Exercise: fix a retailer's discovery and service gaps

Original fictional scenario. No cloud account, API calls or paid services are required. Difficulty: intermediate · Estimated duration: 15–20 minutes

Northwind Home, an invented furniture retailer, lists five complaints from its customer operations review. Each needs one primary Google Cloud offering.

#	Complaint
1	Shoppers type "sofa for small flat" and the site search returns nothing unless they type a product code.
2	The help chatbot invents a returns window instead of quoting Northwind's own policy.
3	Half of all calls are "where is my order?" and wait in the queue for a person.
4	Representatives handling disputes search four manuals mid-call and write notes for ten minutes afterwards.
5	Managers cannot say why call volume rose last month or how callers felt.

Your decision

Assign one offering to each complaint and give the one-line reason.
Order a single customer contact through the products you chose for complaints 3–4, naming what happens when the first step cannot resolve the issue.
One manager proposes Grounding with Google Search for complaint 2. Explain why it is the wrong source.
If Northwind also wanted the chatbot to mention a public holiday announced this morning, what would change, and what would stay the same?

Rubric (10 house points)

5 points: one correct offering per complaint, with its reason.
2 points: the contact flow — virtual agent first, handoff to a person with Agent Assist when unresolved.
2 points: explain that Google Search grounding uses public web data, which does not contain Northwind's policy.
1 point: add web grounding for the public fact while keeping policy answers grounded on Northwind's documents.

Reference solution

1 → Agent Search over Northwind's site: it builds search for public websites and replaces keyword matching with conversational search. 2 → grounding on Northwind's own documents in Agent Search, so answers come from its policy. 3 → a virtual agent built in CX Agent Studio, answering order-status questions without a person. 4 → Agent Assist, which suggests documents and responses during the call and summarizes the interaction when it ends. 5 → CX Insights, which analyses caller sentiment and call topics.

The contact flow: the contact center platform queues and routes the contact to the virtual agent; if it cannot resolve the issue, the conversation is handed off to a human agent, who receives Agent Assist suggestions.

Rejected alternatives: Grounding with Google Search for the policy (public web data cannot know Northwind's rules); Agent Assist for complaint 3 (it helps representatives, but the goal is to avoid needing one); Insights for complaint 4 (it analyses afterwards, not during the call).

For the holiday announcement, add Grounding with Google Search for current public facts; the returns policy stays grounded on Northwind's documents. Grounding source follows the question.

Sources and scope

All organisations, names, numbers, thresholds, the rubric and the conclusions above are house-authored. The product facts the reasoning depends on are grounded in:

Fix a retailer's discovery and service gaps — decision exercise

Exercise: fix a retailer's discovery and service gaps

Your decision

Rubric (10 house points)

Reference solution

Sources and scope