Cosentino is preparing to use Microsoft Discovery in its R&D operation to screen possible formulations for architectural surfaces before committing them to physical trials, positioning the Almería-based manufacturer as the first Spanish industrial company announced as a user of Microsoft’s scientific AI platform. The immediate takeaway for IT leaders is more restrained than the headline: this is a real deployment direction for enterprise AI beyond office copilots, but it is not yet a demonstrated materials-discovery result.
Microsoft’s account of the partnership describes Discovery as the next phase of a much longer Cosentino technology program spanning Microsoft 365, Azure, Power BI, plant sensors, IoT, security, Fabric, Copilot and GitHub Copilot. Reporting from La Ecuación Digital and Silicon, both following a June visit to Cosentino’s Cantoria facilities, adds the detail missing from the latest announcement: at the time of that visit, Cosentino had not begun a Microsoft Discovery pilot. It had a planned use case and a substantial historical data estate, not a verified new material or measured reduction in laboratory work.
That distinction is important because the company and Microsoft are presenting an ambition to reduce experiments, lead times and R&D costs. None of the material published so far establishes how many physical trials Discovery will eliminate at Cosentino, whether a candidate formulation can scale reliably from lab to production line, or when a Discovery-assisted material will reach the market.
Microsoft introduced Discovery at Build 2025 as an enterprise platform for scientific and engineering research. Its core proposition combines AI agents, graph-based knowledge retrieval, simulation, high-performance computing and proprietary enterprise data. The agents can search research records, organize evidence, generate hypotheses, propose candidates and analyze simulated results.
For Cosentino, the relevant data is unusually specialized. The company manufactures engineered surfaces for architecture and design, including Silestone and Dekton, where outcomes depend on raw-material composition, particle size, resins, pigments, pressure, temperature, curing or sintering conditions, surface treatment and visual appearance. A formulation that looks promising in a small laboratory sample must still perform predictably when manufactured as a large commercial slab.
La Ecuación Digital reports that Cosentino’s R&D organization holds more than four million records related to raw-material characterization, alongside formulations, historical tests, patents and reports. That is the asset Microsoft Discovery is being brought in to exploit: not a general-purpose chatbot with access to a few documents, but a system intended to reason across a large body of proprietary scientific and industrial history.
The practical value is likely to lie in narrowing the search space. If the platform can identify prior experiments that rule out an ingredient combination, surface a relationship between raw-material properties and a known defect, or suggest candidates worth simulating, it can help researchers decide where to spend laboratory time. That is different from proving a new material can be produced at industrial scale.
Cosentino’s own manufacturing reality makes that final step non-negotiable. A surface can meet a target color, strength or stain-resistance result in a controlled sample and still fail when production variables change across a continuous industrial process. Microsoft Discovery can help decide which experiments should happen; it does not remove the requirement to validate durability, repeatability, safety, manufacturability and cost.
There is also a timing issue in the language. Microsoft Discovery was initially announced in May 2025, and Microsoft said it reached general availability on June 2, 2026. Cosentino’s public announcement comes after the platform became generally available, but the company appears to be at the start of practical adoption rather than reporting a mature production implementation.
Microsoft has publicized other Discovery users in sectors where scientific and engineering iteration is central, including BHP, Syensqo and GSK. Those reference cases reinforce that Cosentino is joining a broader push to sell agentic AI into R&D, not receiving a bespoke technology built only for the Spanish manufacturer.
What makes Cosentino useful as a case study is its industrial setting. The company’s research connects directly to materials that must be manufactured in high volumes, shipped globally, specified by architects and fabricators, and backed by product-quality expectations. That creates a much tougher benchmark than using AI to summarize literature or generate early-stage ideas.
The public record does not yet provide the numbers that would make the claim operationally meaningful: model accuracy, simulation fidelity, the number of physical tests avoided, research-cycle reduction, cloud-computing consumption, or the cost of the Discovery environment. Until those figures emerge, the “first Spanish industry” headline says more about early adoption than about business impact.
The agent is meant to retrieve CRM and project data, prepare for meetings, draft messages and proposals, check schedules, track opportunities and help unblock commercial proposals affected by inventory or logistics issues. In theory, it saves the user from opening Teams, email, CRM, scheduling tools and other systems separately.
This is a familiar enterprise-agent pattern: the user sees one conversational interface, while the agent must act as a governed layer over systems of record. The benefit is convenience, but the risk is concentrated in permissions and execution. An agent that can view customer information, draft communications and create a revised proposal needs strict identity controls, source-level authorization and an auditable record of what data it retrieved and which actions it initiated.
Cosentino says it has embedded security by design and uses Microsoft’s security platform to protect data, identities and operations. That is the right direction, but the announcement does not disclose CLAR’s permission model, whether it can take irreversible actions without confirmation, which systems it can write to, or how the company evaluates incorrect recommendations. Those are the questions that determine whether an enterprise agent remains a productivity layer or becomes a new route for data exposure and operational mistakes.
The company says Microsoft 365 Copilot, Copilot Chat and GitHub Copilot have also been deployed, and attributes a 36% time saving in its commercial area to these productivity initiatives. That number is a Cosentino claim, not an independently published productivity study, and the company has not supplied the baseline, measurement period, sample size or calculation method. It should therefore not be treated as evidence that similar deployments will produce the same gain elsewhere.
Those underlying systems matter more to the eventual usefulness of CLAR and Discovery than an AI agent’s interface does. A materials-research platform cannot reliably recommend a formulation if historical test results are fragmented, poorly labeled or missing contextual details. A commercial agent cannot safely reconcile inventory, customer commitments and logistics if the authoritative systems disagree or access rules are inconsistently applied.
Cosentino says its SAP-to-Azure work improved critical times by up to 97.29%. As presented, that is another vendor and customer metric without an explanation of which processes were measured or what “critical times” means. The figure may describe a legitimate operational improvement, but it cannot be assessed from the available disclosure.
The strategic pattern is clear nonetheless. Cosentino is not treating Discovery as an isolated experiment bolted onto an old research archive. It is attempting to combine cloud migration, data governance, process mining, employee copilots, bespoke agents and R&D simulation into one operating model. Microsoft calls organizations following that model “Frontier Firms,” but that label is Microsoft marketing rather than an external certification.
The milestone worth watching is not another description of AI agents or a declaration that the company is moving toward an “Autonomous Enterprise.” It is a controlled account of a real R&D program: the baseline development cycle, the candidate space considered, the number of experiments actually avoided, the performance of the resulting material, and whether the formulation survived scale-up into production.
Until then, Cosentino has made a credible technology commitment and assembled the data, cloud and workflow pieces that such a program requires. It has not yet shown that Microsoft Discovery created a new material, shortened a material-development cycle or reduced physical experimentation in its factories.
That distinction is important because the company and Microsoft are presenting an ambition to reduce experiments, lead times and R&D costs. None of the material published so far establishes how many physical trials Discovery will eliminate at Cosentino, whether a candidate formulation can scale reliably from lab to production line, or when a Discovery-assisted material will reach the market.
Microsoft Discovery enters the part of R&D where the hard work remains physical
Microsoft introduced Discovery at Build 2025 as an enterprise platform for scientific and engineering research. Its core proposition combines AI agents, graph-based knowledge retrieval, simulation, high-performance computing and proprietary enterprise data. The agents can search research records, organize evidence, generate hypotheses, propose candidates and analyze simulated results.For Cosentino, the relevant data is unusually specialized. The company manufactures engineered surfaces for architecture and design, including Silestone and Dekton, where outcomes depend on raw-material composition, particle size, resins, pigments, pressure, temperature, curing or sintering conditions, surface treatment and visual appearance. A formulation that looks promising in a small laboratory sample must still perform predictably when manufactured as a large commercial slab.
La Ecuación Digital reports that Cosentino’s R&D organization holds more than four million records related to raw-material characterization, alongside formulations, historical tests, patents and reports. That is the asset Microsoft Discovery is being brought in to exploit: not a general-purpose chatbot with access to a few documents, but a system intended to reason across a large body of proprietary scientific and industrial history.
The practical value is likely to lie in narrowing the search space. If the platform can identify prior experiments that rule out an ingredient combination, surface a relationship between raw-material properties and a known defect, or suggest candidates worth simulating, it can help researchers decide where to spend laboratory time. That is different from proving a new material can be produced at industrial scale.
Cosentino’s own manufacturing reality makes that final step non-negotiable. A surface can meet a target color, strength or stain-resistance result in a controlled sample and still fail when production variables change across a continuous industrial process. Microsoft Discovery can help decide which experiments should happen; it does not remove the requirement to validate durability, repeatability, safety, manufacturability and cost.
The “first” claim is a market-positioning claim, not a performance metric
The statement that Cosentino is the first industrial company in Spain to incorporate Microsoft Discovery has been repeated by Microsoft-aligned coverage and Spanish technology outlets. There is no public registry of Discovery customers in Spain that would independently establish a nationwide first, so it should be read as a claim about Microsoft’s known deployment pipeline rather than an independently certified industry ranking.There is also a timing issue in the language. Microsoft Discovery was initially announced in May 2025, and Microsoft said it reached general availability on June 2, 2026. Cosentino’s public announcement comes after the platform became generally available, but the company appears to be at the start of practical adoption rather than reporting a mature production implementation.
Microsoft has publicized other Discovery users in sectors where scientific and engineering iteration is central, including BHP, Syensqo and GSK. Those reference cases reinforce that Cosentino is joining a broader push to sell agentic AI into R&D, not receiving a bespoke technology built only for the Spanish manufacturer.
What makes Cosentino useful as a case study is its industrial setting. The company’s research connects directly to materials that must be manufactured in high volumes, shipped globally, specified by architects and fabricators, and backed by product-quality expectations. That creates a much tougher benchmark than using AI to summarize literature or generate early-stage ideas.
The public record does not yet provide the numbers that would make the claim operationally meaningful: model accuracy, simulation fidelity, the number of physical tests avoided, research-cycle reduction, cloud-computing consumption, or the cost of the Discovery environment. Until those figures emerge, the “first Spanish industry” headline says more about early adoption than about business impact.
CLAR is the deployment with clearer near-term operational consequences
Cosentino’s more mature AI initiative is CLAR, an internal commercial-management agent built on Microsoft technology. According to CIO and the company’s descriptions, CLAR is designed for use from laptops, tablets and phones, including voice interaction while sales staff are traveling.The agent is meant to retrieve CRM and project data, prepare for meetings, draft messages and proposals, check schedules, track opportunities and help unblock commercial proposals affected by inventory or logistics issues. In theory, it saves the user from opening Teams, email, CRM, scheduling tools and other systems separately.
This is a familiar enterprise-agent pattern: the user sees one conversational interface, while the agent must act as a governed layer over systems of record. The benefit is convenience, but the risk is concentrated in permissions and execution. An agent that can view customer information, draft communications and create a revised proposal needs strict identity controls, source-level authorization and an auditable record of what data it retrieved and which actions it initiated.
Cosentino says it has embedded security by design and uses Microsoft’s security platform to protect data, identities and operations. That is the right direction, but the announcement does not disclose CLAR’s permission model, whether it can take irreversible actions without confirmation, which systems it can write to, or how the company evaluates incorrect recommendations. Those are the questions that determine whether an enterprise agent remains a productivity layer or becomes a new route for data exposure and operational mistakes.
The company says Microsoft 365 Copilot, Copilot Chat and GitHub Copilot have also been deployed, and attributes a 36% time saving in its commercial area to these productivity initiatives. That number is a Cosentino claim, not an independently published productivity study, and the company has not supplied the baseline, measurement period, sample size or calculation method. It should therefore not be treated as evidence that similar deployments will produce the same gain elsewhere.
Cloud, Fabric and SAP are the less glamorous prerequisites
The broader Cosentino-Microsoft relationship predates the current AI push by years. Microsoft documented Cosentino’s use of Microsoft 365, Teams, OneDrive, SharePoint, Planner, OneNote and Power BI in 2020, alongside industrial IoT work to connect factory operations. The newer program includes SAP migration to Azure, Microsoft Fabric as a data foundation, Celonis for process analysis and contact-center modernization through Azure Communication Services and Dynamics 365 Contact Center.Those underlying systems matter more to the eventual usefulness of CLAR and Discovery than an AI agent’s interface does. A materials-research platform cannot reliably recommend a formulation if historical test results are fragmented, poorly labeled or missing contextual details. A commercial agent cannot safely reconcile inventory, customer commitments and logistics if the authoritative systems disagree or access rules are inconsistently applied.
Cosentino says its SAP-to-Azure work improved critical times by up to 97.29%. As presented, that is another vendor and customer metric without an explanation of which processes were measured or what “critical times” means. The figure may describe a legitimate operational improvement, but it cannot be assessed from the available disclosure.
The strategic pattern is clear nonetheless. Cosentino is not treating Discovery as an isolated experiment bolted onto an old research archive. It is attempting to combine cloud migration, data governance, process mining, employee copilots, bespoke agents and R&D simulation into one operating model. Microsoft calls organizations following that model “Frontier Firms,” but that label is Microsoft marketing rather than an external certification.
The first useful proof will be a product-development benchmark
Cosentino’s Discovery announcement is significant because it takes enterprise AI into formulation work, where bad recommendations consume laboratory capacity, waste expensive inputs and can fail much later during manufacturing validation. It is also early enough that the company has not yet supplied evidence that the platform changes those outcomes.The milestone worth watching is not another description of AI agents or a declaration that the company is moving toward an “Autonomous Enterprise.” It is a controlled account of a real R&D program: the baseline development cycle, the candidate space considered, the number of experiments actually avoided, the performance of the resulting material, and whether the formulation survived scale-up into production.
Until then, Cosentino has made a credible technology commitment and assembled the data, cloud and workflow pieces that such a program requires. It has not yet shown that Microsoft Discovery created a new material, shortened a material-development cycle or reduced physical experimentation in its factories.
References
- Primary source: Demócrata
Published: 2026-08-05T13:03:19+00:00
Loading…
www.democrata.es - Related coverage: laecuaciondigital.com
Loading…
www.laecuaciondigital.com - Related coverage: silicon.es
Loading…
www.silicon.es - Related coverage: learn.microsoft.com
Loading…
learn.microsoft.com - Related coverage: directortic.es
Loading…
directortic.es - Related coverage: learn.microsoft.com
Loading…
learn.microsoft.com - Related coverage: cio.com
Loading…
www.cio.com - Related coverage: news.microsoft.com
Loading…
news.microsoft.com - Related coverage: laecuaciondigital.com
Loading…
www.laecuaciondigital.com - Related coverage: cosentino.com
Loading…
www.cosentino.com - Related coverage: cosentino.com
Loading…
www.cosentino.com - Related coverage: devblogs.microsoft.com
Loading…
devblogs.microsoft.com - Related coverage: content.cosentino.com
Loading…
content.cosentino.com - Related coverage: spainisin.com
Loading…
www.spainisin.com - Related coverage: adoption.microsoft.com
- Related coverage: static.cosentino.com
Loading…
static.cosentino.com - Related coverage: azure.microsoft.com
Transforming R&D with agentic AI: Introducing Microsoft Discovery | Microsoft Azure Blog
Learn more about the new enterprise agentic platform, , accelerating research and development
azure.microsoft.com
- Related coverage: blogs.microsoft.com
Powering America's Genesis Mission: Microsoft's commitment to scientific discovery - The Official Microsoft Blog
Today, we’re excited to share a long-term commitment to the Department of Energy’s (DOE) Genesis Mission, backed by a $60 million investment designed to accelerate AI for science and the breakthroughs it can deliver for the country. This deepened commitment includes Microsoft’s new Scientific...blogs.microsoft.com - Related coverage: blogs.microsoft.com
Microsoft Build 2026: Be yourself at work - The Official Microsoft Blog
Platforms shift when developers build. We explore, choose tools, dream, create. This platform shift comes with more information than ever, ready at your fingertips. This shift, it’s about building fast AND THEN: it’s about building, operating, optimizing and observing. Securing your...blogs.microsoft.com - Related coverage: news.microsoft.com
Loading…
news.microsoft.com - Related coverage: microsoft.com
Loading…
www.microsoft.com - Related coverage: techcommunity.microsoft.com
Loading…
techcommunity.microsoft.com - Related coverage: techcommunity.microsoft.com
Loading…
techcommunity.microsoft.com - Related coverage: cdn-dynmedia-1.microsoft.com
Microsoft Digital Defense Report 2025 – Governments and Policymakers Executive Summary
A surreal, abstract blue structure made of evenly spaced, translucent blades or fins, forming a dynamic spiral or wave shape, glowing with vibrant blue neon light. The composition is minimal, with smooth gradients and a dark blue background for high contrast. The scene has a futuristic and...cdn-dynmedia-1.microsoft.com
- Related coverage: msthesource.thesourcemediaassets.com
Loading…
msthesource.thesourcemediaassets.com - Related coverage: tomsguide.com
Microsoft Build 2025: All the big AI news announced | Tom's Guide
Here's everything Microsoft announced at its AI-focused developer eventwww.tomsguide.com