CC5 Control Activities
Your risk assessment says an unauthorized production change could expose customer information. Someone suggests adding an approval. Someone else suggests another scanner. Both may help, but neither tells you whether the proposed control will interrupt the way that risk could actually occur.
Before buying another tool or adding another checkbox, walk through the work. Who can change production? Where does the change enter the system? Can the person requesting approval also bypass it? What happens when a release is urgent?
Those questions turn a general concern into a useful control-design conversation.
Start with the risk, not the control library
AICPA criterion CC5.1 concerns choosing and developing activities that reduce risks to a level management considers acceptable. Its points of focus connect those choices to risk assessment, the organization’s circumstances and relevant business processes. They also address the combination of manual and automated measures, prevention and detection, where controls operate, and separation of incompatible responsibilities.
Here is a practical way to work through that idea. Take one meaningful risk from your assessment and write down the event you want to prevent or detect. Then describe what someone would do differently because the control exists. If you cannot explain that connection, you may have selected a familiar control without solving the problem in front of you.
In our release example, an approval could stop an unreviewed change if the deployment process enforces it. An approval recorded in a ticket is less useful if an engineer can deploy directly and nobody examines that path. A scanner might identify certain coding weaknesses, but it would not necessarily establish that the release was authorized.
Ask the people who run the process to demonstrate the ordinary route and the exceptions. You want a design that works during a rushed release, a staff absence and a tooling outage, not only during the demonstration.
Be precise about what the control does
Imagine a control description that says, “Changes are reviewed.” You cannot tell who reviews them, what they examine or whether the review happens before deployment. The team may agree with the statement while each person means something different.
A more useful description could say that a qualified engineer other than the author reviews changes to the production application before the deployment workflow permits release. The reviewer checks the proposed change, relevant test results and security implications. Emergency changes follow a defined exception path, with restricted authority and a subsequent review within a time chosen for the business’s risk.
That is an illustrative design, not a prescribed SOC 2 control or universal deadline. Your environment may call for a different approach. The important question is whether the activity addresses the risk and can be performed consistently.
Also distinguish the control from the project that creates it. Configuring a deployment gate is implementation work. Requiring and enforcing the review on subsequent releases is the continuing control. A completed setup ticket does not show how later releases were handled.
Walk through a failure before calling the design finished
Suppose you add an automated approval gate. Now try a few uncomfortable questions. Can an administrator disable it without anyone noticing? Does it cover every production repository? Does the deployment identity accept releases from somewhere else? Does the reviewer have enough information to challenge a change?
You may decide to combine the gate with restricted administrative access and a review of exceptional changes. Each activity should have a distinct purpose. Adding three reviews of the same incomplete information does not necessarily produce a stronger result.
Keep the tradeoffs visible. A control that blocks routine work without a workable exception process may encourage people to find another route. An exception process that anyone can use without scrutiny may quietly become the normal route. Test both sides with the people responsible for the service.
The following fictional examples show how that conversation might differ as a company grows. Headcount is only a starting point. A small team operating a high-risk service may need more rigor than a much larger business with a simpler environment.
Five people: make the two-founder review meaningful
Consider a bootstrapped software company with two founders and three employees. Both founders build the product and administer production. They cannot create a large approval hierarchy, and pretending that they have one would not improve their security.
The founders identify the changes most likely to affect customer access and data. For those changes, one founder prepares the work and the other reviews the proposed behavior and test results before release. They configure their normal deployment route to require that review. They also identify who can override the gate and how the other founder will learn about an override.
For emergency work, they agree on the limited circumstances in which one founder can proceed, the records to retain and how promptly the other will review the result. They consider whether an external specialist is needed for changes neither founder can competently assess. Mutual approval has limits when both founders can alter the system or agree to bypass the process.
In Vanta’s risk register, they can keep the production-change risk, its owner and treatment decision together and link the relevant controls. This gives their next discussion a useful starting point: what risk did we address, what remains and what needs to change?
Keep a few real release records, the gate configuration and any exception reviews. Those records should show the reviewer’s decision and what was reviewed, not merely that someone clicked an approval button. The founders can then examine whether the design matches the way their team actually works.
At 100 people: design the handoffs as carefully as the approval
Now imagine a 100-person service provider. Engineering manages releases, customer support handles urgent requests, and a small security team advises on risk. A customer escalates a defect, support promises a quick fix, and engineering treats the request as an emergency. The control weakness may begin before anyone opens a pull request.
Bring those teams into the design discussion. Decide who can classify a change as urgent, what information the reviewer needs and when security expertise is necessary. Support should not promise a release that depends on skipping a required review. Engineering should not have to interpret an urgent commercial message as authorization to bypass the process.
The company might define a normal review route, an exception route and a recurring examination of exception use. It could also restrict who can alter the deployment gate. The frequency and depth of that examination should reflect the volume and significance of changes, rather than an arbitrary schedule copied from another company.
Vanta’s Action Tracker can help assign the work needed to implement a risk treatment plan, with owners and due dates. It also brings together controls linked to risk scenarios. The team still needs to confirm that the finished workflow enforces the intended review. Closing the implementation task is not the same as testing that workflow.
Useful evidence includes the agreed responsibilities, a walkthrough of both release routes, configuration records and examples of completed reviews. If the walkthrough exposes an unmonitored deployment path, record it and resolve the design gap rather than declaring success because the main route works.
At 1,000 people: make a shared control’s boundaries clear
At a 1,000-person business, a central platform team may provide the deployment service used by many product teams. The platform team owns the gate, while product leaders own the quality and authorization of their releases. Neither responsibility replaces the other.
Suppose the platform service enforces peer review, but an acquired business still deploys through a separate tool. A companywide statement that “all releases require approval” would overlook that boundary. Map which teams and production environments use the shared service, identify the exceptions and decide how those exceptions will be controlled.
The central team can provide evidence that the gate operates as designed. Product teams need evidence that appropriate reviewers assess their changes. Leaders should also know who can change the shared rules and how a change to those rules is evaluated. A technical setting cannot establish whether a reviewer understood the application or challenged an unsafe proposal.
Where Vanta is used, linked risks and controls can help make these dependencies visible during review. Avoid treating one green status as proof that every business unit is covered. Confirm the underlying scope and keep local exceptions connected to the broader risk decision.
The resulting evidence could include a coverage map, ownership records, shared-service configuration, local review samples and documented exceptions. You are building a defensible explanation of how the controls work together, not a larger collection of screenshots.
Know what you have designed, and what remains to be done
Before you move on, ask the control owner to explain the design without referring to a framework checklist. What can go wrong? How does this activity change that outcome? Who performs it? What information do they use? What would tell you that it failed?
Keep the answers concise enough to use. Record important assumptions, including dependencies on a supplier, a shared platform or another team. If management accepts a remaining risk, document the decision and its limits rather than quietly describing the risk as eliminated.
CC5.1 addresses control selection and development. CC5.2 considers supporting technology controls, while CC5.3 addresses putting expectations into practice through policies and procedures. Your next step is to make the chosen design operational, then evaluate whether it works. A well-written description is the beginning of that work, not the finish.
If your control list has grown but you still struggle to explain what it protects, Genius GRC can help you connect the risks, responsibilities and practical safeguards. Start with one process that matters to your customers and build from there. Talk to Genius GRC: https://geniusgrc.com/