo modelo, monta tu propio IQ sobre los datos de la empresa, y pon encima una capa que observe y gobierne agentes y gasto (Agent 365). > > Me llama por dos motivos. Enlaza con el Eco de Satya Nadella sobre la frontera sin ecosistema, que Althoff cita, y deja ver hacia dónde empuja Microsoft la adopción, con el foco movido del modelo al sistema que lo gobierna. Es la mirada del vendedor, conviene leerlo sabiéndolo, pero los ejes que plantea son los correctos. > > El artículo original de Judson Althoff: [Achieving success with AI](https://blogs.microsoft.com/blog/2026/06/16/achieving-success-with-ai/) > [!abstract]- Resumen esquemático > > #### Tesis > > - Los dos elementos esenciales de cualquier solución de IA: Inteligencia + Confianza. > - La IA pensada para crecer debe amplificar la inteligencia propia de la organización y hacer que esta se componga desde dentro, en vez de ceder ese valor a unos pocos modelos. > > #### Tres preguntas que repiten los clientes > > - ¿La IA amplificará mi inteligencia y lo que hace única a mi empresa, o la usará en su propio beneficio aprendiendo de mis flujos y mi propiedad intelectual? > - ¿Puedo confiar en un retorno duradero y en que todo corra dentro de mi gobierno y mi seguridad? > - ¿Cómo consigo visibilidad, control y flexibilidad para gestionar los costes y maximizar el valor? > > #### Palancas para gestionar el coste a escala > > - Diversidad de modelos: no depender de un único modelo ni de su harness; asignar a cada tarea el modelo adecuado (por ejemplo GPT-5.5 o Claude Opus 4.8) según su fuerza y su economía. > - Tu IQ (Microsoft IQ): convertir datos en bruto en inteligencia usable y dar a los agentes el contexto por delante en vez de reconstruirlo; resultado medible en velocidad, precisión y menos tokens. > - FinOps: gestionar el gasto de IA como capacidad central de la empresa, sobre todo al pasar de precios fijos a modelos por uso. > > #### Modelos de negocio > > - Licencia por usuario (USL): paquete de capacidades a un precio fijo por usuario y mes; es la base. > - Licencia por uso: para agentes que trabajan largo y en muchos pasos, donde el coste se alinea con el trabajo hecho. > - Microsoft 365 Copilot y GitHub Copilot combinan ambos; Copilot Cowork, ya en disponibilidad general, requiere la USL y luego va por uso. > > #### Agent 365: el plano de control > > - Un único sitio para observar, gobernar, gestionar y asegurar los agentes, sean de Microsoft, de otro proveedor o propios. > - Se apoya en el stack que las empresas ya usan: Entra (identidad), Defender (amenazas), Purview (gobierno del dato), Intune (endpoints), y se extiende a la gestión de costes. > - Idea de fondo: gestionar el trabajo humano y el de los agentes como un solo sistema, con visibilidad de rendimiento y de coste. # Artículo original: Achieving success with AI Fuente: [Achieving success with AI](https://blogs.microsoft.com/blog/2026/06/16/achieving-success-with-ai/) The two most important elements in any AI solution are Intelligence + Trust. I first made this statement in November at our Ignite conference and my conviction is strengthened by every conversation I have with customers. Through my travels, three consistent topics are being raised when considering the adoption of AI solutions: 1. Will AI amplify the intelligence of my organization and the attributes that make my company unique within its industry to grow my business; or will it use my intelligence for its own benefit, learning from my most important business flows and leveraging my intellectual property? 2. Can I trust that the outcomes are providing durable return on investment and that these solutions are running within the confines of my governance and security standards? 3. How do I get the visibility, control, flexibility and business model innovation needed to manage the costs associated with AI and maximize value? I consistently advise customers that they need to build their own IQ on a platform of intelligence that is model-diverse, open and heterogeneous at every layer of the stack. Models are commoditizing. No company should be dependent upon any one model or any one model’s harness. Over the weekend, Satya warned of [a world where every company across every sector is ceding value to a few models that eat everything they see](https://snscratchpad.com/posts/frontier-ecosystem/). AI that is intended for growth should amplify the intelligence of an organization so that it compounds from within. Companies also need an observability platform that provides governance, management, security and Financial Operations (FinOps) to ensure the ROI with AI. This enables AI to be trusted within the environment over which it reasons and puts the business in control of the outcomes. Intelligence + Trust is embedded across Microsoft 365 Copilot, GitHub Copilot and Copilot Studio, where model diversity aligns cost and performance to each task. Microsoft IQ optimizes workflows, so context is routed efficiently and reduces unnecessary compute. Agent 365 is the control plane to observe, govern, manage and secure agents. We have built a system to manage AI spend as a core enterprise capability, not an afterthought. It is delivered across clouds and model providers without locking customers into a single approach. ### Managing costs at scale As agent usage scales, organizations need a clear set of levers to manage cost: **Model diversity.** Any given inferencing model, model harness or agentic loop on its own does not help build out an organization’s IQ in ways that compound its intelligence. Both Microsoft 365 Copilot and GitHub Copilot are model-diverse by design without locking customers into a single provider. Different models — like GPT-5.5 or Claude Opus 4.8 — serve distinct roles with different economics. Matching the right intelligence to each task optimizes performance and cost. **Your IQ.** Agents struggle with raw data. Significant compute is spent interpreting structure and context before useful work begins. The [Microsoft IQ platform](https://www.microsoft.com/en-us/ai/microsoft-iq?msockid=3f1a5e10529f619810e74df553fd60a1) empowers your IQ by turning raw data into usable intelligence, continuously building a semantic understanding of how your organization operates across Microsoft 365 and line-of-business systems. It provides agents with the context they need upfront rather than requiring them to reconstruct it. The result is [measurable](https://www.microsoft.com/en-us/microsoft-365/blog/2026/06/02/announcing-the-new-work-iq-apis/?msockid=3f1a5e10529f619810e74df553fd60a1): faster execution, higher accuracy and lower token usage. This is how intelligence compounds within your organization. **Financial operations**. FinOps became critical when companies moved to the cloud and requires even greater attention as AI shifts from fixed pricing to usage-driven models. With Foundry and Agent 365, we are providing tools to help our customers optimize their AI costs today. ### Frontier business models Business models are evolving as we use AI to drive business outcomes. The User Subscription License (USL) has become the foundation, providing a package of capabilities for a predictable per-user-per-month fee. Usage-based licensing has emerged for long-running, multi-tasking agents, where cost aligns directly to the work performed. Microsoft gives customers a unique combination of business model flexibility and integrated product experiences that is unmatched in the market. Microsoft 365 Copilot and GitHub Copilot use both models — a USL offering with not only value and capabilities, but flexible consumption on top. Today we’re announcing the [general availability of Copilot Cowork worldwide](https://www.microsoft.com/en-us/microsoft-365/blog/2026/06/16/copilot-cowork-is-now-generally-available), which requires the Microsoft 365 Copilot USL and is then usage-based. Our model-diverse strategy allows customers to purchase capacity with the flexibility to use the right model for the job based on model strengths, economics and the latest innovations. Microsoft **Agent Factory** provides a single consumption model spanning Microsoft 365 Copilot (including Cowork), GitHub Copilot and agents built in Fabric, Foundry and Copilot Studio. Our integrated product experiences put AI in the flow of work for both knowledge workers and software developers and manage capacity fluidly across the two. Historically these personas have been distinct, but increasingly the line between them is blurring. Coding is becoming a mainstream knowledge worker skill and chat and Cowork are becoming modalities important for software development. With Microsoft 365 and GitHub, we offer market-leading tools for both roles and make it easy to seamlessly manage capacity based on availability and need. ### Agent 365: The control plane As organizations adopt agents from Microsoft, another provider or build their own, a control plane is essential. Agent 365 gives IT and security leaders a single place to observe, govern, manage and secure agents across the organization. It builds on the Microsoft stack that enterprises trust: Entra for identity, Defender for threat protection, Purview for data governance and Intune for endpoint management. We are extending Agent 365 to include cost management, so organizations can monitor and manage agent spend alongside security and compliance. As the Frontier Firm operating model takes hold, leaders will manage human and agentic work as a single system, with visibility into both performance and cost. — The two most essential elements in any AI solution are Intelligence + Trust. At Microsoft, this conviction shapes how we design every layer of our AI platform. Microsoft IQ enables organizations to harness their own unique IQ, bringing context to data and embedding AI directly into the flow of work to deliver faster, more accurate and more trusted outcomes while safeguarding assets and protecting intellectual property. Agent 365 provides that trust layer, ensuring every agent and AI artifact is observed across the environment so organizations can move decisively from experimentation to enterprise impact with confidence. As Jay Parikh put it at Build, [AI alone will not change your business. The system running it will](https://blogs.microsoft.com/blog/2026/06/02/ai-alone-wont-change-your-business-the-system-running-it-will). We have built this system for our customers and partners, where intelligence compounds from within and every agent operates with control, visibility and trust. Together, we can scale human ambition and define how AI delivers measurable business impact across every role, organization and industry. *Judson Althoff is the chief executive officer of the commercial business at Microsoft. He is responsible for the product strategy, sales, services, support, marketing, operations and revenue growth of the company’s commercial business, which operates in more than 120 regional and national subsidiaries globa