An access control policy is a short written document that says how your company decides who gets access to which systems, how that access is changed and removed, and how you check it. For a small business it fits on two pages. There is a template at the end of this post that you can copy and adapt.
The template is a starting point, not legal or compliance advice. If you are preparing for SOC 2, ISO 27001 or HIPAA, your auditor, assessor or counsel decides whether your policy and your practice meet the standard.
What the policy is for
A policy does three jobs, and only one of them is for auditors.
It settles arguments in advance. Who can approve admin rights in the CRM? How fast must a leaver lose access? Can a contractor have a company Google account? Decide once, write it down, stop deciding every week.
It tells new admins what "normal" is. When the person who set everything up leaves, the policy is how their replacement knows the rules.
It is the yardstick for audits. SOC 2 auditors test whether you did what your own policy says. ISO 27001:2022's Annex A has a control titled "Access control" (5.15), and a written policy is the usual way to show it. Customers' security questionnaires nearly always ask "Do you have a documented access control policy?"
That last point cuts both ways. A policy promising monthly reviews that you run once a year is worse than one that promises yearly reviews and keeps to it. Write what you will actually do.
What auditors and frameworks look for
The three frameworks small companies meet most often ask for the same things in different words. None of them hands you a policy; each expects you to write your own and then follow it.
For SOC 2, the auditor reads your policy and then tests a sample against it: a few joiners, a few leavers, a review or two. The policy's job is to make promises the evidence can keep. Do you need SSO for SOC 2? covers what else the audit asks about access.
For ISO 27001, the 2022 Annex A has the "Access control" control (5.15) for the policy itself, "Access rights" (5.18) for granting, changing, reviewing and removing, and "Segregation of duties" (5.3) for conflicts such as approving your own access. Your certification body decides whether what you wrote and what you did meet them.
For HIPAA, the Security Rule's administrative safeguards in 45 CFR 164.308, as we read them on 27 October 2026, include "Access authorization" (policies and procedures for granting access to electronic protected health information), "Access establishment and modification", and "Termination procedures" for ending access. Each maps onto a clause of the template below. Access control under HIPAA and ISO 27001 goes through both frameworks for small teams.
The core clauses: granting, changing, removing, reviewing
Every access control policy, from a 10-person agency to a bank, covers the same four moments in an account's life. If you know the joiners, movers and leavers model, this is the same thing, plus a check.
Granting. Access is given for a reason, by someone allowed to approve it. The cleanest rule for a small company: each role has a standard set of tools and roles (an "access bundle" or role matrix; role-based access control explains how to build one), and anything beyond the standard needs the tool owner's approval. Default to the lowest role that works, which is the principle of least privilege in one line.
Changing. When someone changes job, they get the new role's access and lose the old role's. The losing part is the one that gets skipped, so the policy should say it explicitly.
Removing. Leavers lose access by a stated deadline. "On or before their last working day" for employees, "when the engagement ends" for contractors, "immediately" for involuntary departures. The policy should also say who is responsible, and that removal covers every tool, not just email. Our offboarding checklist has the full list of steps.
Reviewing. Someone checks, at a set frequency, that what people have still matches what they should have, and fixes the gaps. Quarterly for systems with customer data, money or code, yearly for the rest is a common choice. The user access review guide shows how to run one.
Admin and shared accounts
These two get their own clauses because they are where most real damage comes from.
Admin accounts. Limit owner and admin roles to named people, two per important system so there is always a backup. Require multi-factor authentication on every admin account. Where your identity system supports it, keep admin work on a separate account from daily work. Record each admin and the reason.
Shared accounts. Ban them by default, then allow exceptions with conditions: a named owner, the password kept in the company password manager and nowhere else, MFA where the tool supports it, and a password change whenever someone with access leaves. Shared accounts covers how to run that in practice.
Service accounts and API tokens. Integrations need access too. Give each one an owner, the narrowest permissions the tool offers, and a place on the review list.
Evidence
A policy with no records behind it is a promise. For each clause, decide what you will keep:
| Clause | Evidence to keep |
|---|---|
| Granting | The request and approval (ticket, email, chat message) for anything outside the standard bundle |
| Changing | The date of the role change and of the old access being removed |
| Removing | The leaver's date and, per tool, when access was removed; a screenshot or log entry |
| Reviewing | Exports, reviewer sign-off, findings and the changes made (access certification) |
| Admin accounts | The current admin list per system, with reasons |
| Shared accounts | The register of approved shared accounts, owners, and the date of the last password change |
Store it somewhere you will still find in a year: one folder per quarter is enough. Audit logs in your tools help, but whether a tool has one, and how long it keeps it, varies by tool and plan, so do not rely on them alone.
How long it should be
As short as you can make it while still answering the questions above. Two pages is plenty for a company under 50 people. Long policies copied from enterprise templates fail in a predictable way: they promise processes nobody runs, and the gap between paper and practice becomes the finding.
Have a named owner, review it once a year, and change it when your practice changes, not just when an auditor asks.
How to adapt the template
The numbers in brackets are choices, not standards. Here is how to think about the ones that matter most.
The removal deadline (5.1). "By the end of their last working day" is the common choice for planned departures, because it is a date everyone already knows. For an involuntary departure, removal at or before the conversation is the safer rule. Pick something you can meet on a Friday afternoon with the founder on holiday, because that is when it will be tested.
The mover window (4.2). Five working days gives the person time to hand over work without leaving old access in place for months. If your moves usually involve longer handovers, write a longer window, but write one.
The contractor inactivity rule (5.2). Contractors often stop working without anyone announcing an end date. A rule that removes access after a set period without activity catches the engagements that fade out instead of ending. It only works if someone looks at last-activity dates, so put that check in the quarterly review.
The review frequency (8.1). Quarterly for critical systems is common, and yearly for the rest. If you know you will not manage quarterly, write twice a year and do it.
Who the "ops/IT contact" is. In a 5-person company this is the founder, and clause 3.4 means someone else signs for the founder's own access. At 50 people it is usually a named operations or IT lead, with a backup, and system owners are real people in each team.
Template
Copy this into a document, replace everything in [brackets], and delete what does not apply. It is a plain-language starting point for a small business, not legal or compliance advice; have it checked by whoever will audit or advise you before relying on it.
[COMPANY NAME] ACCESS CONTROL POLICY
Version [1.0] - Approved by [name, title] on [date]
Policy owner: [name, title]. Reviewed at least once a year.
1. PURPOSE AND SCOPE
1.1 This policy sets how [Company] gives, changes, removes and reviews access to its systems and data.
1.2 It covers all employees, contractors and anyone else with an account in a system [Company] uses for work, and all such systems, including free tools that hold company data.
1.3 [Directory, e.g. Google Workspace] is the source of truth for who works at [Company]. A person who is not active there must not have access to any company system, except approved contractors listed in [location of contractor list].
2. PRINCIPLES
2.1 Least privilege: people get the lowest role that lets them do their current job, and no more.
2.2 Every account belongs to one named person. Exceptions are in 6.
2.3 Each system has a named owner who approves non-standard access. The list of systems and owners is kept in [location].
3. GRANTING ACCESS
3.1 Each role has a standard set of systems and roles, kept in [location of role matrix / access bundles].
3.2 New starters receive their role's standard access by [their first day].
3.3 Access beyond the standard needs approval from the system owner, recorded in [ticketing tool / channel / email].
3.4 Nobody approves their own access.
4. CHANGING ACCESS
4.1 When someone changes role, their manager tells [ops/IT contact] by [the effective date].
4.2 They receive the new role's access, and access they no longer need is removed within [5] working days.
5. REMOVING ACCESS
5.1 When someone leaves, access to all systems is removed by [the end of their last working day]. For involuntary departures, access is removed [before or at the time they are told].
5.2 Contractor access is removed when the engagement ends, or after [30] days without activity, whichever comes first.
5.3 [Ops/IT contact] is responsible for removal and records the date for each system.
5.4 Company data owned by the leaver is transferred to [their manager] before their accounts are deleted.
6. ADMIN, SHARED AND SERVICE ACCOUNTS
6.1 Owner and admin roles are limited to [two] named people per system. The list of admins is kept in [location].
6.2 All admin accounts use multi-factor authentication.
6.3 Shared accounts are not allowed unless approved by [policy owner] and listed in [register]. Each one has a named owner, its password is kept only in [company password manager], it uses multi-factor authentication where the system supports it, and its password is changed whenever someone with access leaves.
6.4 Integrations and API tokens have a named owner and the narrowest permissions available, and are included in reviews.
7. AUTHENTICATION
7.1 Multi-factor authentication is required for [the directory and all systems that support it / list of systems].
7.2 Passwords are unique per system and stored in [company password manager].
8. REVIEWS
8.1 [Ops/IT contact] runs an access review [quarterly] for [systems holding customer data, money or source code] and [yearly] for all other systems.
8.2 System owners or managers review and sign off each list. Findings are fixed within [10] working days.
8.3 Records of each review are kept for [3 years] in [location].
9. EXCEPTIONS AND BREACHES
9.1 Exceptions to this policy need written approval from [policy owner], with a reason and an end date.
9.2 Suspected misuse of access is reported to [contact] at once.
Adopting it this week
Writing the policy takes an hour. Making it true takes a little longer, and the order matters.
Start by filling in the brackets honestly, choosing numbers you already meet or can meet next month. Then build the two lists the policy points to: the systems with their owners (clause 2.3), and the admins per system (clause 6.1). Building the admin list will usually find someone to downgrade, which is the policy paying for itself on day one. Write the role matrix next, even if it covers only your two most common roles; clause 3.1 is empty without it.
Once the lists exist, have the policy approved by whoever runs the company, with a date on it, and tell the team where it lives. Put the first review in the calendar, three months out, with the owner's name on it. The first review is when you find out which promises were realistic, and the yearly policy review is when you fix the ones that were not.
Common mistakes
The most common one is copying a long enterprise policy and promising things nobody does, such as monthly reviews, a ticket for every access request, or a security committee that never meets. Close behind: no named owner, so the policy is never reviewed; no mention of contractors, who end up with the loosest access in the company; ignoring free tools, which still hold company data; and treating email as the whole of offboarding, which leaves every tool with its own password untouched.
Where SmashSSO fits
SmashSSO helps you do what clauses 3, 4, 5 and 8 promise. It reads your people from Google Workspace, read-only, and shows who has an account in each tool and with which role, matched by email. Access bundles by role give new starters the standard set, role changes and removals run from the same place, and anyone suspended in Google Workspace shows where they still have access. Steps that would add a paid seat wait for approval, and every change goes into an audit log that serves as evidence.
Changes run through a tool's member API where you have added a token, or through an agent in your own Chrome, which needs that browser open. It is not SSO, it does not write or store your policy, and it has no SOC 2 report of its own (security). It costs a flat fee from $199 a month, billed annually (pricing).