AI removes the routine volume that made outsourcing attractive, so the keep-outsourced or bring-in-house decision is worth reopening, and for most companies the answer is a mix rather than all of one.

Say a retailer outsourced its support to a BPO five years ago because hiring 80 agents was slow and expensive. Today AI handles half its contacts on its own. The retailer is paying a partner mostly for the harder half, and the original reason for the deal, raw scale, has shrunk. That is the question this article is about: when AI changes the economics, do you keep outsourcing customer service or bring it back in-house?

This is a different question from how to renegotiate an existing contract. We cover that separately in our notes on what AI automation means for your outsourced contact centre. Here the question is whether to keep the BPO relationship at all.

Why companies outsourced in the first place

The case for a BPO was rarely about quality. It was about three things. Scale, because a partner could add or cut agents faster than an internal team could hire or lay off. Cost, because labour in the BPO's locations was cheaper, and outsourcing has historically cut per-ticket costs by a wide margin. And focus, because running a contact centre is hard, and outsourcing let a company hand the staffing, training, and shift management to someone whose whole business is that.

Each of those reasons was strongest when support was mostly high-volume, repetitive work. That is exactly the work AI now changes.

What AI changes about the logic

AI takes the routine, repetitive contacts first. Gartner has projected that conversational AI would reduce contact centre agent labour costs by 80 billion dollars in 2026. When that volume goes, the outsourcing case weakens at each point.

Scale matters less, because AI absorbs the surges that used to need a flexible pool of agents. Cost matters less, because the cheap-labour advantage applied to volume that AI now handles for less, which narrows the cost gap between an outsourced and an in-house model. The remaining human contacts are the complex, judgment-heavy ones, and those are harder to hand to a partner who does not know your product deeply.

So AI does not automatically end outsourcing. It shrinks the part of the job the BPO was best at and grows the part where closeness to the business matters more.

The case for keeping a BPO

Outsourcing still earns its place for several reasons. A good BPO is adapting rather than standing still: as Nick Clark's Service Matters newsletter describes, the stronger operators are building real AI capability and treating human-plus-AI as their delivery model. Running the AI layer well is itself hard, and a partner who has done it across many clients may do it better and faster than you would alone. Surge capacity, out-of-hours coverage, and extra languages are still easier to buy than to build. And if support is not central to how you compete, handing the whole operation to a specialist still frees your attention.

The case for bringing it in-house

The pull the other way is real too. The contacts AI leaves for humans are the ones that need product knowledge, judgment, and brand voice, and an internal team is usually closer to all three. Insourcing is a real option, and some companies are choosing it as the cost gap narrows. The cost argument has weakened, so the saving that justified the distance is smaller. And the customer data, transcripts, and edge cases that flow through support are what train and improve your AI, which is harder to capture cleanly when the operation sits with a partner.

How to decide

The decision is rarely all-or-nothing. Most companies will land on a mix: AI for routine contacts, an internal team for the complex and brand-sensitive ones, and a BPO for surge, out-of-hours, and languages. To find your mix, start from the work, not the contract.

What to do before the decision: split last year's contacts into routine, complex, and surge or after-hours. Estimate how much of the routine layer AI will absorb. Then ask, for the volume that remains, where does product knowledge and brand voice matter most, and where is flexible capacity worth more than closeness. Routine goes to AI. The judgment work is the strongest candidate to bring in-house. Surge and coverage are the strongest to keep with a partner. That split is your answer.

Where this leaves the outsourcing question

AI does not settle the outsource-or-not question on its own. It removes the routine volume that made the original case simple and leaves you with a smaller, harder set of contacts to place. For most companies the practical answer is a deliberate mix, decided by looking at the work that remains rather than renewing the old arrangement out of habit.