Goodwill is discretionary, which is exactly why it needs rules. The agent applies your goodwill matrix to the situation, checks what the customer has already received, and drafts a proposed gesture with the reason. A person approves anything with monetary value. Requests that are really about the firm's fault go to complaints, because redress owed is not goodwill.
What is a goodwill gesture, and who owns it?
A goodwill gesture is something a firm offers a customer that it does not owe them: a credit, a voucher, a fee waiver, an upgrade, a handwritten apology. It recognises an inconvenience or keeps a valued relationship. Because it is discretionary, it is prone to inconsistency: customers who complain loudly or ask the right adviser get more, and others get nothing for the same experience.
The head of customer operations usually owns the goodwill policy and budget, with finance. A goodwill gesture matrix sets what may be offered in which situation and who may approve each amount. Team leaders apply it. The agent makes the matrix consistent, in the spirit of our guide to the AI support agent that knows when to say no.
What are the steps, and who acts at each one?
- Agent, depth 1: recognises the request, explicit ("I think I deserve something for this") or implicit, where your policy says a gesture should be considered.
- Agent, depth 1: separates goodwill from redress. If the customer says the firm got something wrong and caused them loss, that is a complaint, and any payment is redress through the complaints process.
- Agent, depth 1: reads the situation: what happened, how long it lasted, what it cost the customer in time or money, and any vulnerability.
- Agent, depth 1: checks history: gestures given to this customer within the period your policy sets, and any open complaints.
- Agent, depth 2: applies the matrix and proposes a gesture type and value, with the matrix row that supports it.
- Agent, depth 3: offers non-monetary gestures your matrix allows the agent to give directly, such as a formal apology or a priority callback.
- Person, adviser or team leader: approves monetary gestures within their authority. Larger amounts go up the authority chain.
- Agent, depth 3: tells the customer the outcome and how the gesture will be applied, from the approved record.
- Person, operations lead: reviews gestures for consistency across customers.
| Step | The agent does | Depth | Stays with a person |
|---|---|---|---|
| Recognise | Spots the request | 1 | Nothing |
| Separate | Routes redress to complaints | 1 | Complaints decision |
| Situation | Reads what happened | 1 | Nothing |
| History | Checks past gestures | 1 | Nothing |
| Matrix | Proposes type and value | 2 | Approval |
| Non-monetary | Gives allowed gestures | 3 | Nothing |
| Approve | Nothing | none | Monetary gestures |
| Tell | Confirms outcome | 3 | Nothing |
| Review | Nothing | none | Consistency review |
Which controls matter, and what evidence does each leave?
Matrix row on every proposal. Evidence: the row cited, so reviewers can compare like with like.
Credits are money. Issuing a credit or refund is depth 4, held for a person. Can an AI agent issue a goodwill credit? sets out when that could change. Evidence: held verdicts and approvers.
History checked. Repeat gestures beyond the policy period need senior approval. Evidence: the history check on each case.
No promises before approval. "I'll get you a credit" is a commitment. The agent says it has requested a gesture. Evidence: scope-gate results.
Redress routed. Evidence: routing records to complaints.
What should you measure?
- Gestures by matrix row, type and value.
- Approver changes to proposed values, up or down.
- Customers with repeated gestures, and why.
- Gestures given without a request, where your policy says they should be offered proactively, such as after a known failure. Customers who do not ask should not lose out.
- Complaints after a gesture, a sign it was treated as a substitute for fixing the problem.
Where does it usually go wrong?
Loudness wins. The agent proposes more for angry customers. Anger is not a matrix input; impact is.
Goodwill replaces redress. A small voucher is offered where the firm owes the customer money. That can leave a customer short and still entitled to complain.
Gestures are offered to stop cancellations. Retention offers are commercial decisions with their own rules; keep them separate, as in handling cancellation requests with an AI agent.
Returns become gestures. A customer outside the returns window asks for an exception. That is a policy exception, decided as in handling returns requests with an AI agent, not a goodwill credit.
Different channels give different answers. A customer who calls gets a credit from an adviser; one who chats gets the matrix result from the agent. If advisers do not use the same matrix, the agent makes the inconsistency visible rather than fixing it. Apply the matrix to people and agents alike.
Approvals sit in a queue. The customer is told a gesture has been requested and then hears nothing for a week, which undoes the goodwill. Set a response time for approvals and tell the customer what it is.
The matrix is never revisited. Situations not in the matrix pile up as "other". Review those cases and add rows.
Where Pryme fits
In Pryme, the goodwill agent's role allows proposing from the matrix and giving the non-monetary gestures you list. Monetary gestures sit at depth 4 and are held for a named approver, whose own authority limit is one of the customer's limits in the minimum chain. Every proposal, approval and change is filed for consistency review.
Questions people ask
Should the agent tell customers about the goodwill matrix?
No. The matrix is an internal tool. The agent explains the outcome and the reason in plain terms.
Can the agent decline a goodwill request?
Where the matrix gives nothing for the situation, the agent can say no gesture applies and explain why, with a route to a person if the customer disagrees. Monetary approvals stay with people.
How do we stop the matrix becoming an entitlement?
Keep values modest and tied to impact, review repeat recipients, and make sure gestures sit alongside fixing the underlying problem.