Before grounding an agent on business data, what should you assess?
Why A is the answer
Data accuracy, relevance, timeliness, cleanliness and availability
Accuracy, relevance, timeliness, cleanliness and availability decide grounding quality. Poor data leads to poor answers, however good the agent is.
In detail
- Grounding quality depends on accuracy, relevance, timeliness, cleanliness and availability.
- Assign an owner for each knowledge source.
Use case
An illustrative scenario, not a real incident.
An HR agent quotes a 2021 leave policy, because old files sit beside new ones. Reviewing the data first would have removed them.
See it in different ways
flowchart LR
A["Business data"] --> B{"Accurate, relevant, timely?"}
B -->|No| C["Clean it first"]
B -->|Yes| D["Check security and permissions"]
D --> E["Ground the agent"]mindmap
root((Grounding data))
Accuracy
Relevance
Timeliness
Cleanliness
Availability
Access controlmindmap
root(("Grounding data"))
n1["Content"]
n2["Accurate"]
n3["Relevant"]
n4["Clean"]
n5["Delivery"]
n6["Timely"]
n7["Available"]Key point: Assess data before grounding.
Why the other options are wrong
- B. Which developers prefer which IDE and language for the project
- Developer tools do not affect data quality.
- C. The visual design of the Copilot interface in each app
- Interface design does not affect data quality.
- D. The number of Teams channels that will mention the agent
- Channel counts do not affect data quality.