Enterprise Future Guide: Maintained knowledge becomes shared operating infrastructure

Koko: Welcome to Koko Knows, and this is the Enterprise Future Guide. I'm Koko, here with Sam and Max. Today's hypothesis: maintained knowledge becomes shared operating infrastructure. The claim is that reusable definitions, policies and decision context could matter more than raw connectivity as agents operate across what you'd call headless software, applications consumed through APIs rather than fixed screens.

Max: I'm Max. What does it actually cost to keep that knowledge current? If an agent pulls a stale supplier definition or a bad prior approval, you've scaled that mistake across every workflow touching it.

Koko: That's the core risk. Foundation Capital argues that exception records and approval context could become valuable new systems of record. But Forrester's abstract on the AI CDO makes a distinct point: the semantics and lineage mandate belongs to the CDO, not merged into the CIO's infrastructure role. Both must be funded separately.

Sam: I'm Sam. Worth flagging that Foundation Capital is an investor thesis, and the Forrester piece is abstract-only access, so the confidence level is limited. HFS Research adds an independent structural argument from a public roundtable: context needs an owner, a budget and measurable quality. Fourteen responses, so not a large survey, but the accountability logic holds.

Max: Even with clear owners, there's a gap between the CDO holding semantics and the CIO holding the platform. Think of a procurement workflow where approved supplier definitions live in one system and spend limits in another. An agent bridging them without reconciliation is a control failure, not an efficiency gain.

Koko: That conditional example is exactly where the alternative scenario bites. Local context may simply be too specific to justify enterprise-wide reuse, and maintenance costs could outrun the value captured. What would change the hypothesis is fewer recurring exceptions across more than one consumer workflow, sustained and measurable over time.

Sam: And connectivity alone doesn't establish shared meaning or reliable authority. Andreessen Horowitz argues agent performance depends on maintained context, but that's an investor argument, not a demonstrated enterprise return. The cost and governance of maintenance still need testing on real workflows.

Max: Supplier exit is the part that gets missed before scale. If agents depend on an external context graph, the CFO and procurement need exportable evidence and a tested reconciliation path, not just a contract clause discovered during offboarding.

Koko: So by 2030 the decision is whether maintained knowledge is owned infrastructure or a recurring cost you pay again for every new capability. By 2035, if that ownership is absent, the Stranded Intelligence scenario from the Enterprise Future Guide becomes more likely: agents proliferate, definitions conflict and review burden absorbs the gains.

Sam: The action is concrete: the CDO, CIO and a business owner co-fund one maintained knowledge product, test context quality and access rights across a real workflow, then agree with the CFO and procurement on evidence export and supplier exit before expanding anywhere else.

Max: That's a bounded decision gate. One workflow, verified quality, exit terms confirmed. If that doesn't reduce recurring exceptions, the hypothesis narrows before the budget grows.

Koko: That's the episode. This micro expands the knowledge engineering chapter from the parent episode, Enterprise Future Guide: Capabilities, Leadership and Value in 2030 and 2035. The companion reading, Enterprise Capabilities, covers the full operating contract including authority, economics and portability. Both are in the Koko Knows feed.