# How many hosting accounts hold company apps right now?

**Most companies do not know.** You can find the answer in an afternoon by checking records your company already holds, then asking teams to declare what those records miss. Count each hosting account, identify every app and owner, and record its data, access, and connections. Then secure, move, retire, or delete each result.

Last checked: 2026-09-03.

## Where should you look first?

Treat discovery as a bounded records check. Look for hosting accounts, deployed apps, custom domains, credentials, and deploy links that sit outside the approved path.

Copy this checklist into a ticket:

- [ ] Search expense reports and company card records for hosting, deployment, database, storage, and domain charges.
- [ ] Review the identity provider's application list for hosting services employees signed into.
- [ ] Check agent and connector installations visible through the identity provider or endpoint tools.
- [ ] Review DNS and certificate records for company-related domains that no team claims.
- [ ] Search outbound network logs for regular access to hosting providers.
- [ ] Search company code repositories for deploy hooks, tokens, workflows, or configuration tied to hosting platforms.
- [ ] Ask team leads which internal apps their teams use and where those apps run.
- [ ] Ask employees to declare apps under a time-limited amnesty, with no punishment for reporting them.
- [ ] Record personal accounts even when the company reimbursed the charge.
- [ ] Give each search source an owner and mark when the check is complete.

Do not wait for one system to produce a full answer. Card records find paid accounts but miss free plans. Identity records find sign-ins but may miss personal email accounts. Repository searches find deploy hooks but miss apps built without a company repository. Direct questions close those gaps.

## What should you record for every app?

The first record does not need a full threat model. It needs enough facts to identify the owner, place the app in a risk lane, and decide what happens next.

Copy this checklist for each app:

- [ ] Purpose
- [ ] Named business owner
- [ ] Current users
- [ ] Data the app reads, stores, or sends
- [ ] Outside services the app calls
- [ ] Whether anyone outside the company can open it
- [ ] Hosting provider, account, and runtime location
- [ ] Person or team that holds the hosting credential
- [ ] Company systems the app can read or change
- [ ] Last known use
- [ ] Next action and action owner

Record "unknown" when you cannot answer a field. An unknown data class or credential owner still helps triage. Do not hold the inventory open while you seek perfect detail.

Count both accounts and apps. One personal hosting account may hold several apps, while one app may depend on separate hosting, storage, and domain accounts. You need both views to remove personal credentials and preserve useful work.

## What should you do with each result?

Start with access and data, then judge business value. Microsoft's remediation flow is Discover, Assess, then Act. Its listed actions include sanctioning, blocking, applying file controls, monitoring, and adding an app to governance. See [Tutorial: Discover applications and shadow IT](https://learn.microsoft.com/en-us/entra/global-secure-access/tutorial-internet-access-application-discovery).

| What you find | First action | Next action |
| --- | --- | --- |
| Publicly reachable app holding company data | Remove public access first | Identify the owner, preserve needed data, and assess the app |
| Useful, low-risk internal app | Give it a named owner | Move it to the approved standard lane |
| App connected to a company system | Limit access while you assess it | Send it through the existing repository and security review |
| Duplicate app | Confirm which version the team uses | Export needed data and retire the duplicate |
| Abandoned app | Confirm no active business or records need | Delete the app, its data, credentials, and unused domain records |
| Restricted use | Stop access as company policy requires | Apply formal review or prohibit the use |

Preserve evidence needed for the decision. Record who changed access, where data moved, and who approved deletion. The aim is not to punish the builder. It is to put each app under a known owner and rule.

[Understanding the (In)Security of Vibe-Coded Applications](https://arxiv.org/abs/2606.23130) finds recurring flaws such as placeholder logic, unfiltered input, and exposed secrets. An approved runtime can contain these risks, while code review alone cannot, and better models reduce flaws but do not remove them.

## Why does an amnesty matter?

Punishing the first person who declares an app ends the inventory. Employees will infer that silence carries less personal risk than disclosure, and the remaining apps will stay hidden.

State the terms before asking. Reporting an app during the amnesty should start triage, not disciplinary action. This does not excuse harmful conduct or change records duties. It gives people a safe way to help IT find apps built before the approved route was clear.

You caught the problem in time. Use that chance to learn how teams built and hosted the apps, then fix the path that led them there.

## When is the discovery pass complete?

The pass is complete when each likely source has an owner, each result has a record, and each record has a next action. A raw list of domains or vendor names is not enough.

Track unresolved items separately: unknown owner, unknown data, unknown credential, or unreachable account. Give each item a named investigator and due date under your normal process. If an app cannot be assessed at once, reduce its access while the work continues.

Repeat the checks on a set schedule and after material changes to identity, endpoint, expense, or network systems. The first pass finds the backlog. Later passes test whether the approved path works.

## Why is the count not the main result?

The count gives leadership a clear starting point, but it does not prevent the pattern from returning. Every useful app needs somewhere approved to go. If the only choices are a personal account or a weeks-long review, low-risk tools will drift outside the inventory again.

Keep the existing review pipeline for connected, sensitive, public, and high-impact apps. Give private, low-risk tools a standard lane with company sign-in, named ownership, controlled connections, protected secrets, isolation, inventory, and logs from the moment they run.

Charming provides that standard lane while employees keep using the AI agents the company already approved. Use [the risk-lane guide](/which-ai-built-apps-need-code-review) to sort the apps you find, then put the rules into [the policy template](/ai-built-app-policy-template).

## FAQ

### Should IT disable every app as soon as it finds one?

No. Remove access first when an app is public and holds company data. For other apps, assess data, access, connections, use, and ownership before choosing the action.

### Should personal hosting accounts remain in use after discovery?

Company apps should move to an admin-managed account or approved platform. Remove personal credentials once you preserve needed apps and data.

### What if nobody claims an app?

Reduce access, identify its users and data, and check records needs. Delete it when no owner, active use, or retention duty remains.

### How often should the company repeat discovery?

Set a schedule that fits your risk and existing control cycle. Repeat the checks after changes that may create new blind spots.

### Does an amnesty mean ignoring policy violations?

No. It creates a safe route for declaration and triage. The company still applies its data, access, records, and review rules.

## What should you read next?

- [Decide which apps need code review](/which-ai-built-apps-need-code-review)
- [Write the company policy for AI-built apps](/ai-built-app-policy-template)
- [Respond when vibe coding has become shadow IT](/vibe-coding-shadow-it)
- [Set up governance for employee-built apps](/govern-ai-built-apps)
- [See how Charming supports IT teams](/work/it)
