AegisAI Perspectives · 089
Choose email security for a small or mid-sized business by what it adds to Microsoft 365 or Google Workspace filtering and the admin time it costs weekly.

The best email security solutions for small and mid-sized businesses are the ones that add to Microsoft 365 or Google Workspace filtering8 and cost a small IT team little weekly admin time. AegisAI, for example, reads mail through the Microsoft 365 or Google Workspace API after delivery, with no MX record change and about five minutes to connect23. Whatever you shortlist, test it read-only on your own mail before you pay, and score it on the seven criteria below.
The built-in features in every Microsoft 365 subscription with cloud mailboxes prevent broad, volume-based, known email attacks. Defender for Office 365 Plan 1 is already included in Microsoft 365 Business Premium and Microsoft 365 E39. Plan 2 adds post-breach investigation, hunting, response and automation.
Google Workspace lists protection against suspicious attachments and scripts from untrusted senders under its advanced phishing and malware settings16. Gmail also puts a question mark next to the sender's name on unauthenticated mail.
Write down what your plan already includes before you compare anything. A paid layer has to catch the targeted messages that are left, the ones written to get one person to act.
Ask the vendor: which messages that our built-in filter delivered last month would your tool have flagged?
Ask first, because a routing change puts every inbound message behind the new tool. If a non-Microsoft service modifies inbound messages before delivery to Microsoft 365, Microsoft says to identify it as a trusted ARC sealer so those messages do not automatically fail authentication checks15. That is setup and testing a small team carries on its own.
A tool that connects through the platform API reads mail after delivery and leaves MX records where they are. Get the rollback plan in writing either way: which DNS records, connectors or rules change, and how long it takes to put them back.
Ask the vendor: what changes in our mail routing on day one, and what do we undo if we cancel?
Measure it in minutes per week during the trial, because it is the cost that keeps running after the contract is signed. On a team of one to three, the person who reviews quarantine also resets passwords and onboards new hires.
Count every queue review and release request the candidate creates, plus every policy edit. A release request for a real supplier invoice costs the same minutes as a real attack. For the full picture of running email security with nobody on a queue, read how to choose email security for a lean IT team with no SOC.
Ask the vendor: in a tenant our size, how many messages a week need a person to decide?
Test it on one convicted message, because removal decides how long a missed attack sits in inboxes. In Microsoft 365, post-breach response and automation come with Defender for Office 365 Plan 2, so check which plan you hold before you count on them.
Confirm that removal reaches every mailbox the message landed in, including forwards. Check that each removal is logged with who approved it and when.
Ask the vendor: who approves a removal, and which access does the tool need to make it?
Only if the verdict says what the email asked the reader to do and which signs gave it away. A generalist admin answers user questions and approves releases, so a bare score sends every question back to IT.
Pick five verdicts from the trial and hand them to someone outside security. If they cannot say what each sender wanted, the tool will cost you time on every release request.
Ask the vendor: show us a verdict on a payment change request with no link and no attachment.
Ask before you grant access, because the answers decide whether your own security review passes. Get in writing where message content is processed and stored, how long it is kept, and whether any model trains on it.
Match the API scopes the tool requests to the job it does. Monitoring needs read access. Removal needs write access, and you should be able to revoke it in your admin console.
Ask the vendor: which scopes do you request in monitoring mode, and which only when we turn on removal?
Run the candidate beside your current filtering on your own mailboxes, then turn on removal for one group. These steps fit a team of one to three.
Turn on the platform baseline first. In Microsoft 365, start with the Standard preset security policy for all users11; presets stay off until you turn them on. In Google Workspace, review the advanced phishing and malware protection settings.
Configure SPF, DKIM and DMARC records, in that order, for every custom domain (parked domains and subdomains included).
Grant access through a named admin role and write down how to revoke it. For Defender for Office 365, Microsoft recommends the Security Administrator role in Microsoft Entra for others who configure settings.
Connect the candidate read-only to every mailbox, beside your current filtering.
Log each message the candidate flags that your built-in filter delivered, and each one your filter caught that the candidate missed.
Record the admin minutes the candidate costs each week, release requests included.
Give five verdicts to a non-specialist and ask what each email wanted.
Grant write access for one group, remove one convicted message, and confirm every copy is gone before you widen removal.
Score every candidate on the same rows, with your trial evidence in the last column.
| Criterion | Why it matters at your size | How to test it |
|---|---|---|
| Mail flow risk | A routing fault reaches every mailbox, and one or two people fix it | Ask which MX records, connectors or ARC settings change |
| Admin time per week | The same people also run devices and accounts | Log minutes per week during the trial |
| Overlap with built-in filtering | Built-in features already stop broad, known attacks | Count flagged messages your filter delivered |
| Removal after delivery | A missed attack stays in inboxes until someone pulls it | Remove one convicted message and check every copy |
| Readable verdicts | A generalist answers user questions and approves releases | Have a non-specialist explain five verdicts |
| Data handling | Security reviews and customer questionnaires ask about mail data | Get storage, retention, model training and API scopes in writing |
| Access and rollback | Removal needs write access that someone grants and revokes | Name the admin role and test revoking it |
Illustrative example, figures are fictional. Harbor Lane Logistics (fictional) has 180 staff and 214 mailboxes on Google Workspace, with an IT team of two. It scores three options against the table: Gmail's built-in protection alone, a gateway that needs an MX record change, and an API-based tool that starts read-only. The gateway drops out on mail flow risk, because its rollback needs a DNS change during business hours. Over a two-week trial, the API-based tool flags 19 messages Gmail delivered, 4 of them requests to change a supplier's bank details. Its review queue takes 40 minutes a week. The office manager explains 5 of 5 verdicts correctly. The IT lead grants removal for the finance group first and pulls one convicted message from 11 mailboxes, forwards included. Written answers on data handling arrive before the contract.
Business email compromise is the second largest loss category in FBI complaint data. It caused $2,770,151,146 in reported losses in 2024, behind investment fraud3. IC3 logged 21,442 business email compromise complaints that year. Phishing and spoofing was the most reported crime type, with 193,407 complaints. All complaints together reached 859,532, with $16.6 billion in reported losses.
IC3 advises confirming any change to payment or account details with the real recipient through a separate channel5. Build that rule into the trial: count the payment change requests each candidate flags.
It costs the licence plus the admin hours plus the setup and rollback work. Work it out per year: the price per mailbox per month times your mailboxes times the months in a year, plus weekly admin hours times the weeks in a year times your hourly staff cost, plus setup hours.
Check whether you already pay for a layer. Defender for Office 365 Plan 1 comes with Microsoft 365 Business Premium and E3. A tool with a low licence and a daily quarantine queue costs a two-person team more than its price line shows.
A gateway sits in the mail path after an MX record change, so every inbound message passes through it first. An API-based tool connects to Microsoft 365 or Google Workspace and reads mail after delivery. With a gateway that modifies messages, Microsoft 365 needs it set as a trusted ARC sealer. With an API tool, removal depends on the write access an admin grants. For twelve checks on both approaches, read how to choose an email security solution in 12 checks.
AegisAI reads mail through the Microsoft 365 or Google Workspace API after delivery, with no change to mail routing (MX records)20. It runs in monitoring mode first, and removal needs write access your admin grants and can revoke. One bad email is pulled from every inbox it reached, forwards included.
You can run that trial on your own tenant through the demo page.
The best one is the tool that passes a read-only trial on your own mail. Score it on the seven criteria in the table, from mail flow risk to rollback, and on what it flags beyond your built-in filtering.
It covers a lot. Google Workspace protects against suspicious attachments and scripts from untrusted senders, and Gmail flags unauthenticated senders with a question mark. Test a candidate beside it and count what it flags that Gmail delivered.
Yes. Defender for Office 365 Plan 1 is included in Microsoft 365 Business Premium and E3. Turn on the Standard preset security policy first, since presets are off by default.
A named admin with a defined role. For Defender for Office 365, Microsoft recommends the Security Administrator role in Microsoft Entra for others who configure settings. Write down how to revoke any write access you grant a vendor.
Set it up anyway. Microsoft says to configure SPF, DKIM and DMARC, in that order, for all custom domains. NIST SP 800-177 Rev. 1 says its email guideline is useful for small or medium sized organizations.
Confirm the change with the real recipient through a separate channel before acting. Business email compromise caused $2,770,151,146 in reported losses in 2024.