Comparison
When Not to Automate Calls
Some businesses should not do this, and some calls should never be handled by a system at any business. Both are worth stating plainly before anyone signs anything.
Two separate questions
- Whether your business is a poor fit overall
- Which specific calls must always reach a person
- Whether the volume justifies the effort
- Whether anyone will maintain it
- Whether your systems can support it
Businesses We Would Turn Away
Very low call volume is the clearest case. If your phone rings a handful of times a week and everyone who calls knows you personally, the configuration effort will never be repaid and voicemail is genuinely adequate.
Businesses where every call is genuinely unique are the second. If no two enquiries follow a similar pattern and the right answer always depends on judgement your team has never written down, there is nothing stable enough to automate.
The third is where nobody will own it. A knowledge base that is never updated becomes a source of confidently wrong answers, and a deployment with no internal owner degrades quietly until customers notice.
Calls That Should Always Reach a Person
| Situation | Why automation is wrong | Where this applies |
|---|---|---|
| Anything about an individual child | Safeguarding requires human judgement without exception | Preschools, paediatric clinics, tuition centres |
| Clinical urgency or symptoms | Triage is a clinical act, not a routing decision | All healthcare settings |
| Security incidents | Response cannot depend on classification by a system | Security services, property management |
| Allergen and dietary confirmation | An improvised answer can cause serious harm | Restaurants, catering |
| Callers in distress | Emotional situations need a person, whatever the subject | Everywhere |
| Grievances and personal HR matters | Confidentiality and judgement are inseparable here | Internal shared services |
| Regulated advice | Legal, medical, financial and visa advice require qualified people | Professional services |
These are not configuration preferences. They are exclusions we build in deliberately, and we will not remove them at a client’s request.
Signals to Wait
Reasons to postpone
- Your service list is about to change substantially
- You are mid-migration to a new booking system
- Nobody has capacity to approve content
- Call volume has not been measured at all
- The team has not been told it is happening
Reasons to proceed carefully
- Start with out-of-hours only
- Automate one enquiry type, not four
- Keep escalation thresholds deliberately low
- Review transcripts weekly at first
- Measure against a recorded baseline
Why We Publish This
A deployment that should not have happened costs the client money and costs us a reference. Identifying poor fits early is straightforwardly in both parties’ interest.
It also sets the tone for how the escalation rules get written. A supplier willing to say no to a client is more likely to say no to a caller when the situation requires it.
How those boundaries are implemented is described under emergency escalation and knowledge base management.
Questions People Ask
Have you actually turned business away?
Yes, where volume was too low or where the calls involved matters we will not automate.
Can we override the exclusions?
No. Child safeguarding, clinical triage, incidents and regulated advice are not configurable.
What if only some of our calls fit?
That is normal. Automate those and route the rest, which is how most deployments work.
How low is too low a volume?
There is no fixed threshold. Compare your monthly exposure against the cost and decide from there.
How do we know if we are a poor fit?
Tell us about your calls. We will say so if the answer is no.
Ask Whether You Should
Describe your calls and volume. If this is not right for your business, we will tell you and explain why.