Use learned knowledge for general reasoning
Good for concepts and patterns; unsuitable as the authoritative source for current contact or employer facts.
Waterfall contact data grounding
Ground AI agents with waterfall contact data.
When an agent needs a work email or business phone number, TargetWise routes the requested field through eligible specialist sources and returns a structured result. Company enrichment and reverse email lookup provide supporting context through separate routes.
A not-found result is a stop signal, not permission to synthesize.
Illustrative grounded agent context
IllustrativeRequestFind a work email for this known contactProfessional profile supplied · work email requested
TargetWise routeOne entity per lookup
Identity resolutionPerson and company matched
Work emailReturned to workflow
—Work phoneNot requested
15+ sourcesavailable for contact waterfall routing
Structuredfields instead of loose passages
Explicitreturned or unavailable context
25maximum candidates per search page
Grounding, plainly defined
Connect the agent’s task to an external, checkable result.
Grounding means giving the model retrieved context for the specific entity in front of it, rather than trusting its training data to supply a current professional fact.
Language models are useful for interpretation, synthesis and planning. They are not a reliable source for a person’s current work email, business phone, employer or a company’s present details. Those values are exactly where an explicit enrichment request belongs.
TargetWise receives the identity context, uses the applicable contact, company or reverse-email route and returns structured fields with a clear outcome, request ID and retrieval timestamp. Company provenance is reported only when the selected route supplies it.
Grounding methods and controls
Choose the retrieval form that matches the question.
Model memory, document retrieval and structured enrichment each have a different role. Contact and company fields need the structured path.
Retrieve explanatory text and passages
Useful for policies, notes, knowledge bases and documents where the relevant evidence is unstructured text.
Resolve a known entity into named fields
Use when the workflow needs available contact or company context in a structured response.
Supply enough context to reduce ambiguity
Profiles, domains and person-company combinations help distinguish similarly named subjects.
Use only what the decision needs
Select required contact fields and filter returned company-profile fields instead of consuming everything by default.
Treat no-match as grounding too
A clear absence prevents a guessed value from silently entering the workflow.
Grounding sequence
Resolve, retrieve and branch without ambiguity.
Keep identity, retrieval and decision logic visible as separate steps.
Identify
Extract one person or company identifier from the task.
Request
Choose the operation and requested fields; select contact fields or filter the returned company profile.
Retrieve
TargetWise calls the product endpoint and returns structured fields with explicit missing values.
Decide
Branch on the returned outcome using deterministic acceptance rules.
No-match discipline
The safest grounded answer can be “not found.”
A system loses its grounding when it treats missing data as a prompt for creative completion. The workflow should make absence explicit and decide what additional evidence—if any—is appropriate.
Do not retry an unchanged request in a loop. A useful next step requires a new identifier, a manual check or a decision to proceed without that field.
- Return the no-match state to the caller
- Ask for a stronger identifier when the task permits
- Do not infer an email pattern or phone number
- Log the request outcome for later review
Safe outcome handling
What the workflow should do next
1Matched — apply field acceptance rules
2Partial — use only returned returned fields
3Not found — stop or strengthen identity
4Identity doubt — reject or route to review
Where structured grounding helps
Give the agent just enough business context for the task.
Grounding is most useful when a task needs a filtered candidate set or a selected record lacks specific context.
Ground a signup in company context before routing.
Resolve the known domain or business identifier when the workflow needs firmographic context for a rule.
Ground a meeting brief in retrieved company data.
Request context when the brief is created and distinguish returned fields from model-generated narrative.
Complete one known professional record safely.
Request the missing field and preserve an explicit no-match path instead of filling it speculatively.
Add a structured business-data tool to a copilot.
Let the host control tool permissions while the model decides when the current task would benefit from retrieval.
Frequently asked questions
What to know before you build.
Grounding improves reliability by making retrieval and absence explicit. It still needs application-level controls and appropriate human judgment.
Test real grounding cases
Use the Free plan on records your team already understands.
A representative test shows where structured enrichment adds decision-ready context and where the workflow needs stronger identifiers or a no-match branch.