The question usually arrives framed as a cost comparison: if the agent works from home, the provider isn't paying for a seat, and that should show up in the rate. It's a reasonable way to start and a poor way to decide. Infrastructure savings are the smallest and most visible part of this; what actually changes between the two models is where the risk lives, how the work is supervised, and which processes can run without losing control.
It's also worth separating three things that get called the same in everyday conversation. In Colombia, teletrabajo (Law 1221 of 2008), trabajo en casa (Law 2088 of 2021) and trabajo remoto (Law 2121 of 2021) are distinct legal figures. They differ on formalisation requirements, who provides and maintains the equipment, how the connectivity allowance is handled, and what applies on the right to disconnect. An outsourcing contract that says "agents work from home" without naming the figure is leaving open something that won't be settled later by email. This is general framing, not legal advice: the applicable figure is defined case by case with counsel.
What actually changes
Four things, and none of them is the floor lease.
The data perimeter. On site, control is physical and verifiable: restricted access, no paper, no phones on the operations floor, cameras, a managed network. At home, that perimeter is replaced by logical controls — virtual desktop, blocked downloads and clipboard, second factor, session monitoring — and by a promise about a physical environment nobody audits daily. It isn't that one is secure and the other isn't; it's that one is demonstrated with a visit and the other with technical evidence.
The speed of informal learning. A new agent on site solves half their questions by turning their chair. That channel doesn't exist at home and has to be built deliberately: a staffed floor chat, a supervisor reachable on a call, far better written working material. When it isn't built, the ramp stretches and escalation rises — the exact opposite of what saving on seats was meant to achieve.
Supervision. On site the supervisor sees the operation; at home they measure it. These are different activities, and the second one needs instruments: schedule adherence, quality monitoring by sample, real-time dashboards, short start and end of shift huddles. None of it is optional remotely and all of it costs.
Team stability. The home model widens the labour market beyond commuting distance from a site and hands the agent back their travel hours. That helps. It also removes the daily bond with the team, which is one of the things that holds tenure together. The net effect depends on how it's managed, not on the model, and it connects to what we covered in BPO attrition.
Which processes tolerate which model
The useful question isn't "remote yes or no" but "this process, remotely, with which controls?". Criteria that usually decide it:
- Data sensitivity. If the agent sees financial data, medical records, identity documents or anything the client classified as restricted, the remote model demands technical controls that have to be written into the contract, not assumed.
- Complexity and ambiguity. Processes requiring judgement and full of exceptions ask for proximity, at least during the ramp.
- Interdependence. If a case passes through three people before closing, remote coordination needs a tool and a ritual; on site it gets solved by walking over.
- Volume and standardisation. Repetitive, well documented processes measurable per case are the ones that work best from home.
- End client or regulator requirements. Sometimes the constraint isn't technical but contractual, and it surfaces late if nobody asked.
A remote model without written controls isn't a remote model: it's an on-site operation with the site removed.
What the home model requires and doesn't always get quoted
If the remote rate is cheaper for the same capacity under the same controls, something is being left out. The list of what remote work has to fund is concrete: equipment issued and managed by the provider — not the agent's personal computer — a backed-up data connection, technical support during operating hours, virtual desktop or session control licences, more supervision per agent during the ramp, and an answer to what happens when power or internet drops in the city where half the team lives. That last point isn't theoretical and belongs to the business continuity plan: a site has a generator and a second carrier; forty houses don't.
Hybrid, and its traps
The mixed model is often the right answer and also the worst executed. It works when the split follows process logic: ramp and certification on site, steady-state operation from home; restricted-data processes always on site; predictable peaks covered with additional remote headcount. It fails when the split is one of convenience and nobody defines who decides, how much notice a change of modality requires, and how a single operating culture is kept across two environments. A hybrid with no written rule ends up as two different operations sharing a name.
What to ask before signing
- Under what employment figure does the team work, and who provides and maintains the equipment?
- Which technical controls prevent information being extracted from a remote station, and how are they audited?
- What's the plan when power or connectivity fails, and how fast is capacity recovered?
- How many agents per supervisor remotely, and how does that number change during the ramp?
- Which part of the process requires being on site, and for what reason stated in terms of risk rather than habit?
- How is adherence measured without turning it into surveillance? This connects to what we covered on shrinkage and adherence.
And a warning about monitoring: there's a temptation to compensate for absence with software that permanently captures screen, keystrokes or camera. Beyond the personal data implications for the worker themselves, it's the fastest way to damage the climate and raise attrition. Remote supervision that works measures the output of the work, not presence in front of a screen.
How smartBPO works it
We decide per process, not by blanket policy. We classify each flow by data sensitivity, complexity and interdependence, and the modality follows from that; when a process can run from home, we issue and manage the equipment ourselves, with a controlled desktop and the extraction restrictions written into the technical annex rather than assumed. We run the ramp and certification with close support and only then move to distributed operation, with a supervisor available online through the whole shift. We write the power and connectivity contingency plan before go-live, stating what capacity recovers and how fast. And we measure by case outcome and audited quality, not by screen capture. We don't promise the remote model is cheaper: we show what controls it costs to sustain it, and let the comparison happen with both operations described at the same level of detail.