Bring the Work Home
For three decades, enterprises scaled by sending work out. Agentic AI changes where the advantage lives — and leaves leadership teams with four decisions to make in the next four quarters.
For most of the last thirty years, the playbook for running a large enterprise's operations was settled. Standardize the process, document it, and move it to wherever the work could be done reliably at the lowest cost. The global delivery model that grew out of that playbook was one of the great management achievements of its era. It was built on a simple premise: labor was the expensive input, and it could be moved.
Agentic systems quietly invert that premise. When software can read the ticket, check the policy, draft the change, execute it within a defined scope and write down what it did, the scarce input is no longer hands. It is context: the records, rules, exceptions and judgment that sit inside the enterprise — next to the data, the controls and the people accountable for the outcome.
That is the heart of what InMag calls insourcing. Not a rejection of partners, and not a headcount exercise. A decision about where capability and accountability should live now that the economics have shifted.
The arbitrage is inverting
Outsourcing contracts were priced around effort: full-time equivalents, tickets closed, hours consumed. Agents break the link between effort and output. A process that once needed a large team following a runbook can increasingly be run by a small internal team directing agents and checking their work.
The organizations that capture that shift keep the gain. The ones that let it happen inside someone else's delivery center often see it arrive as a modest rate reduction at renewal — if they see it at all.
The scarce input is no longer hands. It is context — and context lives inside the enterprise.
What insourcing is not
It is not a wholesale repatriation of everything that was ever outsourced. Some work is better done by specialists, and some contracts carry expertise an enterprise could not rebuild quickly. It is also not a promise that agents will do everything; most programs that work start with narrow, well-instrumented processes and widen from there.
And it is not a technology purchase. Every insourcing program we have seen discussed seriously turns, within weeks, into a conversation about controls, contracts and careers.
Four decisions for the next four quarters
- Pick the first process deliberately. High volume, rules-heavy, well-understood exceptions, and an internal owner who wants it. Avoid the process everyone hates but no one owns.
- Own the evidence from day one. Every agent action should be scoped, attributable and logged in a form an auditor would accept. Autonomy grows only as fast as the evidence does.
- Read the contract calendar. Renewal dates, termination rights, knowledge-transfer clauses and pricing mechanisms set the pace of any insourcing program more than any model release does.
- Design the new roles before you need them. The people who know the process best are the natural orchestrators of the agents that will run it. Treat that as a career path, not a redundancy plan.
What to do on Monday
Put three names on a page: the process you would bring home first, the person who would own it, and the contract it currently sits in. If any of the three is hard to fill in, that is where the work starts.
Then find the peers who are a step ahead of you. The most useful insourcing lessons right now are not in vendor decks. They are in rooms where leaders can say what actually happened. That is what InMag is for.
This essay opens Issue 01, “The Insourcing Era.” Disagree, or have a case that complicates it? Pitch a response.
