06AI SALIENT AI Talk to us

Use case 02 · AI for Marketing + service

How AI can answer customer questions.

Help customers find approved answers through a website or support channel, with clear source ownership and a useful path to a person when the question needs judgement.

Start with one workflowDiscuss this use case
Tiny staff routing glowing information from a giant book into a speech bubble
02One recognizable result

The friction

What is happening now?

Customer answers are often spread across public pages, service policies and staff knowledge. A chatbot should use an approved customer-facing set, not assume that every internal document or visitor message is suitable source material.

A practical system

Build customer answers from approved public sources.

  1. 01

    Choose a customer question set

    Collect recurring public questions and current approved answers. Record the owner and review date; keep private staff sources outside the initial scope.

  2. 02

    Define answer and handoff rules

    Decide what the chatbot may answer, what needs clarification and what must reach customer service. Include missing, conflicting and out-of-scope examples.

  3. 03

    Test the customer journey

    Check the answer and the next action on the chosen channel. Verify source updates, staff-unavailable handling and whether a person receives enough context to continue.

  4. 04

    Fix the underlying gaps

    Review wrong or unanswered questions with the content owner. Update the approved source and retest; do not turn conversation text directly into new facts.

Tiny people assembling a knowledge assistant
Build the controlled workflow.
Tiny people improving customer support workflows
Observe the result and improve it.

Proof

Measure answer usefulness and customer handoffs.

  • Correct answers against reviewed customer examples
  • Unsupported answers and questions that should have reached a person
  • Handoffs received with sufficient context and an accountable owner
  • Repeated customer questions after an answer, alongside staff corrections

Human safeguards

Keep people in control.

  • Approved-source retrieval
  • Visible source links
  • Escalation rules
  • Feedback and review logs

Related Salient AI service

The capability behind this use case.

Explore website and messaging chat

Before the pilot

Questions before automating customer answers.

Can customer questions improve the assistant?

They can reveal gaps, but a content owner must verify the proposed answer and update the approved source. Review and testing belong between a conversation and a change to what the assistant may say.

What if we also need an employee assistant?

Treat it as a separate audience and access design. Internal sources require user permissions and source-level testing; they should not be attached to a public customer widget by default.

Related resources

Explore the next step in this workflow.

One useful beginning

Bring us the real bottleneck.

We will map the people, information, exceptions and measure of success before recommending what should be built.

A practical first stepSchedule an opportunity session