Your Custom Agent
The five standard agents cover the common functions. The sixth is whatever your organisation actually does that nobody else does — and it is frequently the one that matters most.
The five standard agents cover the common functions. The sixth is whatever your organisation actually does that nobody else does — and it is frequently the one that matters most.
What this agent is for
Every organisation has a function that does not map onto support, sales, HR, technical or finance. A hospital has appointments and preparation instructions. A university has admissions. A logistics company has customs documentation. A government body has eligibility and service applications.
These are usually the highest-volume contact category in the organisation, and they are almost never served by off-the-shelf assistant templates — which is why so many deployments cover the easy functions well and leave the actual business untouched.
The questions it handles
Whatever you are asked most
The starting point is always the same: what do people actually contact you about? The answer is rarely what the website emphasises.
Eligibility and qualification
"Do I qualify" is the most common question in government, education, insurance and finance, and it is answerable from published criteria.
Process and next steps
"What happens now" — the question people ask after every submission, and the one that generates the most repeat contact.
Documentation requirements
What someone needs to bring or upload, which prevents the rejected application and the second visit.
Appointments and scheduling
Where connected, availability and booking rather than a phone queue.
Sector-specific status
Whatever your equivalent of "where is my order" is — a case, a claim, an application, a shipment, a work order.
A typical conversation
- Start from your contact logNot from a template. A month of real contact, categorised and counted, tells you what this agent should be. Every deployment that started from a template discovered the same thing later, having wasted the interval.
- Define what it must never doBefore defining what it should do. For a specialised function the boundaries are usually clearer and more consequential than for a generic one.
- Give it your real materialProcedures, criteria, forms and guidance — the documents your own staff use, not the public summary of them.
- Set the tone for your audienceA hospital assistant and a logistics assistant should not sound the same, and neither should sound like a generic chatbot.
- Connect what it needsUsually one system holds the answer to most of the volume. Connect that first and widen later.
- Launch narrow and read everythingThe first month of transcripts will change your assumptions about what this agent is for. That is normal and worth planning for.
How this agent is configured
| Tone | Set to your sector. The register appropriate for a clinic is not the one for a freight forwarder, and both differ between Arabic and English. |
|---|---|
| Knowledge | Your own operational material — procedures, criteria, forms, guidance — rather than a template populated with generic content. |
| Escalation triggers | Defined for your risk profile. In clinical, financial and government contexts these are usually narrower than for commercial functions. |
| Never states | Determined by your regulatory position. This list matters more for specialised functions than for generic ones and should be written before launch. |
| Connections | Whichever systems hold the answers. Usually one holds most of the volume. |
| Languages | Arabic and English, with sector terminology handled correctly in both — which frequently requires deliberate work rather than translation. |
Where it hands over to a person
Because a custom agent handles the function that defines the organisation, its escalation boundaries deserve more thought than the standard ones rather than less. The generic rules still apply — complaints, distress, money, explicit requests — but there will be sector-specific ones that matter more.
In clinical settings, anything that could be read as advice about a condition. In government, anything that could be read as a determination of entitlement. In finance, anything resembling advice. In education, anything about an individual admission decision. These lines are usually obvious to people inside the organisation and invisible to anyone building from a template, which is exactly why they should be written down during deployment.
The general principle holds: the assistant explains what the published criteria say. It does not apply them to an individual and announce an outcome. That distinction protects both the organisation and the person asking.
What it should never do
Make a determination
Explaining eligibility criteria is helpful. Telling someone they qualify is a decision with consequences.
Give professional advice
Clinical, legal and financial advice are regulated acts, not conversational conveniences.
Commit on behalf of the organisation
Dates, outcomes, approvals and exceptions belong with people authorised to give them.
Answer outside its material
A specialised agent improvising is more dangerous than a generic one, because the questions carry more consequence.
Handle distress
Sector-specific distress — health worries, financial hardship, immigration anxiety — needs a person quickly.
Guess at sector terminology
Where a term has a precise meaning in your field, an approximate answer is a wrong answer.
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
Start from a month of real contact. Every organisation believes it knows what it is asked about, and almost every one is surprised by the actual distribution. The counting exercise takes a day and determines whether the agent is useful.
Involve the people who currently answer these questions. They know the phrasings, the edge cases and the questions that always turn out to be something else. That knowledge is not written down anywhere and is the difference between an agent that works and one that technically functions.
Write the never-list with whoever owns risk in your organisation. For a specialised function this is not a formality — it is the document that lets the deployment pass review, and producing it late is the most common reason a working pilot stalls before production.
Common questions
Whatever your organisation actually does — admissions, appointments, eligibility, claims, customs, applications, case status. Usually the highest-volume contact category and the one templates never cover.
It starts from your contact log rather than a template. The standard agents are a useful starting shape; a custom agent is shaped by what you are actually asked.
It can explain them. It should not apply them to an individual and announce an outcome — that distinction protects both you and the person asking, and it is configured explicitly.
The material and the boundaries take longer than the configuration. Organisations with current procedures and a clear risk position move quickly; the delay is usually internal agreement rather than technical work.
Yes. Large organisations frequently need one per service line, each with its own material, tone and limits.
Then the first step is counting your contact rather than configuring anything. A day spent categorising a month of real questions is worth more than any amount of specification.
Configure your own agents
One assistant per function, each with its own voice and limits.