Ambience, Snowflake, and Albert Invent all describe the same shift: a governed internal layer that every app reuses

Healthcare's version of the argument

Ambience Healthcare's Mike Ng argues that healthcare AI requires shared, open infrastructure. The post says new AI applications in healthcare face challenges in accessing reliable clinical context, in working across existing systems, and in operating safely at scale. Health systems are introducing AI across documentation, clinical decision support, coding, and care coordination, and without shared infrastructure, the post says, health systems risk rebuilding foundations for every new application. AI capabilities are advancing rapidly, the post notes, but bringing AI advances into everyday patient care remains a challenge. The gap is not capability; it is the layer between capability and care. m:…8647

Ambience's answer is Chorus, which the company says is designed around a System of Context, a System of Action, and a System of Governance. The vision, the company says, is an open foundation that health systems, researchers, and developers can build on. Ng has six years of experience building alongside leading health systems, according to the post. The three-system framing is doing real work: context is what the AI knows, action is what it can do, and governance is what it is allowed to do. A foundation that omits any one of the three will be rebuilt by the next app that needs it. m:…8647

Snowflake's version, with receipts

Snowflake tells the same story inside its own finance team, with numbers. The company says its sales finance team reduced monthly forecast and variance analysis time from more than a week to a few hours, built 36 AI skills in two months, and now publishes a midmonth update to the topline forecast. The mechanism is a shared profile: the CoCo profile contains definitions for bookings, pipeline, quota, and other core metrics, and it reduced the steps per AI answer from approximately 3.0 to 1.3. The analytics team certifies data sources to ensure answers are consistent and traceable. Without a profile, the company says, each extra step in an AI session adds about 0.05 credits, roughly 10 cents. The profile is the foundation; the skills are the apps. m:…4022

The adoption numbers the company cites are strong enough to take seriously and self-reported enough to discount: 87% of first adopters still use CoCo daily three months after launch, and the corporate FP&A team uses the same approach for long-range planning. The company's measurement post generalizes the point into a framework measuring AI value through adoption and fluency, where adoption measures breadth of usage and fluency measures sophistication of delegated tasks. Organizations should categorize AI work into foundational, applied development, and agentic levels, the authors propose, and can be categorized into four positions based on adoption and fluency. The distinction matters because a company can have high adoption and low fluency: everyone uses the tool, nobody delegates anything hard to it. m:…4022 m:…2610

The company's examples of fluency are concrete. Snowflake CoCo reduced diagnostic time for data pipelines from 30 minutes to under 16 minutes, improved over 500 pipelines in Q1 and Q2, and generated over 100 merged PRs in Q1 and Q2. Marketing weekly usage of CoCo reached 70% by May 2026, 93% of marketing staff report daily AI tool usage, and marketing teams built an inbound automation engine handling 2,000 to 3,000 daily emails. The company also published data-eng-bench, an open benchmark of 103 data engineering tasks, and says CoCo leads general-purpose coding assistants on quality for cost in that benchmark. The constraints the authors name are managerial, not technical: discovery is the primary constraint once dozens of reusable workflows exist, reviewer attention is a critical management constraint for agentic workflows, and power users should teach workshops rather than relying on vendor training. m:…2610

Capital follows the foundation builders

Albert Invent's funding announcement fits the same pattern in materials R&D. The company announced a growth equity investment from Applied Ventures, LLC, which is the venture capital arm of Applied Materials, Inc. Applied Ventures sees opportunity for Albert's technology in the semiconductor ecosystem, and the investment will be used to expand Albert's platform and global footprint. The platform combines scientific expertise, scientific ontology, and AI, targets industries including energy and semiconductors, and its AI capabilities are aimed at shortening development cycles and reducing time to market. The platform includes zero data retention, fine-grained access controls, and reasoning audit trails. Those last three features are the governance layer, stated as product features: a foundation that enterprises can trust with their science. m:…4018

Read together, the three stories describe a maturing enterprise AI market. The first wave sold apps. The current wave sells the layer the apps share: context definitions, certified data, permissioned actions, and audit trails. The buyer question changes accordingly. Instead of asking which app has the best demo, ask which foundation your next ten apps will reuse, who governs it, and what it costs to change it later. The companies in this story all answer that question the same way: build it once, govern it visibly, and let every team build on top. m:…8647 m:…4022