You know your clients’ employees are using AI. What you may not know is which tools they’ve chosen, what they’re doing with them, and what information they’re sharing.
A Shadow AI audit helps you identify AI use that the client hasn’t reviewed or approved. You monitor activity for an agreed period, review the findings with your clients, and use them to build an AI policy for the client’s environment.
AI governance is becoming table stakes for MSPs. A Shadow AI audit gives you a repeatable way to bring that service to your client base. The audit’s findings help you recommend approved tools; decide what to monitor, redact, or block; and agree on what you’ll manage for each client. If you haven’t deployed Longwave internally, start there. Your complimentary Longwave NFR gives your team a way to learn the product and see your own AI usage before advising clients. As you begin rolling out to clients, choose whether to white-label the audit and all client reports under your own brand or present them with Longwave branding.
1. Plan your client rollout
First, plan the deployment across your client base, grouping clients according to your team’s capacity and each client’s requirements.
A good place to begin is with firms handling sensitive information, clients already asking you about AI, and clients who believe their employees barely use it. That last group (i.e., AI skeptics or deniers) still needs approved tools and clear rules for how employees can use AI, even if adoption is low.
2. Tell the client what the audit includes
Explain up front what the client will get: a review of employee AI usage, recommendations, and a written AI policy.
Choose an announcement that fits how you already work with the client. Longwave provides you with a full co-marketing kit, including templates for the emails below:
- Opt-out: “We’re starting this on [date]. Let us know if you’d rather not.” This fits clients where your existing agreement covers this kind of monitoring and you already roll out security tools without a separate approval meeting.
- Opt-in: “We’d like to include you. Reply yes and we’ll get started.” This fits clients who approve changes individually or need input from in-house IT, HR, or compliance.
Include all client users in the audit and agree on how staff will be informed. Decide which AI apps’ prompts to retain, who at the client or your MSP can view them, and how long they’ll be kept.
3. Set the audit window and book the workshop
Book the AI Policy Workshop before you deploy. The client should know when you’ll review the findings and make policy decisions.
Match the audit window to the client:
| Audit window | Good fit |
|---|---|
| 3–14 days | Larger teams or clients you know are using AI heavily. |
| 30 days | Most clients. A good default if you’re unsure. |
| 60 days | Small teams or light AI users who need more time to collect enough activity for a useful review. |
Leave yourself a few days before the workshop to review the report and prepare your recommendations.
4. Deploy Longwave in Silent mode
Create the client tenants and deploy the Browser Extension and Desktop Agent through your existing RMM or MDM, using Longwave’s deployment instructions. Keep each tenant in Silent mode throughout its audit.
In Silent mode, Longwave monitors activity without showing policy warnings, redacting prompts, or blocking employees’ AI tools. You can understand how people are working before applying new rules.
Open the tenant’s Users page to confirm that all users’ browsers and desktops are coming online. Check coverage for the client’s operating systems, browsers, and AI applications so you know what the findings will include.
5. Review the findings and prepare your recommendations
A few days before the workshop, generate the client’s Shadow AI Audit Report.
Use the activity data and any retained prompts to answer four questions:
- Which tools are in use? Compare observed usage with the client’s approved tools. Include applications beyond ChatGPT and Copilot.
- What work are employees doing with AI? Look for useful workflows and teams that could benefit from approved tools, training, or licenses.
- What sensitive information is being shared? Review relevant prompts and alerts in context. Bring examples that help the client decide what needs protection, such as client contracts, financial information, or source code.
- What should change? Prepare recommendations for approved tools, data rules, and any exceptions particular teams may need.
If the report shows little activity, check deployment coverage before drawing conclusions. If coverage is sound and usage is still light, say so. The client can still choose approved tools and establish rules before adoption grows.
Share the report, your recommendations, and your quote for the proposed ongoing service ahead of the workshop.
6. Build the policy with the client
Use the workshop to make decisions with the client. Bring a draft AI Acceptable Use Policy; Longwave provides standard policy documents, or you can upload your own.
Work through five decisions:
- Approved tools: Which AI tools should employees use, and do they need licenses?
- Data rules: What information can employees share? What should be redacted or kept out of AI tools entirely? Where are file uploads allowed?
- Roles and exceptions: Do any teams need different rules? Who approves a new tool or an exception?
- Accountability: Who owns the policy, and what happens when it isn’t followed?
- Review cadence: When will you review usage together and revisit the policy?
Document these decisions in one policy for the client’s environment. Employees should know which tools to use, what information they can share, and what to do when they need an exception.
7. Set the policy live
After the client approves the policy, distribute it to staff and configure Longwave’s controls to match. That can include monitoring usage, removing sensitive data from prompts, flagging or blocking unvetted tools, and controlling file uploads. If the client chooses a sanctioned AI tool, set it up and mark it Approved in Longwave.
This is where the orchestra starts to play together. The policy you configure in Longwave is the policy your client approved, and this becomes the experience employees have when they use AI.
Start with Longwave’s global defaults to establish common policy settings across clients, then use tenant-specific overrides where needed.
Customize the warning and block dialogs, too. Tell employees why they’re seeing the message, link to the client’s policy, and explain which approved tools to use or who to contact for help.
You can test the changes with a few users first to check how the controls affect their work. Notify staff, then switch the tenant to Active mode.
Deliver ongoing AI governance
Agree on what you’ll manage after the policy goes live: reviewing usage, handling requests for new tools, updating controls, and revisiting the policy with the client.
Longwave can automatically deliver client reports weekly or monthly. Use the AI Advisory Report to discuss what’s changed. Are employees using new tools? Are the controls interrupting legitimate work? Are there slow or manual tasks where an approved AI tool could help?
Decide how AI governance fits into your client offering: within your existing security stack, as a standalone service, or alongside an AI subscription as part of Managed AI Services. The audit gives you the information to recommend an approach that fits each client.