The Pulse
AtScale Launches AI Suite Built Around Governed Business Metrics
AtScale announced AtScale for AI on October 15, combining tools for semantic modeling, question clarification, governed computation and answer testing. Ava and ACE are available now, while CoModel and Canary are planned for a later release.
AI.info Team · Written with AI ·
AtScale is betting that enterprise AI needs clearer business definitions more than it needs a smarter model. The company’s new AtScale for AI suite aims to make AI systems use approved metrics and calculations, while its chief executive argues that models cannot supply business rules an organization has never defined.
That pitch arrives with a product caveat: only two of the suite’s four named capabilities are available now. The other two, which help create semantic models and test AI answers, are still in development.
AtScale argues the problem is business meaning
Announced October 15, AtScale for AI brings together AtScale Ava, AtScale CoModel and AtScale Canary, powered by the company’s AI Computation Engine, or ACE. The suite is designed to connect natural-language questions to shared business definitions, compute answers against those definitions and test results for accuracy.
AtScale frames the issue with a familiar question: “What was our revenue last quarter?” A system needs to know which meaning of revenue applies, which period counts and how the figure should be calculated. The company argues that a stronger language model cannot resolve those choices unless the business has set out its definitions.
“The AI industry keeps telling enterprises to buy a smarter model, but a model can’t reason its way to your definition of revenue,” said Chris Lynch, CEO of AtScale.
Chris Lynch, CEO of AtScale
Two capabilities are ready; two are pending
Ava, whose name stands for Ask, Verify, Answer, handles ambiguous requests before calculation. If a user asks for “revenue,” the release says, Ava can clarify whether the question concerns bookings, recognized revenue or contract value, then send the resolved request to ACE.
ACE applies governed metric definitions and calculation logic to enterprise data, with the aim of returning consistent results across AI agents, applications and analytics tools. AtScale says it also optimizes query performance and computational cost. Ava and ACE are available now; the company says CoModel and Canary are coming soon.
CoModel is intended to use AI to draft and maintain semantic models describing metrics, dimensions and relationships. Proposed changes remain subject to human review. Canary is designed to compare AI answers with validated results and customer-provided question-and-answer pairs, helping teams identify errors and track regressions.
Customer examples come with a proof question
AtScale’s launch release includes an account from HSBC’s UK commercial banking data and analytics head, Dr. Kevin McLafferty. He said that on a 10-terabyte table in Google Cloud, accuracy on AI-driven deterministic analytics rose from 65% to above 99% when AtScale served as the semantic layer. That is a customer claim in the company’s announcement, not a description of an independently reported test.
Carrefour’s data and analytics director, Nicolas Treanton, said the retailer already uses a semantic layer to govern thousands of key performance indicators. Extending that foundation to AI, he said, lets agents draw on the same definitions as analysts and business-intelligence tools.
AtScale’s accuracy case depends on what ships next
The launch also reflects AtScale’s support for Apache Ossie, formerly called Open Semantic Interchange, as a shared standard for preserving business logic across AI models, agents and data platforms. The company says its separate evaluation of seven large language models found that adding an AtScale semantic layer improved accuracy for every model, but the launch announcement does not include the underlying scores or methodology.
For now, buyers can use Ava to clarify questions and ACE to compute governed metrics. The suite’s broader promise—using model-assisted modeling and ongoing answer tests to improve enterprise AI—depends on CoModel and Canary moving from planned capabilities to available products.