Managed Cloud Operations vs. an In-House Platform Team
A practical decision guide for small businesses: separate operational cover from platform development, test the handoff, and compare managed, in-house and hybrid options.
# Managed Cloud Operations vs. an In-House Platform Team
If your small business needs reliable cloud support but has little recurring platform development work, start by evaluating managed operations. If several product teams need the same deployment tools, environments and infrastructure changes every week, consider building an in-house platform capability. When both are true, keep platform design in-house and buy clearly defined operational cover.
The useful distinction is between running today's services and building tools that change how teams deliver tomorrow. Hiring for one will not automatically solve the other.
1. Write down the work before choosing the team
Take a representative stretch of recent tickets, changes and incidents. Sort each item into one of three buckets:
- Run: respond to alerts, patch systems, check backups, run restores and maintain documented services.
- Build: create reusable deployment templates, provision repeatable environments, improve developer tooling and maintain shared infrastructure code.
- Decide: set recovery objectives, approve meaningful production risks, prioritise improvements and decide when to communicate an incident.
Record the systems affected, the hours when work can wait, and the knowledge needed to do it safely. Separate the amount of work from the coverage it requires. An uncommon incident can still need after-hours cover; a high-volume task that can wait until morning may be better solved by automation. The SRE Workbook chapter on on-call discusses designing coverage and escalation around the work responders actually handle.
2. Match the operating model to the bottleneck
Choose managed operations when the gap is repeatable work or coverage. This is a reasonable starting point when your environment is understood and someone can document what the provider is allowed to do. Ask which alerts are monitored, which actions can be taken without waiting for approval, and who owns incidents involving application code. An acknowledgement is not the same as a fix.
Build in-house when the gap is ongoing platform development. Consider this when multiple teams repeatedly face the same delivery problem and you can give a team time to maintain what it builds. Before hiring, name the first shared capability it will deliver, who will use it and who will maintain it. If you cannot identify repeated demand, solve one team’s immediate problem before building a platform for everyone.
Split the work when both gaps are real. An external team could handle infrastructure triage and approved recovery procedures while internal engineers maintain deployment tooling and diagnose product defects. Use a named incident lead and primary and backup contacts in both teams. The Cloud Adoption Framework's operating-model guidance distinguishes responsibilities for shared platform services from responsibilities for individual workloads. Use that distinction to write clear boundaries; it does not dictate who must employ the people doing the work.
3. Test the handoff on one service
Pick an important service and rehearse two scenarios without changing live traffic: a routine job fails, and a release causes visible errors outside business hours. For each scenario, write down who sees the signal, who can act, which changes require approval, where escalation goes if nobody answers, and who decides the service has recovered.
For example, in a hypothetical order-processing application, the operations team could investigate a storage alert and perform a documented mitigation. An internal application owner would handle a faulty release. If the runbook cannot say who takes over when both symptoms occur together, fix that handoff before committing to the model. Use the on-call handover guide to document that boundary, and the small-team recovery test plan to exercise the restore separately.
4. Compare equivalent coverage and plan the exit
Use the same service list, support hours and change responsibilities for each option. For managed services, ask what is included, what is charged separately, who supplies application expertise, and how transition support works. For an internal team, account for recruitment, leave and backup cover, onboarding, and time for improvements while incident work is ongoing. Include the time someone inside the business will still spend managing the arrangement.
Keep cloud accounts, infrastructure code, monitoring configuration and runbooks accessible to your organisation. Before expanding a managed arrangement, have an internal operator demonstrate that they can find and use those materials without relying on a provider-controlled login. The UK Government's Sourcing Playbook describes assessing in-house, external and hybrid delivery and planning transitions early. Those are useful planning prompts here, not requirements for a private business.
Start with the smallest scope that closes the gap you identified. At the end of the pilot, review which tasks were handled without ambiguity and which still depended on an unavailable owner. Use those findings to expand coverage, assign work internally or change where the boundary sits.
For related guidance, explore observability and reliability services and cloud foundations.