The Future of Managed Operations: From Outsourcing to Autonomous

The managed-operations industry — what we used to call BPO — was built on a single, simple insight: it is cheaper to have someone in another time zone do the back-office work of a Fortune 500 than to do it in-house. For thirty years, that arbitrage powered the industry. Companies offshored payroll, AP, claims processing, customer support, and a long tail of other repeatable workflows. Vendors built businesses with hundreds of thousands of seats.

That era is ending. Not slowly — quickly. And what replaces it is genuinely new.

This is our take on where managed operations is going, why the shift is faster than most people think, and how to think about whether to use a managed provider, build in-house, or do something in between.

A short history of why outsourcing worked

It’s worth being precise about what outsourcing actually did. It is not, as some critics suggest, simply “labor arbitrage.” It is three things bundled together:

1. Cost arbitrage. A trained operations professional in Manila or Krakow costs a fraction of one in San Francisco or London.
2. Process specialization. A vendor running 50 clients through the same workflow has more rep counts than any individual client and can build tooling, training, and quality processes accordingly.
3. Capacity flexibility. A vendor can absorb seasonal peaks that would require permanent headcount in-house.

The first of these is fading fast. The second and third are not — and they’re about to get reinvented.

What’s changing: the floor is dropping

The labor arbitrage that powered BPO depended on humans being meaningfully cheaper than software. In the last 24 months, that has flipped for a growing list of workflows.

A simple thought experiment. The going rate for an offshored AP processor is roughly $8–$15/hour fully-loaded. Assuming reasonable productivity, that’s maybe $0.50–$1.00 per invoice processed.

The same invoice can now be processed by an agent for $0.02 in compute, with human review only on the 10–20% of cases that warrant it. Even with overhead, the all-in cost is below $0.20 per invoice. The labor advantage that built the industry is gone for invoices, and it is going for the next ten workflows behind it.

This is not a slow shift. The workflows are tipping over one by one as the underlying technology gets reliable enough.

The hybrid model

What replaces pure-human BPO is not pure-software automation. It’s something more interesting: hybrid operations where agents handle the volume, humans handle the judgment, and the entire system is operated as a single workflow.

In a well-designed hybrid operation:

• Agents handle the 70–90% of work that fits inside known patterns.
• Reviewers handle the exceptions the agent flags — typically 10–20% of volume.
• Specialists handle the genuinely novel cases — typically 1–5% — and the outputs of their work feed back into the agent’s training and policy.

The economics of this model are radically different from either pure offshoring or pure software. The variable cost is dominated by exception handling, which scales with the quality of the agent rather than the volume of work. The fixed cost is dominated by the engineering team that builds and operates the agent, which is small relative to either a human BPO team or a software-license SaaS spend.

The result: hybrid operations are typically 60–80% cheaper than offshored human-only operations at scale, with better SLAs and far better data.

When to use managed ops vs in-house

Given this shift, when does it make sense to use a managed provider versus building yourself?

Build in-house when:

• The workflow is differentiating — i.e. a customer would notice if you were better at it than your competitors.
• You have, or can hire, the technical team to operate agents reliably.
• The data flowing through the workflow is competitively sensitive enough that third-party access is a meaningful concern.

Use a managed provider when:

• The workflow is non-differentiating but mission-critical (i.e. it has to work, but no customer will pay you more for doing it better than table stakes).
• You don’t have the technical team, and building one is not on your roadmap.
• You want capacity flexibility — e.g., your volume varies 5x seasonally.
• You want someone else to absorb the regulatory and compliance burden of the underlying technology.

This is roughly the same answer as a decade ago, but with the technical bar moved higher. “Build in-house” used to mean hiring an operations team. Today it means hiring a hybrid ops engineering team. Many companies that would have built in-house in 2015 should use a managed provider in 2025, simply because the talent for hybrid ops is rare.

What “autonomous operations” actually means

The phrase gets thrown around. Most uses of it are marketing. The honest version:

Autonomous operations is a workflow where:

1. The agent makes the bulk of decisions without human intervention.
2. The system measures its own performance against defined SLAs and quality thresholds.
3. The system flags its own exceptions and routes them to the right reviewer without human dispatching.
4. The system improves itself over time by feeding exception resolutions back into its policy.

It is not: an agent that does everything end-to-end with no humans. We have not seen that model work for any non-trivial enterprise workflow, and we are skeptical of anyone selling it. The interesting capability is not eliminating humans — it is minimizing their involvement to the cases where they add the most value.

What this means for the next decade

A few predictions we are willing to commit to:

The traditional BPO model collapses on itself. Pure-labor providers will either pivot to hybrid operations or shrink. The industry consolidation has already started.

The hybrid ops category gets very large. This is what we are building at Apison Hooks, and what we expect to see a wave of new entrants address. The companies that win will have deep operational expertise and deep technical expertise — the combination is rarer than either skill alone.

Enterprises will treat operations as a buyable service, not a built one. The default for non-differentiating workflows will be to consume them as a managed outcome — “process my invoices to a 99.5% accuracy SLA at this cost per unit” — rather than build them internally.

A new generation of operating systems for ops teams emerges. The current tools — ticketing systems, ERPs, workflow engines — were built for human-led operations. The next generation will be built for hybrid teams, with agents as first-class workers alongside humans.

The interesting question for any operations leader today is not whether this shift is happening. It is which side of it your team will be on three years from now. The companies still running 2015-vintage BPO contracts in 2028 will look as quaint as the companies that were still running paper invoices in 2015.

The technology to do better is already here. The discipline to operationalize it is the bottleneck — and it is exactly what we built our firm to provide.