AI Controls in Practice: Three Worked Examples, From Risk to Audit Evidence
Three worked SOC 2 examples: AI-written code, agent-led access reviews and agents changing customer data. Follow each risk through its control, audit test and evidence.

Three worked SOC 2 examples: AI-written code, agent-led access reviews and agents changing customer data. Follow each risk through its control, audit test and evidence.

AI now shows up in SOC 2 environments in three ways: supporting the people who operate the system, performing compliance controls, and acting directly in production. Each changes the risk differently, and so changes what a credible control looks like.
This guide works through one example of each, from the risk to the control that SOC 2 and enterprise customers would expect, to how an auditor tests it and the evidence that shows it worked.
The proposed 2027 SOC 2 criteria update would make AI risks more explicit through points of focus. These examples show how to address those risks in practice today.
These are illustrative control designs and audit procedures, not a mandatory checklist. The final approach depends on your system scope, risks and service commitments. For the broader criteria mapping, see SOC 2 controls and criteria for AI agents in production.
By how much the AI acts on its own. The more it acts without a person in between, the more the controls have to govern the AI itself, not just the people using it.
| Pattern | Who acts | Who owns the outcome | Where the risk concentrates |
|---|---|---|---|
| AI supports operations | A person, using AI output | The person | Over-trust and volume: people approve more, faster |
| AI performs a control | The AI, with a person overseeing | The control owner | Missing context and silent errors in the AI's judgement |
| AI acts in production | The AI, on live systems and data | The company, and its customers | Authorisation, accuracy and the AI straying from its purpose |
Worked example 1: AI writes code, a human reviews it. This is the simplest case, and it looks like nothing has changed. A person still owns each change, and a separate person still reviews and approves it before it reaches production.
But the risk profile has shifted. AI raises the volume of code and the size of each pull request, so reviewers are stretched. People can also over-trust AI output, while their experience tells them where humans make mistakes, not necessarily where AI does. AI can make confident assumptions with nothing behind them, and those are easy to wave through.
The control framework still fits, but it needs more weight on independent review and the automation around it.
| Risk | Control | Audit test | Evidence |
|---|---|---|---|
| AI-assisted changes reach production without real review | Branch protection requires approval from someone other than the account that pushed the change | Sample merged changes; confirm the approver differs from the author and approved before merge | Repository settings; pull request history with authors, approvers and timestamps |
| Reviewers over-trust AI output, and defects slip through | Review guidance covering AI failure modes; review time scaled to change size; limits on pull request size | Inspect guidance; for a sample, look for substantive review and time-to-approve that fits the change size | Review guideline; review comments; approval timing |
| Volume outpaces human review | Automated tests, code quality checks and security scanning as required checks before merge | Confirm checks are mandatory, not advisory; sample changes for passing checks | CI configuration; check results on sampled changes |
| Insecure code or invented dependencies | Dependency and vulnerability scanning, with findings tracked to resolution | Sample findings and trace remediation | Scan reports; remediation tickets |
An auditor will notice patterns that suggest rubber-stamping, such as a 2,000-line change approved within a minute. Relevant criteria include CC8.1 (change management) and CC7.1 (vulnerability detection). Approval timing is a signal to investigate, not proof on its own that a review was inadequate.
Worked example 2: AI performs a control. Take a user access review. An agent compares access lists against current staff to find people who have left but still have access. In some setups it reports what it finds to a responsible person, who acts on it. In others, it has permission to remove the access itself.
This can make the control stronger than the human version. AI doesn't get tired or skim a long list, and the owner can focus on the actions that matter. But AI makes mistakes people wouldn't. If someone changed roles or was deactivated and that never reached a system the agent can see, the agent may judge their existing access appropriate. People pick up that context from the office; AI only has what's recorded.
| Risk | Control | Audit test | Evidence |
|---|---|---|---|
| The agent lacks context, such as an unrecorded role change, and passes inappropriate access | HR records are the source of truth for joiners, movers and leavers, and feed the agent | Sample role changes and leavers in the period; confirm each appears correctly in the agent's review | HR change records; agent review outputs |
| The agent makes a wrong call | The control owner reviews the agent's findings, and a sample of its 'no issue' results, before signing off | Inspect sign-off; reperform the review for a sample of accounts | Review report; sign-off record |
| An agent with write access removes the wrong access or goes beyond its remit | Least-privilege permissions for the agent; every action logged and reviewed | Inspect the agent's permissions; trace logged actions to review findings | Agent role configuration; action logs |
| A change to the agent's prompt, tools or model silently weakens the control | The agent's configuration is under change management | Sample configuration changes for approval and testing | Change records for the agent |
The last row is the shift auditors care about most. When AI performs a control, the auditor needs comfort over the AI itself, as with any system: how it's configured, who can change it, and whether its outputs are complete and accurate.
The same applies when AI drafts board minutes, updates the risk register or performs a risk assessment. Fed the right context by the right people, it can be more thorough than a person. The risks are excessive detail that others struggle to work with, and missing a critical risk that people know about but nobody has written down. The control is a human owner who supplies the inputs and approves the result.
Relevant criteria include CC6.2 and CC6.3 (granting and removing access), CC4.1 (monitoring), CC3.2 (risk assessment) and CC8.1 (changes to the agent).
For the operating model, see how to build an agentic compliance program. For the context an agent needs, read Compliance Brain.
Worked example 3: an agent changes live customer data. Now an agent runs part of a call centre. It works as a data administrator: when a customer needs a record fixed or updated, it makes the change in live production data while talking to them. It can be fast and effective, and it's now part of your system, acting on your customers' data.
| Risk | Control | Audit test | Evidence |
|---|---|---|---|
| The wrong people direct changes through the agent | A clear authority matrix: who can request and approve each type of change; requesters verified before the agent acts | Inspect the matrix; sample changes and confirm each requester was authorised | Authority matrix; change logs showing requester and approval |
| The wrong people can use, configure or change the agent | Access to the agent restricted to approved users and isolated to each customer's environment; configuration access reviewed | Inspect access lists and isolation settings; review access review records | Access lists; environment configuration; access review sign-off |
| The agent misunderstands a request and makes the wrong change | Audit logging of every change with before and after values; human approval for high-impact changes; a way to reverse changes | Sample changes and trace each to the request, approval and log | Tamper-resistant logs; approval records; reversal records |
| Prompt injection, or the agent drifting from its purpose | Scope limited to approved tools, data and change types; guardrails on inputs; alerts on out-of-scope attempts | Inspect the allowed scope and guardrails; review alerts and how they were handled | Tool and permission configuration; guardrail rules; alert and incident records |
| Changes to highly sensitive data, or accounts customers haven't approved, causing a breach | Agent permissions follow data classification; customers opt in before the agent can act on their accounts | Inspect permission mapping against classification; sample customer approvals | Classification policy; permission matrix; customer approval records |
The agent also belongs in your system description, with its purpose, boundaries and the controls around it. Relevant criteria include CC6.1 and CC6.3 (access), CC7.2 to CC7.4 (monitoring and incident response), CC8.1 (changes to the agent), and Processing Integrity, Confidentiality and Privacy where they're in scope.
For choosing those additional categories, see the SOC 2 Trust Services Criteria explained.
When an agent works in production, the model provider behind it becomes a technology dependency to assess within your system scope. It's a third party to you and a fourth party to your customers. Customer data, privacy and confidentiality commitments need to be considered through that chain. How the provider is treated in the SOC 2 report depends on the service and reporting approach; it does not mean the provider's own controls are automatically tested by your auditor.
That brings the model provider into the controls you already run for critical vendors:
Enterprise customers increasingly ask about this directly. Clear answers about the model provider, and the controls around it, help their risk teams get comfortable. Our guide to getting through enterprise security reviews explains how this supports customer due diligence.
The following eight questions turn the worked examples into a practical review of your own environment. The answers should point to real owners, settings, decisions and records, not just a policy saying AI is governed.
Identify where AI supports a person, performs a control or acts directly in production. Record each material use case, its purpose and its boundaries. That makes the scope visible before you decide what controls and evidence are needed. The ISO 42001 startup guide gives a worked structure for an AI use case and agent register.
Name an accountable human owner. Show how that person supplies context, reviews findings and checks that the task worked, including a sample of results the agent marked as having no issue. Delegating execution to AI does not remove ownership of the outcome.
Show the agent's permissions, approved tools, data boundaries and authority to act. Trace those permissions to an approval by the appropriate owner. For customer-facing agents, include customer authorisation and isolation between customer environments.
Treat consequential changes to the agent as system changes. Keep approval and testing records that show the configuration was checked before use, including whether the change weakened a control or expanded the agent's authority.
Identify the sources the agent relies on and who keeps them current. In an access review, that includes joiner, mover and leaver records. Check actual changes against the agent's inputs and results, rather than assuming a connection to a system makes the context complete.
For consequential actions, retain a trace of the request, requester, approval and result. For live data changes, include before and after values and a practical reversal path. Auditors need evidence that can be traced back to what happened, not just an agent's summary. See collecting evidence with connectors, browser agents and a human in the loop.
Define the allowed scope, guardrails and alerts for out-of-scope attempts. Keep records of how alerts were reviewed and incidents handled. Monitoring should show that someone acted on the signal, not only that the system generated it.
Identify the providers behind each material AI use case. Check contractual data use, training, retention, processing and storage terms, and align permissions and disclosures with those commitments. Keep the due diligence and data terms available to support customer questions.
AI can do more and more of the work, but people must own the outcomes. A control is only as good as the context, scope and oversight around the AI performing it.

Former PwC manager, founder of AssuranceLab and former Sensiba partner and Chief Innovation Officer. Now founder of Throughline, focused on compliance and assurance for AI-native companies.
Beyond credentials, frameworks and pricing, ask who will do the work, how offshore teams operate, when auditors are available, whether controls fit your business and what happens when plans change.
Paul Wenham explains the proposed 2027 SOC 2 criteria, AI points of focus, governance and change management updates, transition effort and Throughline's approach to fees.
How certification bodies set ISO 42001 audit days under ISO/IEC 42006, how AI roles and in-scope headcount drive the price, and when scoping to engineering and product makes sense.
SOC 2 for startups, from planning to reporting: typical USD audit fees, timelines, internal team hours, AI-assisted readiness and keeping a first report on track.
SOC 2 for mid-sized companies: typical USD costs, timelines and team hours, plus existing controls, multiple products, regulatory drivers and auditor changes.
SOC 2 for enterprises: typical USD costs, timelines and team hours, plus ISO 27001 reuse, multi-entity scope, vendors, procurement and customer deadlines.
Typical SOC 2 costs and timelines for startups, mid-sized companies and enterprises, by report type. Compare planning, readiness, observation, fieldwork and reporting, plus related costs in USD.
Build an AI management system around your real use cases: an agent register, risk and impact assessments, shared responsibilities, governance policy and Statement of Applicability.
ISO 42001 isn't yet an enterprise requirement. What it covers, how it compares with SOC 2 and AIUC-1, and when it's worth it for an AI company.
How to cover production AI agents in SOC 2: governance, access, change management and monitoring, with indicative criteria mappings, customer responsibilities and practical audit expectations.
Australian companies need a SOC 2 report issued under the AICPA standards by a US-licensed CPA firm, delivered by a strong local team in Australian time zones. Here's how the provider options compare, and what to check.
Experienced CTOs don't choose a SOC 2 audit firm on price or speed. They look for tech stack expertise, a tailored approach, flexible scheduling, responsiveness and year-round partnership, because those factors decide how much time the audit really costs.
What Security, Availability, Confidentiality, Processing Integrity and Privacy each cover, how often they're included, and how to decide which belong in your SOC 2 report.
Which SOC 2 report should come first? Six client cases, practical decision tables and an illustrative 18-month cost comparison explain when to start with Type 1 and when to go straight to Type 2.
How long a SOC 2 Type 2 observation period should be, how to work backwards from a customer deadline, when auditors test, why reports run late and how much buffer to build in.
A category-by-category SOC 2 budget in USD, covering audit fees, readiness, platforms, penetration testing, consultants, internal assessments and operational costs. Learn what is required, what is optional, and what a quote can leave out.
Five signals of real quality in assurance and security, with practical guidance on building a compliance program that supports customer TPRM, repeatable enterprise reviews and revenue growth.
The person leading your SOC 2 audit is the most underrated factor in choosing an audit firm. Companies compare price, timelines and platform integrations, then barely ask who will actually do the work.
SOC 2 is a risk report, not a certificate. Understand how enterprise customers use it, why having a report is only base camp, and how to turn your audit into a trust asset for growing revenue.
What readiness should cover, when to involve your auditor, which gaps matter most, and how readiness, Type 1 and Type 2 can work as one continuous engagement. A practical guide to preparing for a report your customers can rely on.
CPA firm registration, signing licenses and peer review are the starting point, not proof of quality. A practical guide to checking credentials and asking the questions that matter before choosing your SOC 2 auditor.
Neither is better for everyone. The right choice depends on your customers, how you run security and where you are headed. For AI-native and AI-centric companies, SOC 2's flexible criteria usually make it the stronger starting point.
There's a basic psychology behind why most people don't like audits. Fix this, and people are surprisingly happy to do them again.
Enterprise buyers use SOC 2 as an initial qualifier, then look deeper. A quality report is what carries you through that assessment and turns your compliance investment into real trust.
Audit quality isn't service quality, and it isn't only professional standards. The real test is whether a reader of your report thinks: this tells me what I need to know.
Enterprise security teams review you in the background and rarely give feedback. What they look at beyond SOC 2, what counts as a genuine pen test, and how a quality report does the heavy lifting.
Compliance from first principles, the 95% problem, the compliance metadata AI needs, rethinking screenshots, spreadsheets and policies, and why audits should meet clients where they are.