Outsource customer service or keep it in-house?

Originally published on the Supportifier blog.

The question usually comes when something has grown too big: tickets are growing faster than the team, evenings and weekends are unstaffed, or a campaign is launching in new markets. Then someone turns up with a quote from a contact centre, and the head of customer service has to say whether it is a good idea. This is about what decides.

Why do companies outsource customer service in 2026?

Not primarily to save money. In Deloitte's Global Outsourcing Survey 2024, 34 percent named cost as the main reason, against 70 percent in 2020. The most common reason was rising customer expectations. In practice, three things drive the question:

• Volume swings. A store with 300 tickets a week in March and 1,200 in December cannot hire for December.
• Opening hours fall short. Customers shop evenings and weekends; the team works office hours.
• Languages multiply. Nordic customers expect answers in their own language.

What do you hand over when you outsource?

You hand over the execution – and if you're not careful, the knowledge and customer insight too. Three things should always stay with you:

• The knowledge base. Reviewed answers in a system you own. The provider answers from it, not from its own documents. A structured help centre is a natural home for it.
• The exceptions. Compensation, policy exceptions and risky complaints need a clear hand-over route.
• The insight. Which questions come in and which answers are missing – every week, not in a quarterly report.

The Swedish industry association Kontakta's buyer guide shows that much of the work sits with the buyer before a provider is chosen: if you can't describe ticket types, volumes per hour and what a good answer is, you get a quote for the wrong thing.

What does it cost, and what does the part not in the quote cost?

Quotes are priced per hour, per ticket or per staffed seat. The per-ticket model gives the provider revenue per contact while you want fewer contacts, so write into the contract that recurring questions must be reported with proposals to remove them.

Then come the costs not in the quote: maintaining the knowledge base, training at every staff change, weekly quality checks, and a person at your end who owns the relationship. Without that person outsourcing doesn't get cheaper, only less visible.

How does AI change the calculation?

AI makes the question less about "who answers" and more about "who owns the knowledge the answers are built on". According to Gartner, 91 percent of customer service leaders feel pressure to implement AI. The simple, recurring tickets that made outsourcing profitable are the same ones self-service and AI take first – and AI only works on good source material. Deloitte reports that 83 percent of companies use AI in outsourced services, but only a quarter see lower cost or higher quality yet. Without a reviewed knowledge base you get neither good AI nor good outsourcing.

What does GDPR require when a provider handles your customers?

A data processing agreement, documented instructions and control over sub-processors. The provider is a processor; you remain the controller. Ask where agents sit and where ticket data is stored, which sub-processors are involved, and what happens to the data when the contract ends. Simplest is when the provider works in your own system – a shared inbox where tickets and history stay with you and access can be closed the day the collaboration ends.

When does in-house, outsourcing or hybrid fit?

Hybrid fits most companies with a few hundred tickets a week: you own the knowledge and the difficult tickets, a partner takes peaks, evenings or languages. Outsourcing doesn't remove the need for someone who owns customer service; the job shifts from answering to steering.

What to do

1. Count volume per hour and weekday for three months, per channel.
2. Sort ticket types into lookups, standard answers and exceptions. Only the first two can be handed out.
3. Write the knowledge base before you ask anyone.
4. Decide metrics in advance – response time, first-contact resolution, satisfaction, hand-over share – with weekly reporting.
5. Require in the contract that the knowledge is yours, that history can be exported, and that the DPA covers sub-processors.
6. Pilot a limited part, such as evenings and weekends in one channel, for three months.

Planning a self-service layer for customers as part of it? See the customer portal.