What are the risks of BPO

TL;DR

What Are the Risks of BPO? The Questions Buyers Should Ask

  • Five risk families cover most BPO failures: quality drift, attrition, data exposure, hidden cost, and now over-automation.
  • Attrition is the quiet one: at 35-38% industry turnover and ~30-50% of salary per replacement, a churning provider staffs your account with permanent novices.
  • AI adds a new contract clause - where the technology sits in your journey, and who decides. Klarna showed what cost-first automation does to quality.
  • Every risk on this list is inspectable before signature. The six questions below expose most of them.

Outsourcing has a marketing problem in both directions. Providers undersell the risks for obvious reasons; horror-story journalism oversells them for the same. What a buyer actually needs is neither reassurance nor terror. It is a map of where BPO deals genuinely fail, and a set of questions that make each failure visible before the signature rather than after.

The risks below are drawn from what the industry's own data keeps saying - and every one of them can be inspected in diligence.

What are the main risks of BPO?

Five families cover most failures. Quality drift: service that starts strong and erodes as attention moves to newer clients. Attrition: the provider's people churn, taking product knowledge and customer familiarity with them. Data risk: your customers' information now sits in someone else's systems, under laws like GDPR and South Africa's POPIA. Hidden costs: management overhead, rework and change requests that erode the headline saving. And over-automation: providers deploying AI on cost logic that damages the experience your brand carries. Each risk is inspectable before contract - if you ask.

Quality drift

Service that dazzles in month one and sags by month nine, as the A-team quietly moves to newer logos.

Ask: who exactly serves my account after go-live, and how senior are they in month six?

Attrition

The provider's people churn, and product knowledge and customer familiarity walk out with them.

Ask: what is your attrition by tenure band, on accounts like mine?

Data risk

Your customers' information now sits in someone else's systems, under GDPR and South Africa's POPIA.

Ask: how is my data segregated, and can I see the last audit?

Hidden cost

Management overhead, rework and change requests, quietly eroding the headline saving.

Ask: what did your last three clients spend beyond the rate card?

Over-automation

AI deployed on cost logic, in front of your customers, with your brand carrying the result.

Ask: where does AI sit in my journey, and who signs off changes?

Notice what is absent from that list: the destination itself. Country risk gets the headlines - and gets its own treatment for South Africa here - but deals rarely die of geography. They die of the five items above, which are provider-level, and which travel with the provider whether the floor is in Durban, Manila or Newcastle.

The quiet one: attrition

Is attrition the biggest hidden cost in outsourcing?

It is a strong candidate. Industry benchmarking puts contact centre agent turnover at 35-38% in recent years, and the cost of replacing one agent at 30-50% of an annual salary - roughly $20,800 in SQM's model - once hiring, training and lost productivity are counted. A provider running 40% attrition on your account is permanently staffing it with novices at your customers' expense, because every departing agent takes their product knowledge with them. Attrition rarely appears in a rate card, and it is the first number to request.

Attrition converts directly into every metric a buyer does track, which is what makes it the master risk. First-contact resolution falls, because novices resolve less. Handling times stretch. Complaints climb. Quality scores sag between calibrations. The provider's response is a fresh training cohort, which restarts the cycle. From the outside this presents as mysterious quality drift; from the inside it is simply the first-90-days problem running on a loop, billed monthly.

The diligence move is straightforward. Ask for attrition split by tenure band and by account, over twelve months. A provider proud of its retention will produce the data before you finish the sentence. One that offers a single blended annual figure, or talks about "industry norms", has answered a different question - and told you the answer to yours.

38% Global contact centre agent turnover in 2022 - the level at which an outsourced account is permanently staffed by novices SQM Group, 2024
30-50% Of an agent's annual salary - the cost of replacing one leaver, once hiring, training and lost productivity are counted (~$20,800) SQM Group, 2024
64% Customers who would prefer companies did not use AI in customer service - the number behind the new over-automation clause Gartner, July 2024

The diligence six

What questions should you ask a BPO provider before signing?

Six expose the most. What is your agent attrition, split by tenure band, and what happens in an agent's first 90 days? Which named team will serve my account, and how senior are they after go-live? How do you measure quality beyond average handling time, and can I see a calibration? How is my customers' data segregated, and how do you comply with POPIA and GDPR? Where and how do you use AI, and will you disclose changes? And what does your last client exit look like - who left you, and why? Fluent answers are informative; flinches more so.

  1. What is your agent attrition, split by tenure band - and what happens in an agent's first 90 days? Exposes whether your account will be staffed by veterans or a permanent training school.
  2. Which named team will serve my account, and how senior are they after go-live? Exposes quality drift before it happens - the demo team and the delivery team are rarely the same people.
  3. How do you measure quality beyond average handling time - and can I sit in on a calibration? Exposes whether quality means your customers' experience or the provider's dashboard.
  4. How is my customers' data segregated, and how do you comply with POPIA and GDPR? Exposes whether data protection is a walkthrough they give unprompted or a promise they improvise.
  5. Where and how do you use AI - and will you disclose changes on my account? Exposes whether the automation decision that carries your brand is contracted or just roadmapped.
  6. What does your last client exit look like - who left you, and why? Exposes more in one answer than a day of reference calls: how they fail, and how they talk about it.

A note on the data question, because it is the one with regulators attached. Outsourcing does not outsource accountability: under GDPR, and under South Africa's POPIA for processing there, your organisation remains answerable for what happens to customer data in a supplier's systems. The practical tests are segregation (whose other clients share the environment?), access control (who can see your data, from where?), breach notification (contractual clock, not statutory maximum), and evidence of audits rather than assertions of them. A mature provider treats this as a standard walkthrough. Treat improvisation as a finding.

The quality question hides a subtlety worth pulling out. Average handling time is the metric providers volunteer because it is the one they control. What predicts your customers' experience is first-contact resolution, quality calibration against your standards, and - the leading indicator underneath both - who is actually on the phones and whether the work fits them. Sit in on a calibration session before you sign. An hour of listening tells you what a quarter of dashboards will not.

Quality drift, viewed from the inside, is the first-90-days problem running on a loop - billed monthly.

The new clause: AI

How does AI change the risks of outsourcing?

It adds a new clause to every contract. Providers are under margin pressure to automate, and cost-first automation is a documented failure mode: Klarna claimed its AI replaced 700 agents in 2024, then reversed course in 2025 after conceding lower service quality. Gartner finds 64% of customers would prefer companies did not use AI in service at all. Buyers should ask where AI sits in their journey - behind agents as an assistant, or in front of customers as a wall - and contract for disclosure, human escape hatches and experience metrics, not just cost per contact.

The distinction to write into the contract is placement. AI behind the agent - suggested responses, instant knowledge retrieval - has strong evidence behind it: a 5,172-agent study found 13.8% productivity gains with better customer sentiment. AI in front of the customer, replacing agents on cost logic, is the pattern Klarna publicly retreated from. A provider should be able to tell you which pattern it runs, on your account specifically, and commit to telling you when that changes. If the answer is a roadmap slide rather than a clause, the decision is being made without you - and it is your brand on the line when 53% of customers say AI use would make them consider a competitor.

Risk, reframed

Run the five families back through and notice what they share. Quality drift is a people problem - attention and experience draining from your account. Attrition is a people problem by definition. Even the AI risk is a people problem: it is what happens when a provider stops investing in the humans your customers actually talk to.

Which suggests the sharpest single question in diligence is not on the standard list: how well does this provider know its own people? Operators who can show you, with data, how their agents' working preferences map to the queues they staff - who fits the emotionally heavy work, who is drifting, where the next resignations are building - are operators managing the root of every risk above rather than its symptoms. That visibility is what Sariio MAPS provides: each agent's preferences from a ten-minute survey retaken at least twice a year, mapped at team and account level. Some providers can already show you such a map; it is what we build for contact centres. Ask the ones who cannot what they are looking at instead.

The companion pieces: the South Africa-specific buyer's view is in is outsourcing to South Africa legit; the attrition mechanics are in why call centre agents really quit; the AI placement evidence is in how can AI improve customer experience.

Sources