How do you give the same answer whoever is asked?

A customer rings on Monday and asks whether delivery is included. Yes, says the person who picks up. They ring on Wednesday to confirm and a different colleague says it depends on the order size. Now the customer believes neither of you, and the one who said yes has made a promise the business has to decide whether to keep.

It happens in sales with pricing and terms, in support with policies, in operations with lead times, and it happens more as a business grows, because the number of people answering grows faster than the number of people who know. People were trained. The answer still lives in their heads, and heads differ.

Why answers drift

• Tenure. The person who's been there six years remembers the policy as it was, and the person who's been there six weeks remembers what they were told in week one.
• The exception that became the rule. Someone once made an allowance for a good customer, a colleague saw it, and now it's the answer for everyone.
• Three versions of the document. The price list on the drive, the one in the proposal template, and the one in someone's email signature, all slightly different.
• Pressure. On a call, with a customer waiting, a confident guess feels better than saying you'll check.

What inconsistency costs

The visible cost is the refund or the discount you honour because someone promised it. The larger cost is credibility: a customer who has heard two answers assumes the next one is also negotiable, and starts asking around for the best one. Inside the business, the cost is the hour a week your most senior person spends being the referee, because they've become the only reliable source of truth, and the backlog of questions that wait for them.

One approved source

Another reminder to check before answering won't do it. The check has to be faster than the guess. That means one approved source for each kind of answer: the current prices, the current terms, the current lead times, the policies as they stand. Owned by a named person, changed in one place, and reachable in seconds by anyone who is asked.

A system makes this practical. Ask the question in plain words and it finds the answer in the approved set and gives it back with the source quoted. The person still says it to the customer, in their own words, but what they say comes from the same place every time. We describe how that system is built in can you build a knowledge assistant on your own documents.

The answer carries its source

Every answer shows where it came from: the document, the section, a link. When a reply to a customer is drafted from it, the source travels with the draft, so the agent can check it before sending. When a colleague disagrees with an answer, they look at the source rather than arguing about memory. And when the source is wrong, which it sometimes will be, the correction goes into the document, and everyone's next answer is right.

A worked example

Go back to the delivery question. The sales lead decides the rule: delivery is included on orders above a set value, and charged at the published rate below it. That sentence goes into the price list, which is the one approved source for anything about pricing, and the old copies are removed. On Monday the customer rings and the person who picks up asks the system whether delivery is included. The answer comes back in a sentence, with the price list quoted under it. They read it out. On Wednesday the customer rings again and a different colleague does the same thing and says the same words. The customer hears one answer twice, which is what being a reliable business sounds like from the outside.

A month later the sales lead changes the threshold. One line in the price list is edited, and the next person asked gives the new answer. Nobody has to be told, because nobody was answering from memory. Compare that with the usual way: an email to all staff that half the team reads, and a customer told the old threshold in week three.

Where it goes wrong

• Nobody owns the answer. A source with no owner goes stale in a quarter, and people go back to memory.
• The source is the whole drive. If the system answers from every document you have, it will find the old price list as readily as the new one. The set has to be chosen.
• The answer arrives without its source. Then a wrong answer is an argument about the system instead of a fix to a document.
• The check is slower than the guess. If finding the answer takes longer than a confident guess, people guess. The system has to be faster than memory.

Where it shows up in daily work

In support, replies are drafted from the approved answers, with the customer's details filled in, for an agent to check and send. That's the model in how to use AI in support without annoying customers. In sales, the quote and the terms in a proposal come from the current price list and the current template, not from whichever version the rep last saved. In onboarding, the welcome message says what the contract says. And for a new hire, the source is the colleague who's always available, which shortens the weeks where they're guessing.

Taking the reading

Before anything is built, find out how inconsistent you are. Write down ten questions customers ask, from pricing to policy. Put each one to five people on your team, separately, and write down what they say. Count the different answers, and note which questions produced the most. Most businesses that do this are surprised, and the list of questions with the most spread is the list of documents to fix first. Keep the answers you collected; after the system is in, you'll run the same test and compare.

What to measure

Consistency of answers across the team is one of the four measures on Customer Experience, and the ten-questions reading is how you take it, before and after. Alongside it, watch requests handled without escalation, since many escalations are a question that reached the one person who knew. And watch the refunds and discounts given to honour something a colleague promised. Those should fall to the ones you meant to give.

What to do this week

1. Write down the ten questions customers ask most, across sales, support and operations.
2. Ask five people each of them, separately, and count the different answers.
3. For every question with more than one answer, decide the right one and who owns it.
4. Find where that answer is written down. If it's in three places, make it one.
5. Note the questions that reach your most senior person because nobody else is sure.

If you bring the ten questions and the spread to a call, we'll tell you what we'd put in the approved set first and how we'd get it in front of your team. The triage that gets each customer's question to the right person is covered in how to triage support requests automatically.