Black Friday falls on 27 November this year, and anyone bringing in seasonal staff for customer service needs them in place well before that. Recruiting is rarely the problem. The problem is that the people who arrive have to answer customers within a couple of weeks, in an inbox where the easy questions are already gone. This article is about how to plan the training when time is short.
Why is training harder than it used to be?
Because self-service and AI have removed the easy tickets new agents used to learn on. Someone starting in customer service five years ago began with order status and opening hours, built confidence on a hundred easy tickets and took the hard ones after a month. Today those tickets are resolved in the help centre and the chat, and what reaches the inbox is the exceptions.
The industry sees the same thing. Gartner reports that 85 percent of service and support leaders are expanding agent responsibilities, despite expectations of large headcount reductions because of AI. So the role is getting harder, not easier, and that applies to someone who is there for eight weeks too.
The conclusion for training is that the goal cannot be for the new person to know everything. The goal is that they can find the right answer, recognise what kind of ticket it is, and know when to hand it on.
What do they need on day one, and what can wait?
Three things on day one, and the rest along the way. What has to be in place from the start:
- Finding the answer. Searching the knowledge base, understanding how an article is built and trusting that what it says is what applies.
- Recognising the ticket type. Which five to ten types make up most of the inflow, and which article belongs to which.
- Knowing when to hand over. A short, concrete list: complaints, compensation, threatening language, anything involving money above a set limit.
What can wait is product range knowledge, every function in the system, the exceptions in the policy, and the phone. The most common training mistake is to go through everything in a week of sessions and then drop the person into the queue. Nobody remembers a policy they have not used, and that time would have done more good in real tickets with someone beside them.
How do you shorten the time to the first real answer?
By making the knowledge base the working tool, not the training material. The difference is that the new person should use the articles while answering, not read them in advance. Training then becomes a matter of finding the right thing, which takes a day, rather than memorising, which takes weeks.
Four things that make it possible:
- Ready answers to the most common questions, reviewed and current. A knowledge base with short answers, steps and exceptions is what makes a new person useful in week one.
- AI drafts they review and change. The draft shows what a good answer looks like, and correcting it is the training. Gartner argues that AI in service should mainly be used to make employees better rather than to replace them, and onboarding is the clearest example.
- Limited rights at the start. In the Knowledge-Centered Service method, new colleagues hold a candidate role with limited rights while they are learning, and are licensed to publish freely only once they are judged competent; the Consortium for Service Innovation calls it the licensing model. The same principle works for answers to customers: let the first days' replies go via a colleague.
- A clear way out. Someone who gets stuck must be able to hand over without feeling they failed. Say so out loud on the first day.
How much coaching is needed?
More than you think, and in short sessions. The Consortium for Service Innovation reports that training combined with coaching gives an 86 percent increase in productivity, with an average return of six times the cost of the coaching, and that member companies see a direct correlation between the time spent coaching and what the method delivers.
In practice for a season: fifteen minutes every morning in the first week, twenty minutes at the end of the day to go through two tickets, and after that half an hour a week. The coaching should be about tickets the person actually had, not about policy in general. Appoint a named colleague per new person, and give that colleague fewer tickets in the schedule during those weeks. Coaching that has to fit on top of a full queue does not happen.
Which tickets should a new person not take?
The ones where a wrong answer binds the company or damages the customer relationship. Set the list in advance and keep it short:
| Ticket type | Why not | Who takes it |
|---|---|---|
| Complaints and faulty goods | The answers follow the law and are expensive to take back, as covered in the article on complaints | Experienced agent |
| Compensation and goodwill | Requires judgement and spending limits | Experienced agent or team lead |
| Unhappy customer threatening to escalate | The tone decides the outcome | Team lead |
| Questions about contracts, invoices and payment | Often system lookups and law | Whoever owns the subject |
| Anything about personal data and erasure | A wrong answer is a breach | A designated person |
Everything else a new person can take from week one, with support. Note that the list is not about difficulty but about consequence. A technically fiddly ticket is fine for a new person to get their teeth into; an easy ticket where the answer binds you for three years is not.
How do you know they are self-sufficient?
Through four numbers you already follow, split per person during the first weeks:
- Share of tickets handed on. Should fall week by week. If it levels out high, it is a ticket type without an article, not a person without ability.
- Time to first reply. Compare with the team average, not with zero.
- Share of AI drafts sent without changes. A steady rise means the person has started to trust the source material, which is what creates pace.
- Customer satisfaction per person, read together with the comments and never as a ranking tool between colleagues.
Bear in mind that the inflow in November is different from October. A new person who looks good on quiet tickets may need more support when the queue grows, which is an argument for training in good time rather than in the week before. How to work out how many people you actually need is in the article on staffing.
What to do
- Count backwards from 27 November. Two weeks of training and a week with support in the queue means the person should be in place by early November at the latest.
- Choose the five to ten ticket types they will own, and check that each type has a reviewed article before they start.
- Write the handover list and put it somewhere visible. Keep it under ten lines.
- Appoint a coach per new person and book the coaching in the schedule, with fewer tickets for the coach.
- Let them answer for real on day two, with a colleague reviewing the reply before it goes out during the first days.
- Follow the four numbers per person every week and fix articles rather than people when the numbers stall.
- Ask them what was missing. Someone who has been new for three weeks sees the gaps in the help centre more clearly than anyone else.
Common questions
How long does it take before seasonal staff are self-sufficient?
Count on two weeks to independent answers on the most common ticket types, and a few more weeks before the pace matches the team's. That assumes the answers are written and that someone is coaching; without that it takes twice as long and the quality is uneven. Shorter than that only works if you narrow the remit further, for example by letting them take a single ticket type.
Should new hires start in chat or in email?
Email and forms, for two reasons: the answer can be reviewed before it goes out, and the person can look things up calmly. Chat requires finding the right answer while the customer waits, which is harder, and the phone is hardest of all. Move them to chat when the share of tickets handed on has fallen and the person recognises the ticket types.
What if we do not have time to write articles before the season?
Prioritise the ten most common questions and write those, even if the rest has to wait. Ten reviewed answers are enough to make seasonal staff useful, while fifty half-finished ones are not. Also ask the experienced agents to save their best answers during the training period, so you have the material for next season without extra work.
Is it worth training staff who leave in January?
Yes, if you train against the knowledge base rather than against the people. What remains when the seasonal staff leave is the articles, the handover list and the coaching routine, and they make next season's training shorter. Keep in touch with the ones who did well, too; someone who already knows your systems and answers is considerably cheaper to bring back than a new hire.
Sources
- Gartner Survey Finds 85% of Service and Support Leaders are Expanding Human Agent Responsibilities Despite Expectations of Mass AI Layoffs — Gartner, 2026
- Customer Service Leaders Must Invest in AI to Enable, Not Replace, Employees — Gartner, 2024
- Technique 7.1: KCS Roles and the Licensing Model — Consortium for Service Innovation, read 2026
- Coaching — Consortium for Service Innovation, read 2026
Rickard Collander
Martin Carlsson