Data foundations
Context before agency.
The first week is in the org, reading what already exists. Data Cloud grounds the agent. Agentforce acts. Graph and vectors hold memory. MuleSoft fronts the stores it does not own. The honest no: if the work is a software factory with a named delivery date, we are the wrong partner.
Certified Partner since 2010 · MVP Hall of Fame · 200+ agents in production · UAE and US desks
Split
Context is not agency. Memory is not the warehouse.
- 01
Systems of record
Salesforce · ERP · HR and ops · Files
- 02
MuleSoft
Canonical model · Gateway · Front, do not own
- 03
Context
Data Cloud · Clean Knowledge · Permissioned
- 04
Memory
Graph for relationships · Vectors for similarity
- 05
Agency
Agentforce · Or a custom runtime when the desk is elsewhere
Mermaid source
flowchart TD
A["Systems of record"] --> B["MuleSoft: canonical model and gateway"]
B --> C["Context: Data Cloud or clean Knowledge"]
B --> D["Memory: graph and vectors"]
C --> E["Agentforce: agency"]
D --> EMemory
Relationships in a graph. Similarity in a vector store.
MuleSoft can query both. It does not become the graph engine or the warehouse.
Lineage
Pick a lineage system. We will say which.
Informatica, Data Cloud, or the metadata tool you already run. Two lineages for the same field is a finding.
Failure modes
Three ways this stalls before the agent ships.
Questions
What we actually say.
- Is this the same as data readiness?
- Readiness is the go or no-go. Data foundations is the architecture. If the context layer is missing, you still get a readiness note, not a build.
- Do you move the graph into MuleSoft?
- No. We front it. The store of record stays the store of record.
Next
What to read next.
The brief · one email
Name the systems of record.
We will tell you what is context, what is memory, and what should not be copied.
