Use case
A large share of support volume is the same handful of questions asked repeatedly. Handling those automatically frees the team for the enquiries that actually need judgement.
The visible symptom is rarely the cause. These are the underlying reasons we find most often.
Routine questions answered instantly and accurately, everything else reaching a person faster because the queue is shorter.
How we solve it
Four to eight weeks to production.
Which questions, how often, and how much time each consumes. This decides what is worth automating and what is not.
Weeks 1–2Retrieval over your documented answers so responses are accurate and traceable, rather than generated from general knowledge.
Weeks 2–5What the system will not attempt, and how it hands over to a person with full context. Decided before launch, not after an incident.
ConcurrentReview what it answered and how, and correct the source content when it gets something wrong.
OngoingCapabilities
They are annoyed by a bot that traps them in a loop and will not let them reach a person. Handled properly (accurate answers, an obvious route to a human, no pretence that it is a person) most customers prefer an instant answer to a queue.
A large share of support volume is the same handful of questions asked repeatedly. Handling those automatically frees the team for the enquiries that actually need judgement.
The same questions consume most of the team's day; Response times stretch during predictable peaks; Enquiries outside working hours wait until morning; and Good people spend their time on work that does not need them
Answers exist but are scattered and hard for customers to find; No triage between routine questions and genuine problems; and Volume concentrated into windows the team cannot staff for
Routine questions answered instantly and accurately, everything else reaching a person faster because the queue is shorter.
Which questions, how often, and how much time each consumes. This decides what is worth automating and what is not. (Weeks 1–2.)
Retrieval over your documented answers so responses are accurate and traceable, rather than generated from general knowledge. (Weeks 2–5.)
What the system will not attempt, and how it hands over to a person with full context. Decided before launch, not after an incident. (Concurrent.)
Connected
If the before state above reads like your operation, the next step is establishing which part of it is actually costing you. Describe it and we will tell you where automate customer support would and would not help.
