Trends · Core On positioning

Notice: This post is Core On’s own interpretation and strategy note, based on a public press release. It has not been reviewed or confirmed by the companies mentioned (Accenture, Within, or others) and does not represent their views or plans.

One-line take

Core On focuses on an on-prem plant knowledge agent whose source of truth is Domain KG + Document RAG built from equipment, procedures, and documents. This note sets out what Core On intends to do, prompted by the 2026-09-23 Accenture press release.

Primary source: Accenture newsroom · 2026-09-23. The same-day Trends wire covers the announcement itself.

What the press release says (summary of Accenture’s announcement)

According to Accenture’s release, Accenture invested in Within through Accenture Ventures, and the two companies are partnering to help clients understand how work actually gets done and deploy AI agents. The release describes Within’s platform as building a “Work Brain” from how people work. Investment terms were not disclosed. See the source above for details.

What Core On focuses on

On the plant floor, answers are hard to keep consistent when they rest only on manuals, tickets, and individual memory. Core On v1 focuses on treating Domain KG (equipment, edges, SOPs) + Document RAG as the source of truth and shipping a plant knowledge agent on-prem first, with answers that cite their sources or refuse (cite / refuse). Existing systems such as Jira are connected as connectors rather than treated as SoT. The longer path is CoreEdge (Jetson) with CoreVLA/CoreLM and CoreOS. The intended audience is industrial AI adopters with significant plant/OT, engineering, and security requirements.

Core On’s direction

  • Context first: Start from the premise that agents need grounded context, and organize equipment, procedure, and document knowledge for industrial settings.
  • Plant requirements: Design on-prem first around safety, procedure compliance, and security needs, with citation or refusal as the default answer behavior.
  • Stay in scope: Concentrate on plant knowledge rather than building an office-work context layer.

External one-liner

Core On industrializes AI for how plants actually run—equipment, procedures, and documents as the source of truth, on-prem first, answers that cite or refuse.