한국어Contact

HOPFIA CORE

AI that builds and evolves company knowledge

A knowledge graph agent creates the ontology directly from company data and operating context, then continuously manages and evolves it as new information and workflow outcomes emerge. Worker agents use this living knowledge graph to reason, decide, and execute real work.

Discuss Hopfia Core

HOW HOPFIA CORE WORKS

Agents build the knowledge graph and evolve it through real work

The knowledge graph agent creates an ontology from company data and operating context, then builds and manages the graph. New facts and outcomes discovered by worker agents flow back into it.

01KNOWLEDGE GRAPH AGENT

Understand company data to build and manage the knowledge graph directly

The agent identifies entities, relations, events, and rules in company data, shapes them into an ontology, and connects them to the knowledge graph with evidence. As new information arrives, it adds, revises, and prunes the graph structure so it continues to evolve.

02BUSINESS KNOWLEDGE GRAPH

Connect company concepts, relationships, and evidence in one knowledge graph

Top, core, and domain ontologies use description logic to define concepts (TBox), roles (RBox), and constraints, then classify and infer over live data (ABox). Connected instances retain traceable references to records, documents, and vectors.

03WORKER AGENT

Execute real work and evolve the knowledge graph with every outcome

Each worker retrieves the task-specific subgraph it needs from the 02 business knowledge graph. New facts, events, and learning acquired during execution write back to it, while action results update CRM, ERP, and operational systems.

01 KNOWLEDGE GRAPH AGENT

Turn patterns and knowledge discovered in data into RDF and connect them semantically to the knowledge graph

A discovery engine suited to each data type turns statistics, features, and patterns into meaningful knowledge candidates. The agent converts each candidate into a source-linked RDF asset, then interprets shared URIs and RDFS classes and relations to integrate it with the existing graph.

THREE DISCOVERY LANESDATA EXAMPLE → DISCOVERY ENGINE → DISCOVERED KNOWLEDGE → RDF ASSET
TRANSACTION BASKETS + CUSTOMER JOURNEYASSOCIATION / SEQUENCE
ORDER + EVENT LOG
C-2041SKU-AVIEW
C-2041SKU-ENTBUY
C-3102SKU-ARENEW
→
DATA MININGAssociation RulesSequential Patterns
→
PATTERN
SUPPORT
0.42
CONFIDENCE
0.81
SEQUENCE
A → ENT
→
RDF ASSET · A12customer:C-2041likelyBuysproduct:SKU-ENT
WEEKLY REVENUE + PRICE HISTORYDESCRIPTIVE / DIAGNOSTIC
TIME SERIESW27 112 · W28 119 · W29 136
→
ANALYTICSTime-series AnalysisDriver Attribution
→
STATISTICS
GROWTH
+18.4%
ELASTICITY
−0.63
CHANGE
W29 SPIKE
→
RDF ASSET · B07event:DemandSpikeobservedAtperiod:2026-W29
ACCOUNT BEHAVIOR + OPERATIONS TELEMETRYPREDICTIVE / UNSUPERVISED
FEATURE VECTOR[ usage .82, recency .71, tickets .18, margin .64 ]
→
MACHINE LEARNINGFeature EmbeddingCluster + Anomaly
→
FEATURES
CLUSTER
C3
PROPENSITY
0.78
ANOMALY
0.91
→
RDF ASSET · C31account:C-2041rdf:typesegment:Growth
RDF / RDFS KNOWLEDGE INTEGRATIONRESOLVE IDENTITY · ALIGN CLASS · LINK RELATION · RETAIN PROVENANCE
SAME URI · customer:C-2041ONTOLOGY CLASS · CustomerTEMPORAL KEY · 2026-W29PROVENANCE · A12 / B07 / C31
likelyBuyshasFeatureobservedInindicatesrdf:typesupportedByCUSTOMERC-2041PRODUCTSKU-ENTFEATURE CLUSTERC3 · PROPENSITY .78PERIOD2026-W29EVENTDEMAND SPIKESEGMENTGROWTHSOURCE ASSETSA12 · B07 · C31
INTEGRATED INTO THE LIVE KNOWLEDGE GRAPHSEMANTICALLY LINKED · TRACEABLE · REUSABLE

02 BUSINESS KNOWLEDGE GRAPH

A knowledge graph connecting business concepts and relations to exact data locations

Top and core ontologies define shared concepts while the domain knowledge graph connects customer, product, policy, and approval relations. Reference nodes point to exact locations across CRM, ERP, documents, and vectors.

BUSINESS KNOWLEDGE GRAPHTOP ONTOLOGY · CORE ONTOLOGY · DOMAIN KNOWLEDGE GRAPH · REFERENCE NODES · ACTUAL DATA FLOW
typed astyped astyped astyped asplacescontainsbecomesusesincludesgovernsapproveslocated insegmentsownssuppliesgeneratesvalidated byTHINGENTITYASSETRULEEVENTCUSTOMERCORE CLASSPRODUCTCORE CLASSPOLICYCORE CLASSORDERCORE CLASSACCOUNTACME · B2BPLANENTERPRISEPRODUCTAI SUITEPOLICYQ3 · 12.5%QUOTEQ-77 · FINALAPPROVALCFO · APPROVEDORDERPO-8821 · ₩128MREGIONAPACSEGMENTENTERPRISE B2BOWNERJ. KIMVENDORHOPFIAINVOICEINV-88REF:C-2041TYPE CUSTOMER · URIREF:SKU-ENTTYPE PRODUCT · URIREF:POL-Q3TYPE POLICY · URIREF:PO-8821TYPE ORDER · URIREF:SEG-B2BSEGMENT URIREF:VENDOR-7VENDOR URICRM RECORDcustomer/2041VECTOR INDEXvec_08af2POLICY DOCpricing-policy.pdfERP ROWorders/8821CRM SEGMENTsegments/B2BSUPPLIER MASTERvendors/7TOP ONTOLOGYCORE ONTOLOGYDOMAIN KNOWLEDGE GRAPHREFERENCE NODESACTUAL DATA FLOW
DESCRIPTION LOGICTBOX · RBOX · AXIOMS · INFERENCE
RELATIONREFERENCEACTUAL DATA

03 WORKER AGENT

The agent understands business context and executes the work

Knowledge graph and live business data are assembled into the agent context. The worker agent uses that context to decide and execute, then writes the result back to business systems and the knowledge graph.

KNOWLEDGE GRAPH

forAccountincludesgovernedByappliesInQUOTEq:104ACCOUNTa:2041PRODUCTENTPOLICYQ3REGIONAPAC

LIVE BUSINESS DATA

DOCUMENTRenewal term24 months

MESSAGE“Approve 12.5%”

DATABASEaccount.tierenterprise

RUNTIME ENVELOPE · v3AGENT CONTEXTVALUES + PROVENANCE
L01JSON{L02WORKFLOW"workflow":"governed_action",L03SCHEMA"result":["actions","updates"],L04TAG"graph":[L05GRAPH["q:104","forAccount","a:2041"],L06GRAPH["q:104","includes","sku:ENT"],L07GRAPH["q:104","governedBy","pol:Q3"]],L08TAG"evidence":{L09DATA"renewal_months":24,L10DATA"requested_pct":12.5,L11DATA"tier":"enterprise",L12REFS"refs":["doc:p12","msg:781","crm:a2"]L13RULE},"max_discount_pct":12.5}
WORKER AGENTDECISION LOOP
A
  1. READ CONTEXTground inputs
  2. DECIDEapply rules
  3. ACTwrite updates

GRAPH PATCH

"owner": "a:2041"L05 + L11"approved_under": "pol:Q3"L06 + L07 + L10 + L13

SYSTEM WRITE PATCH /quotes/q104

"discount_pct": 12.5L10 + L13"status": "approved"L02 + L13