AI Engineering
How could organizational context become usable in everyday AI-assisted work? Sauce is a productivity-harness concept under development, combining a repository, company context and a conversational interface. It is not a commercially validated product.
Concept review · Commercial definition is still open.
The work in context
An AI system may have strong general capability while lacking the context that makes work useful inside a particular organization. Sauce explores that gap. Concept notes describe a repository, a layer of organizational context and working standards, and a conversational interface supported by a recurring intelligence loop. Later notes consider configurable, role-specific company brains with services around them. This review page makes the idea concrete enough to critique while keeping proposed product behavior distinct from delivered functionality and commercial availability.
The candidate problem is fragmented context: guidance lives in documents, reusable patterns sit in repositories, and staff reconstruct both when using AI. Validate that hypothesis with one user and workflow first. Serving every role risks developing several products before one is useful. Tech Support, Legal, Supply Chain and HR remain implementation directions, not four established Sauce platforms or product roadmap commitments.
A view of the work
Proposed approach
Choose a candidate role and recurring workflow, then validate the problem with potential users. Describe the decision or task the context would improve. Product development should follow evidence about that need rather than a broad promise of company-wide productivity.
Specify the repository, context and conversational interactions required for the first workflow. Resolve the relationship to existing components and assets. Separate common foundations from customer-specific information and services before treating configuration as a repeatable product capability.
Propose evaluations for relevance, source grounding, permission behavior and user control. Establish how context would be reviewed and kept current. Tenancy, access and data handling are product decisions still to settle, rather than capabilities implied by the storyboard.
Define product decision authority, funded development capacity, maintenance and support before a commercial offer is made. Validate demand and the delivery model. A working demonstration or proposed subscription price would not answer these questions on its own.
Candidate outputs
Illustrative example
A possible storyboard shows a specialist asking how to approach a recurring support issue. A proposed interface could assemble relevant approved guidance and reusable troubleshooting context, leaving source review and action with the specialist. The storyboard would need to test whether the context is current, authorized and useful for that role. It does not establish a working support product, a permission model or a measured productivity benefit. It is a candidate workflow for concept review.
Illustrative scenario, not a reported client result or a product capability claim.
Scope and responsibility
Sauce remains a concept in the review edition. No pricing, subscription availability, production support or delivered product is promised. The page’s activities describe candidate development work, not an available service package. The first user, minimum product, architecture relationships, product authority, capacity, tenancy, permissions, evaluation and maintenance obligations remain decisions to settle. Existing engineering capabilities should be scoped on their own merits rather than sold through an unspecified Sauce promise.
Practical questions
This page does not establish commercial availability. Sauce is a concept requiring user validation, product decisions and operating readiness. An implementation problem can be discussed separately through the existing Engineering capability.
The concept notes include a repository and organizational best-practices context as well as a conversational interface. Which interactions belong in the minimum product remains open and must be validated against a specific workflow.
Confirm the first user, distinct problem, minimum product and demand, then establish funded capacity, architecture, permissions, evaluation, support and maintenance. Those decisions precede claims about a recurring commercial product.