How resolutions work

A resolution is a structured attempt to fix a customer’s problem — not just acknowledge it. It’s the difference between “sorry to hear that” and the customer actually being made whole.

This is the part of FeedbackRobot that does the work most businesses know they should do and don’t, because it’s time-consuming and nobody enjoys it.

What opens a resolution

A resolution typically opens when:

• A negative review appears on a connected platform
• A survey response comes back with negative sentiment
• A customer replies to something with a complaint
• You open one manually

What happens

1. The problem is identified. The AI reads the whole response and works out what actually went wrong — not just that the customer is unhappy, but what about.
2. The customer is contacted. In your voice, naming their specific situation, owning it once, and either acting or asking the one thing needed to act. It never asks whether the issue is resolved. → The recovery playbook
3. It makes it right, or escalates. When it understands the problem, the AI gives an offer you’ve authorised — a promo code, a gift, a credit — or hands the conversation to a person, with a promised callback if you allow that. → Offers and promo codes · Callbacks and hand-offs
4. It closes. When the customer thanks it, accepts the offer, or goes quiet. Silence after an offer counts as settled; silence with nothing given counts as unresolved. A customer writing back later reopens it.

Resolved vs unresolved

Not every resolution succeeds, and the reporting is honest about it. You’ll see:

• Resolved — the customer said thanks or accepted an offer, or went quiet after one was given
• Pending — still in progress, waiting on a reply
• Unresolved — the customer disengaged, went quiet before anything was offered, or the AI ran out of replies with nothing to give

Each closed conversation shows why it closed, and the report counts them by reason.

The unresolved ones are the useful data. A pattern in what fails to resolve usually points at something the business can’t fix with a message — which is exactly what you want to know.

→ Reading your resolution rates

Giving the AI the facts it needs

A resolution can only offer what it knows about. If yours are sympathetic but vague — apologising without resolving — the AI is missing your policies. Upload them and it can act rather than commiserate.

→ Teaching the AI with your own documents

When it escalates to you

The agent escalates rather than guesses when a case involves:

• Allegations of illness, injury, or safety problems
• Legal threats or regulatory issues
• Discrimination or staff conduct complaints
• Requests it can’t fulfil under your settings
• Anything it can’t confidently interpret

Escalated cases appear as needing your attention on the dashboard. → Taking over a conversation

Why this beats replying publicly

A public reply is performance for future readers. A resolution is an actual attempt to fix it for the person affected.

The best outcome for a negative review isn’t a well-written public response — it’s the customer quietly updating their review because you sorted it out. That only happens if someone reaches out, and reaching out consistently is precisely what doesn’t happen when it depends on a human remembering.