You'll sit at the intersection of business analysis, data architecture, and applied AI — working closely with AI/ML engineers, platform architects, and business stakeholders to ensure the agent's "understanding" of the business is accurate, current, and governed.
Key Responsibilities
Semantic Layer for Agent Consumption
Define and maintain the semantic model (metrics, dimensions, entities, synonyms, relationships) that AI agents use to map natural-language business questions to correct data queries (SQL, semantic layer APIs, or knowledge graph traversals)
Design metric and entity definitions with agent-readability in mind: unambiguous naming, explicit synonyms/aliases, disambiguation rules for overloaded business terms (e.g., "revenue" meaning different things in different departments)
Partner with AI engineers to structure the semantic layer for tool-calling/function-calling patterns — ensuring agents select the right metric, filter, and grain when generating queries
Continuously test and tune semantic definitions against agent query failures, misinterpretations, and hallucination patterns
Context Layer & Grounding for AI Agents
Design the context layer that supplies agents with the business rules, caveats, definitions, data lineage, and usage guardrails needed to ground responses (e.g., via RAG retrieval, system prompts, or structured metadata injection)
Curate and structure business context objects — glossaries, policy documents, calculation logic, exceptions/edge cases — into formats optimized for retrieval (chunking strategy, embeddings-friendly structure, metadata tagging)
Define context scoping and prioritization rules so agents retrieve the most relevant and correct context per query, minimizing noise and conflicting information
Work with platform teams to version and govern context content so agent behavior stays aligned as business rules evolve
Agent Behaviour & Quality Assurance
Partner with AI/ML engineers to evaluate agent outputs against expected business answers; identify root causes when agents misinterpret semantic or contextual inputs
Build test question sets ("golden queries") representing real business questions to validate semantic/context layer coverage and accuracy
Define feedback loops where incorrect agent responses trace back to semantic gaps, missing context, or ambiguous definitions — and drive remediation
Support prompt/context engineering efforts by translating business nuance into structured inputs agents can reliably use
Cross-Functional Collaboration
Act as the primary bridge between business domain experts and the AI platform/engineering team, ensuring business intent is faithfully encoded into semantic and context structures
Collaborate with Business Stakeholder, data engineers and architects to ensure the underlying data model supports the semantic layer the agent depends on
Work with data governance teams to ensure context/semantic content meets accuracy, security, and access-control standards (e.g., row-level security respected by agent-generated queries)
Documentation & Governance
Maintain a living business glossary and context repository that serves as the authoritative source for both human users and AI agents
Document the semantic layer's coverage, known gaps, and disambiguation logic for auditability
Establish a review/change-management process for updating semantic and context definitions as the business evolves, without degrading agent performance
Required Qualifications
6+ years as a Business Analyst, Data Analyst, or Knowledge/Ontology Analyst, with recent exposure to AI/LLM-powered platforms, copilots, or agentic systems
Strong grasp of semantic modeling concepts (metrics layers, ontologies, knowledge graphs) and how they differ when designed for machine/agent consumption vs. human dashboards
Understanding of how LLMs consume structured context — RAG architectures, embeddings, chunking strategies, function/tool calling, prompt/context engineering
Experience translating ambiguous business language into precise, structured, disambiguated definitions
Strong SQL skills and ability to work with data models, APIs, or knowledge graphs that agents query against
Experience writing and structuring documentation (glossaries, business rules, metadata) in formats consumable by both people and AI systems
Excellent stakeholder management — able to extract and reconcile business logic from multiple domain experts
Preferred Qualifications
Hands-on experience with semantic layer tools (such as UnifyApp, Atlan, dbt Semantic Layer, Cube, AtScale, LookML) and AI orchestration frameworks (LangChain, LlamaIndex, Semantic Kernel, or similar)
Experience with vector databases and RAG pipeline design (such as Pinecone, Weaviate, pgvector, etc.)
Familiarity with knowledge graph technologies (Neo4j, RDF/OWL) for entity/relationship modeling
Experience evaluating LLM/agent outputs — building eval harnesses, golden datasets, or human-in-the-loop review processes
Background in AI governance, model risk, or responsible AI practices as applied to enterprise data
Exposure to enterprise data catalogs (Collibra, Alation, Atlan, Purview) and metadata management
What Success Looks Like
AI agents consistently interpret business questions correctly because the semantic layer resolves ambiguity before the query is even generated
Agent responses are grounded in accurate, current business context — reducing hallucinations and incorrect metric usage
Business stakeholders trust agent-generated insights because the underlying semantic/context layer is transparent, auditable, and governed
A measurable reduction in agent query failures/misinterpretations over time, tracked via golden-query test suites
This role defines what the agent should know and how it's structured (business/semantic ownership), while assuming AI/ML engineers own the underlying retrieval and orchestration infrastructure.
30-60-90 Day Plan: Sr. Business Analyst – Semantic & Context Layer for AI Agents
Phase 1: Days 1–30 — Learn, Map, Assess
Theme: Understand the business, the data, the current AI agent/platform state, and where the gaps are.
Key Activities
Meet with Busienss Stakeholders, AI/ML engineering, data platform, and architecture teams to understand the current agent stack (LLM used, orchestration framework, RAG setup, semantic layer tooling if any)
Meet with business stakeholders across key domains (Finance, Sales, Ops, Marketing , IT, HR, GTM, etc.) to understand how they currently define and use core metrics/entities
Inventory existing semantic assets: BI semantic layers, data dictionaries, business glossaries, metric definitions — assess quality, coverage, and consistency
Run/observe the AI agent (if live) against a sample of real business questions; log where it fails, hallucinates, or misinterprets terms
Review existing data models, lineage documentation, and access/security controls relevant to agent queries
Identify 3–5 high-value, high-ambiguity business domains to prioritize (e.g., Marketing, Salesforce (L2C), revenue, churn, customer segmentation)
Deliverables
Current-state assessment document: semantic layer maturity, context layer gaps, agent failure patterns
Initial glossary of top ambiguous/overloaded business terms
Initial Context Layer and Semantic Layer – List of Objects, Tables , Fiels , Domain Terms, Rules, process if any , Metrics and Formulae.
Stakeholder map (who owns which metrics/definitions)
List of 15–20 "golden queries" collected from real business users to use as early test cases
Complete 3 Pilot usecases for Context Layer
Success Markers
Clear understanding of where agent inaccuracy stems from (semantic gaps vs. context gaps vs. data model issues)
Relationships established with AI engineering and key business stakeholders
Phase 2: Days 31–60 — Design, Build, Test
Theme: Start building and iterating on the semantic/context layer for the priority domains identified in Phase 1.
Key Activities
Define structured metric and entity specs for priority domains: name, definition, synonyms/aliases, calculation logic, grain, owner
Design disambiguation rules for known overloaded terms (e.g., department-specific "revenue" definitions)
Structure context objects (business rules, caveats, exceptions) for RAG retrieval — work with AI engineers on chunking, tagging, and metadata format
Run the golden-query set against the agent before and after semantic/context updates; document accuracy shifts
Set up a lightweight review/change-management process for updating definitions without breaking agent behavior
Begin building the living business glossary / context repository as the shared source of truth
Establish a feedback loop: agent failure → root cause tagged (semantic gap, context gap, data issue) → routed to fix owner
Deliverables
Semantic layer v1 for 3–5 priority domains, live in the agent's query/tool-calling logic
Context layer v1 (structured business rules/definitions) integrated into RAG or prompt context
Expanded golden-query test suite (40–50 queries) with pass/fail tracking
First iteration of the governance/change-management process doc
Success Markers
Measurable improvement in agent accuracy on golden queries for priority domains (target: define a baseline % and improvement goal, e.g., 60% → 80%)
Business stakeholders in priority domains have reviewed and signed off on definitions
A repeatable process exists for adding new metrics/context to the layer
Phase 3: Days 61–90 — Scale, Govern, Prove Value
Theme: Expand coverage, formalize governance, and demonstrate measurable impact to stakeholders.
Key Activities
Expand semantic/context coverage to additional business domains beyond the initial priority set
Formalize the golden-query evaluation as an ongoing QA process (e.g., run weekly/bi-weekly, track trend over time)
Present before/after agent accuracy results to leadership and stakeholders, using concrete examples of resolved ambiguity/hallucination
Finalize governance model: who approves new metric definitions, how versioning works, how context gets deprecated/updated as business rules change
Train business users and analysts on how to contribute to or query the glossary/context repository
Document known limitations and a roadmap for the next 90 days (e.g., additional domains, knowledge graph expansion, more advanced disambiguation)
Deliverables
Documented, governed semantic + context layer covering majority of high-value business domains
Ongoing QA/eval process with historical accuracy tracking
Stakeholder-facing readout: impact summary (accuracy improvement, reduced hallucinations, adoption metrics)
90-day forward roadmap
Success Markers
Agent demonstrates sustained, measurable accuracy improvement across expanded domains
Governance process is operating without the BA being a bottleneck for every change
Leadership/stakeholders have visibility into and confidence in the semantic/context layer's value
SailPoint is an equal opportunity employer and we welcome all qualified candidates to apply to join our team. All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, protected veteran status, or any other category protected by applicable law.
Alternative methods of applying for employment are available to individuals unable to submit an application through this site because of a disability. Contact applicationassistance@sailpoint.com or mail to 11120 Four Points Dr, Suite 100, Austin, TX 78726, to discuss reasonable accommodations. NOTE: Any unsolicited resumes sent by candidates or agencies to this email will not be considered for current openings at SailPoint.