A Look Back at Outsourcing in the Age of AI

Now that I am no longer working in outsourcing, it feels like the right moment to reflect on that industry (and IT consultancy more broadly). The timing couldn't be better -- we've entered the Age of Agentic AI. The similarities are so striking that it’s almost amusing.

The Promise of AI-Native Companies

In 2025, companies around the world began declaring themselves AI-native and AI-first. Token farms started consuming the majority of new investments. Then came the agents.

Why would anyone need an agent?

Because it lets you shift your position on the output–cost curve:

• Increase output without hiring, reducing costs per unit of work.

• Maintain output with fewer employees, reducing labor costs.

• Scale output quickly by investing capital rather than waiting through the long hiring cycle.

Hiring takes time. Building a team takes even longer. Now you can transfer money and get a ready-to-work "team." Sounds amazing.

But we've had this capability for decades. It's called outsourcing and outstaffing.

You can buy services that range from a black-box "click a button to build a website" solution to having a person working in your office.

The Reputation Gap

Unlike AI agents, which are currently surrounded by optimism, outsourcing has long been infamous -- especially among smaller companies.

Startup incubators and accelerators tell founders to build teams strictly in-house. Investors push management to get rid of contractors to improve valuation.

Looking for a job with only outsourcing experience can be an uphill battle. I've heard countless variations of:

"I liked your answers, but I don't think you'd be able to do this job because you don't have product experience."

Why does this perception persist?

Were there bad outsourcing companies that damaged the industry's reputation? Probably. But is that so different from today's "slope"?

There are bad engineers and there are bad models. We still assume that AI agents are useful. So logically, outsourcing must also have good players.

Life as a "Third Party"

Here's what I saw over more than a decade on the other side.

In theory, you're helping others succeed. In practice, you're often treated as a third party.

Remember investors encouraging companies to build only in-house teams? That mindset permeates daily operations.

Examples:

• Limited permissions: Sometimes it's forbidden to even message another employee on Slack to resolve an issue.

• Tools: Many licenses are reserved for employees.

• Production access: Connect to the DB, run a pipeline? No -- security concerns.

• KPIs and goal meetings: Not for contractors.

• Infrastructure cost data: "We appreciate your interest in optimization, but that’s internal information."

Trying to improve efficiency?

"You want to refactor that? Sounds like you're trying to scam us into paying your company more."

Coding standards inconsistently applied?

"Bob doesn't need to follow them -- his team is knows what they are doing."

Timesheets?

• Contractors: Mandatory. And increase granularity to 3-hour maximums.

• Employees: No need.

Team building?

"We'd love to invite you, but we can't."

Promotions?

"Senior roles are for employees only. Talk to your manager. Forgot to mention, Bob, the junior developer you onboarded three years ago now would be your tech lead. You did all the architectural work, so please help him."

Stock options? Bonuses?

"Your performance is outstanding even without them."

The Parallel with Agentic Teams

This is what the opportunity to get a "remote agentic team" often looked like from my perspective.

So I have questions:

• Is it really that different to grant build pipeline access to a third-party user versus a third-party agent?

• Could AI providers to raise prices just as an outsourcing company would?

• When an LLM suggests refactoring or using a new tool, is that a scam or an improvement?

• Is AGENTS.md fundamentally different from the project guidelines that were supposed to be written anyway?

And the bigger question:

If a company couldn't effectively utilize a team it could talk to in person, will it be able to utilize a team it can't?

Closing Thought

Outsourcing wasn't just about cost arbitrage. It was about elasticity -- about converting capital into capacity.

AI agents promise the same thing.

The real question is not whether agents will work. It's whether organizations that struggled with human third parties are structurally ready for non-human ones.