Customer Support
The highest-volume agent in almost every deployment, and the one where the questions are dullest, most repetitive and most reliably answerable. That combination is exactly what automation is for.
The highest-volume agent in almost every deployment, and the one where the questions are dullest, most repetitive and most reliably answerable. That combination is exactly what automation is for.
What this agent is for
A support agent exists to remove the contact that consumes trained people without using their training. In most organisations, somewhere between half and three-quarters of inbound support contact is a small number of factual questions asked over and over: where is my order, what are your hours, how do I return this, why was I charged this, what does my policy cover.
None of those require judgement. All of them currently occupy someone who could be handling the cases that do. The support agent takes the repetitive layer and leaves the rest, which is a different proposition from replacing the function — and a far more defensible one.
The questions it handles
Where is my order?
Status from the live record rather than an estimate, phrased as a sentence.
What are your hours?
Location-specific opening times, including exceptions and holidays, which is a surprisingly large share of out-of-hours contact.
How do I return this?
The actual process, in order, with what the customer needs to have ready.
Why was I charged this?
An explanation drawn from the billing record, which prevents the escalation that follows an unexplained charge.
Am I covered for this?
Eligibility against the customer's own terms rather than a general policy statement, where the systems allow it.
I need to speak to someone
Recognised as an instruction and acted on immediately, not answered as though it were a question.
A typical conversation
- The customer opens with a problem, not a categoryNo menu. They type or say what is wrong in their own words, in whichever language they think in.
- The agent establishes what is neededIf it can answer from your material, it does. If it needs one detail — an order number, a branch — it asks for that one thing rather than presenting a form.
- It answers from your materialGrounded in your published policies and procedures, in the language of the question, in the tone you set.
- It confirms the problem is solvedA closing check rather than an assumption. This single step catches a large share of the cases that would otherwise return tomorrow.
- If it cannot help, it says so earlyRather than three failed attempts followed by a reluctant escalation, which is the pattern customers find most frustrating.
- Handover carries everythingThe agent who picks it up sees the whole conversation and continues rather than restarting.
How this agent is configured
| Tone | Calm and plain. Support contact frequently begins with someone already annoyed, and enthusiasm reads as tone-deaf. In Arabic, a slightly more formal register than the English equivalent is usually correct. |
|---|---|
| Knowledge | Published policies, procedures, FAQs, product information and — where connected — the customer's own order and account records. |
| Escalation triggers | Complaints, distress, anything about money moving, repeated failure, and any explicit request for a person. |
| Never states | Anything about another customer's account, anything not supported by your material, and any commitment about compensation or goodwill. |
| Working hours | Answers at all hours; makes clear when a human is available and does not queue silently outside those hours. |
| Languages | Arabic and English, with the reply following the language of the question. |
Where it hands over to a person
Support is the agent where escalation rules matter most, because it is the agent most likely to meet someone who is already unhappy. The rule that pays for itself immediately is escalation on repetition: if the assistant has failed to resolve something twice, the third attempt will not go better, and continuing to try converts mild irritation into a complaint.
Complaints themselves should never be handled automatically. A complaint is not a question with an answer; it is a signal that the relationship needs a person. Attempting to resolve one with a well-written policy paragraph is the single most reliable way to make it worse.
The third trigger is anything involving money moving — refunds, goodwill, compensation. The assistant can explain the policy; committing to an amount is a decision with consequences and belongs with someone accountable for it.
What it should never do
Promise a refund
It can explain the refund policy. Committing to one is a decision with financial consequence and belongs with a person.
Handle a complaint to conclusion
A complaint is an escalation by definition. Answering it with policy text reliably makes it worse.
Discuss another customer
Even where the person asking is convinced they are entitled to. Identity and entitlement are separate questions.
Give legal or regulatory advice
Explaining what a policy says is fine. Advising what someone should do about it is a different act.
Guess at an account detail
If it cannot retrieve a value, it says so. A guessed order date is worse than no answer.
Argue
Where a customer disputes something, it escalates. Systems that push back generate complaints about the system rather than the original problem.
Why a separate agent rather than one that does everything
The instinct is usually to build one assistant that handles everything, on the grounds that customers do not care about your org chart. The instinct is wrong, and for a specific reason: tone, knowledge and risk are different per function, and a single agent forces you to average across all three.
Knowledge is the obvious one. The material a support agent should answer from is not the material a sales agent should answer from, and mixing them means each answers from a corpus containing a large amount of irrelevant text. Retrieval quality falls as the corpus widens, so a general assistant is measurably worse at every individual job than a narrow one.
Tone is less obvious and matters more than people expect. The register appropriate for chasing a payment is not the register appropriate for reassuring someone whose service is down. In Arabic this is sharper still, because formality carries more weight and getting it wrong reads as rudeness rather than as informality.
Risk is the one that decides it. Different functions need different limits — what a finance agent may state about an account is different from what an HR agent may state about an employee, and those limits are far easier to set, test and audit per agent than as conditions inside one large one. Separate agents make the question "what is this allowed to say?" answerable.
How this agent should be measured
| Resolution without escalation | Useful, but never on its own. An agent that makes escalation hard scores well here while the relationship deteriorates — and the deterioration surfaces months after the metric improves. |
|---|---|
| Repeat contact within 48 hours | The number that exposes false resolutions. A conversation marked resolved that returns tomorrow was not resolved; it was deferred. |
| Abandonment mid-conversation | Where people gave up. Rises with latency, with menu depth, and with any answer that reads as evasive rather than unhelpful. |
| Escalations requiring repetition | The share of handovers where the customer had to say something twice. This is the single clearest measure of whether handover works. |
| Answer grounding | What proportion of answers trace to a source document. The number a regulator or an internal auditor eventually asks for. |
| Language accuracy | Whether replies came back in the language of the question. The most common complaint about bilingual assistants, and the easiest to instrument. |
| Satisfaction after escalation | Whether people who reached a person were satisfied. If this is high and containment is low, the deployment is working correctly. |
Mistakes worth avoiding when configuring it
Feeding it marketing copy
The questions people ask are answered in policies, procedures and price lists — not on the brochure page. Answer quality tracks the quality of source material more closely than anything else you control.
Writing the answers before the limits
Deciding what the agent must never attempt takes half an hour and prevents the category of incident that cancels projects. Deciding what it should answer takes far longer and matters less.
Optimising for containment
It is the easiest number to improve for the wrong reasons. Make escalation difficult and containment rises while satisfaction falls.
Translating the persona
An English persona rendered literally into Arabic reads as curt; an Arabic one rendered into English reads as ornate. Configure each language rather than translating one.
Launching broad
Narrow deployments that answer a handful of questions extremely well build trust. Broad ones that answer everything adequately lose it at the first confident mistake.
Leaving it unowned
Agents without a named owner drift — material goes stale, tone wanders, and nobody notices until quality has visibly fallen.
What the person on the other side experiences
It is worth thinking about this agent from the customer's side rather than from the configuration screen, because the two views produce different priorities.
From the configuration screen, success looks like coverage: more questions answered, fewer escalations, a rising containment figure. From the customer's side, success looks like something much simpler — they asked, they got a straight answer, and it did not take long. They do not know or care whether the answer came from an assistant or a person, and they will not credit you for automation. They will, however, notice immediately if the answer was wrong, evasive, or in the wrong language.
This is why the honest failure modes matter so much. "I do not have that information, let me get someone who does" costs you a containment point and buys you trust. A confident wrong answer saves the containment point and costs you the relationship, and you will not find out for weeks.
It is also why language handling is disproportionately important here. A customer who writes in Arabic and is answered in English has been told, without anyone saying it, that they are being served by a system built for someone else. In this market that is not a minor usability issue; it is a statement about who the service was designed for.
Typical impact
Setting it up well
Build the knowledge from real contact rather than from your website. Export a month of tickets and chat transcripts, categorise them, and count. The top twenty categories will cover most of the volume, and several will be questions your published content does not answer at all — which is why they became contact in the first place.
Write the escalation list before the answer list. Deciding what the agent must never attempt takes half an hour and prevents the category of incident that gets a deployment shut down. Deciding what it should answer takes longer and matters less.
Read transcripts weekly for the first month. Metrics tell you that something is wrong; transcripts tell you what. Every deployment discovers customers asking something nobody anticipated, phrased in a way nobody predicted.
Common questions
No, and deployments sold that way tend to fail. It removes the repetitive share of contact and leaves the judgement, the complaints and the complex cases — which is where trained people were always most valuable.
It escalates rather than attempting to resolve. Tone is detectable, and attempting to answer a complaint with policy text reliably makes it worse.
Where you connect the relevant system and the customer is identified. Otherwise it explains the process and hands over rather than guessing.
Yes — and that is often where most of the value is, because the alternative at 11pm is not a slower human but no answer at all.
Answers are grounded in your material and checked before sending. Where your material is silent it says so. The instruction to be careful is not the control; the grounding and the check are.
Yes. The reply follows the language of the most recent message, and mixed Arabic-English questions are handled rather than forced into one language.
Configure your own agents
One assistant per function, each with its own voice and limits.