“Our company isn’t big — is spending this money worth it?”
The boss who asks this is usually already stuck in the dilemma. Dilemmas mean there is a pain point; it just is not known how to do the math.
No feature talk today. Only arithmetic. Three ledgers.
Ledger 1: the labor ledger
First, solve a word problem.
How many questions does your customer service answer repetitively every day? How many minutes does each one take on average? At the end of a day, how much time went into repeated work?
Example: three repeated questions, each asked 20 times a day, each reply takes one minute.
That is 60 minutes a day. 22 hours a month. Close to three working days, spent entirely saying the same lines.
And that counts only reply time. It does not count the hidden cost of getting tired of being asked, missing answers, or answering inconsistently.
A knowledge base does not reduce the number of questions. It reduces repeated answering.
Ledger 2: the risk ledger
This one is invisible on ordinary days; the bill arrives all at once when something goes wrong.
Two kinds of risk, both potentially expensive.
One is answering wrong. Shipping said wrong, delivery time said wrong, warranty scope said wrong. The customer screenshots it and sends it out; you owe an apology, a payout, and a correction of what you said.
The other is violating rules. E-commerce platforms have already switched customer service quality checks to full-volume inspection, fines per order — no more sampling; violations are fined order by order, 200 to 2,000 yuan per order. One superlative word, one over-promise, and that can earn you a fine.
This risk is not “might never happen”; it is that when it does happen, you cannot compute the floor of the loss.
What a knowledge base fixes is a very concrete part of this: writing down “the words that must not be said” and “the wording that must be said” clearly, so answering no longer depends on any individual’s memory.
Ledger 3: the asset ledger
The first two ledgers are about saving money; this one is about accumulating something.
In a small company that has run for five years, the most valuable thing is often not the equipment — it is the judgment criteria in the heads of the few veteran staff. Who can be given flexibility, in which situations to keep evidence, which words get you in trouble when said.
All of it lives in people’s heads right now. The people leave, and it is gone.
Organized into a knowledge base, it becomes a company asset — reusable, passable down, handable to the AI to do the work.
What you spend is expense; what you keep is an asset. Those two things must not be mixed in the same calculation.
When should you NOT build one?
I have talked people out of it too. Three situations:
Fewer than 10 questions, and they hardly ever repeat. Not worth it — keep them on a list for now.
You are the only one, you answer everything yourself, and the scale is not growing. Then no. You are the knowledge base.
The cost of a wrong answer is extremely low. Say, pure consultation and lead capture, no fulfillment obligations. Then it can wait.
Apart from those three, if the same kind of question reaches you a third time or more every day, it is time to start.
A cheap way to start
You do not need to buy a system first.
Start with one sheet: write down the 50 most-asked questions, then write an answer for each, one by one.
The act of writing exposes one thing: for some questions, you do not even have a unified wording internally.
That is the first, real layer of value of a knowledge base. The software comes later.
One last line
A knowledge base has never been a technology question. It is an arithmetic question.
Run the three ledgers clearly, and you will have the answer to whether to do it — on your own.
#enterprise-knowledge-base #knowledge-base-cost #AI-customer-service #SME-digitalization
Have you done this arithmetic? What did it come out to? Tell me in the comments and I will help you judge whether it is worth building.
Who we are: we help small and micro businesses build enterprise knowledge bases and land AI customer service. We turn the wording, rules and scripts scattered in employees’ heads into a base that the AI can look up and that is traceable; then we add red-line interception and handover-to-human judgment to the AI customer service.