Before a 2027 renewal, ask your email security vendor ten questions: what got through, how it judges authenticated senders, whether verdicts show reasoning, how much SOC work it leaves, which modules you use, what switching costs, what the bundle measures, whether terms match the roadmap, how it catches unknown attacks, and whether it will test on your own mail.

Most renewals run on the incumbent's dashboard, and nobody asks what the tool missed. Each question below forces evidence from your own environment.

What questions should you ask your email security vendor before a 2027 renewal?

Ask questions that a demo cannot answer. They work for the incumbent you know well and for any challenger you weigh against it. If you are renewing, you already hold most of the evidence: missed incidents, support tickets, analyst hours, and a full contract term of history. These ten questions force evidence from your own environment, and each one ends with what a good answer looks like.

1. What got through in the last 30 days, and how would you know?

Every dashboard reports what it caught. Ask for what it delivered: user-reported phishing that reached inboxes, messages pulled back out of inboxes, and incidents your SOC opened from email. If nobody can produce that number, your current tool is grading its own homework.

What a good answer looks like: A count of missed threats from your own tenant, with examples, produced within days.

2. How does the tool handle first-contact and authenticated senders?

SPF, DKIM, and DMARC confirm which domain sent a message. They say nothing about intent. Ask the vendor to walk through a verdict on an email from a real, compromised supplier account, inside an existing thread, requesting a bank detail change. If you are renewing, start with the last supplier-account email that reached an inbox and ask your current vendor why it was delivered.

What a good answer looks like: A verdict walk-through on a compromised-supplier email that names the signals the tool weighed beyond SPF, DKIM, and DMARC.

3. Can your analysts see why a verdict was made?

A risk score tells an analyst to look. Reasoning tells the analyst what to do. Ask to see the evidence behind a block and behind a delivery: what the tool checked, what it found, and why it decided. Explainable verdicts let you release, escalate, or brief finance in minutes. For the tool you run now, ask your analysts how often they open a second console to work out what a verdict means.

What a good answer looks like: A plain-language explanation on every block and every delivery that an analyst can act on without opening a second console.

4. How much SOC work does the tool leave behind?

Count the user-reported phishing queue and the quarantine release requests. Ask what share of reports the tool closes without an analyst and how long employees wait for a quarantine release. Detection that creates a queue moves the cost onto your team. For the tool you run now, pull 12 months of report and release volumes before the renewal call, so the vendor answers with your numbers.

What a good answer looks like: A measured share of user reports closed automatically, with release decisions in minutes and an audit trail for each one.

5. How much of the suite do you actually use?

List every module on the invoice and mark which ones are turned on and tuned. Check overlap with what your Microsoft or Google licensing already includes.

What a good answer looks like: A module-by-module usage report from the vendor, and a price that drops for the modules you leave off.

6. What does deployment and switching cost?

An MX record change means change windows, routing rules, DNS work, and a rollback plan. An API connection deploys without touching mail flow and can run side by side with your current tool. Ask whether you can run a challenger in parallel for 30 days before you decide. Incumbents price renewals knowing that switching feels expensive, so get a real number before you negotiate.

What a good answer looks like: An API connection with no MX change, and a written offer to run a 30-day parallel test before you commit.

7. What does the bundle optimize for?

Ask which metric the vendor reports to your board. Training completion and click rates measure employee behavior. Attacks stopped measure protection. Buy the bundle only if both halves meet the bar you would set for them separately.

What a good answer looks like: Detection results reported separately from training metrics, with attacks stopped as the headline number.

8. Does the contract length match the vendor's roadmap and ownership?

Ask for the roadmap in writing, a termination right on change of control, and price protection on renewal. If the vendor will not offer those terms on a three-year deal, sign for one year. If you are renewing, compare the roadmap you were shown at the last signing with what shipped. That record is the best predictor of the next one.

What a good answer looks like: A written roadmap, a change-of-control termination right, and a renewal price cap on any multi-year term.

9. How does the tool catch an attack nobody has written a rule for?

Attackers change their messages every time a defense learns to block them, and a three-year contract covers attacks that do not exist yet. Ask the vendor to show a recent catch that no rule, signature, or policy described in advance, and to explain what the tool needed from your team before it could make that call. A tool that improves only when someone writes a new rule falls further behind each quarter. If you are renewing, ask your current vendor which of last year's incidents it caught on first sight and which it caught only after someone wrote a rule.

What a good answer looks like: A real verdict on a new attack pattern, with the reasoning shown, reached without a rule or policy written by your team.

10. Can you test with your own mail before you sign?

Request an independent, read-only assessment on your tenant that scans recent mail your current stack already delivered. Your own mail is the only benchmark that matters.

What a good answer looks like: A read-only assessment on your tenant, with written findings and evidence for each flagged message, delivered before your renewal date.

Why are 2027 email security renewals different?

A 2027 renewal is a multi-year bet on attacks that improve every quarter. Most contracts up for renewal were signed before AI made targeted attacks cheap, and the stacks behind them were tuned for bulk phishing, malicious links, and known-bad attachments. Older tools still handle those well. Those attacks stopped being the ones that cost companies the most.

The attacks doing damage now look like ordinary business email from people you already trust, such as a supplier's real account asking to update bank details in a thread your finance team has used for a year. The FBI's 2025 Internet Crime Report recorded nearly $3.05 billion in business email compromise (BEC) losses across 24,768 complaints, the second-highest loss category after investment fraud.

AI changed the economics of targeted attacks. Researching a company, mapping its suppliers, and writing in the right voice used to take an attacker weeks per target. Language models now do that work in a fraction of the time, for every name on a list. The messages that cost the most come from real suppliers, inside real threads, from senders that pass SPF, DKIM, and DMARC, often with no link or attachment at all. Tools built on sender reputation, payload scanning, and rules read signals these messages do not carry.

Vendors and teams are under pressure too. Acquisitions are consolidating the market, and new ownership can change a vendor's roadmap partway through your contract. Training vendors are bundling email detection with awareness training and AI governance, so a detection decision can become a platform decision. The people running the renewal usually also work the phishing queue, which makes renewing what is already installed the easy path. Expect the quote to favor a three-year term, which keeps you on the same product through 2030.

Evaluate against where attacks are going. Ask every vendor to show how it handles four patterns that are growing now:

• Vendor account takeover: a supplier's real mailbox, compromised, sending a payment change inside an existing thread.

• Conversation hijacking: an attacker replying into a stolen thread, with the history that makes the request look routine.

• Multi-step and multi-channel lures: a harmless first email that sets up a follow-up call, text, or shared document.

• Attacks aimed at AI assistants: instructions hidden in an email for the AI assistant or agent that reads the inbox on an employee's behalf.

Last year's phishing test shows how a tool handles last year's attacks. These four patterns show whether it still works in 2029.

What are the main types of email security, and where does each fall short?

Email security approaches fall into four categories, and each one is strong at something real. Most organizations run the built-in protection in Microsoft 365 or Google Workspace as their first layer, and Gartner recommends at least two layers of protection, so each approach below usually sits on top of that baseline. The useful question is which failure modes each approach leaves for your team to absorb.

No single category wins every column. A sound 2027 decision maps your actual losses and your SOC's actual workload to the column that matters most.

How does agentic email security work?

Agentic email security is an approach to email protection in which AI agents investigate each suspicious message the way a human analyst would, examining intent, context, and linked content, and recording the reasoning behind every verdict. The agents do that work across an entire organization's mail volume at machine speed.

The architecture matters to a buyer because it decides what a tool can read and what it costs your team to run. Here is how Ocean builds it.

• Screening. Frontier AI models check every email against a broad set of signals: who sent it and from where, how it was authenticated, what it says and what it asks for, what it links to or carries, and how it compares with the organization's normal patterns.

• Investigation. Anything suspicious goes to purpose-built agents that work it the way an analyst would. They check the intent and context: the relationship and vendor history, read the thread, weigh intent, examine the sending infrastructure, and follow links and open attachments when a message has them.

• Deep analysis. Ray, Ocean's central intelligence engine, handles the deepest analysis and brings the agents' findings together into a verdict.

• Reasoning. Every verdict comes with its reasoning: what was checked, what was found, and why the agents decided. An analyst reads the case and acts.

The biggest difference is where the knowledge comes from. Ocean learns from the organization's own existing email and business context: who pays whom, which suppliers send from which systems, how a normal request in a given thread reads. Ocean starts working without anyone writing policies or tuning rules, and it can recognize an attack type before anyone has defined it or written a policy for it. Over a three-year term, that matters most, because the attacks of 2029 have not been defined yet.

Ocean complements the protection already built into Microsoft 365 and Google Workspace. Native protection clears spam, malware, and known phishing at scale. Ocean focuses on the highly contextual, sophisticated attacks that appear legitimate and are crafted to bypass tools built on rules and policies, and it investigates each one with the organization's own context. The two layers work better together: native protection handles volume, and Ocean handles the messages that need an investigation. The model also changes the cost of ownership. Ocean connects by read-only API in minutes, with no MX change, and runs alongside your current tool. There are no policies to write or maintain. The same agents triage each user-reported phishing email in about 46 seconds on average and review quarantine release requests automatically, which returns analyst hours to the SOC.

Hold any vendor in the category to the same test. Ask to see the reasoning on a real verdict from your own mail, ask what the tool needed from your team before it started working, and ask how it handled an attack nobody had written a rule for. For Ocean, the fastest way to answer question ten is a free email health assessment: read-only API access, no MX change, and findings from your own mail before your renewal date.

Frequently asked questions

When should you start evaluating email security vendors before a renewal?

Start at least 90 days before the renewal date. That window covers a 30-day side-by-side test on your own mail, a review of the findings with your SOC, and time to negotiate terms. Teams that start 30 days out usually renew by default, because there is no time left to test an alternative or bring the results into pricing talks.

What is agentic email security?

Agentic email security is an approach to email protection in which AI agents investigate each suspicious message the way a human analyst would. The agents examine intent, sender history, the conversation thread, and linked content, then record the reasoning behind every verdict. Agentic tools are built to catch text-only and authenticated attacks that reputation and signature checks miss, and to close routine SOC work such as user-reported phishing triage.

Should you add a second layer of email security on top of Microsoft 365 or Google Workspace?

Yes. Gartner recommends a layered approach to email security, with at least two layers of protection. Microsoft 365 and Google Workspace provide the first layer: they filter spam, malware, and known phishing at scale, and most organizations already run them. A second layer complements that protection by focusing on highly contextual, sophisticated attacks that appear legitimate and are crafted to bypass tools built on rules and policies, such as a payment change from a real supplier's account inside a thread your finance team already uses. API-based tools add that layer without an MX change, and a 30-day read-only assessment on your own tenant shows what the second layer adds.

Do SPF, DKIM, and DMARC stop business email compromise?

SPF, DKIM, and DMARC stop domain spoofing. They confirm that a message came from the domain it claims. They do nothing against business email compromise sent from a real, compromised account or a lookalike domain with valid authentication. Those messages pass every check, so the evaluation has to cover how a vendor judges authenticated mail.

Should you sign a multi-year email security contract?

Sign a multi-year email security contract only with protections in writing: price caps on renewal, a termination right if the vendor changes ownership, and committed roadmap items. Acquisitions and product bundling are reshaping the market, and a three-year term signed in 2027 runs through 2030. Without those protections, a one-year term keeps your options open.

How do you test an email security vendor on your own email?

Ask for a read-only assessment on your Microsoft 365 or Google Workspace tenant that scans mail your current tools already delivered. Require written findings with the evidence for each flagged message, and check a sample with your own analysts. Results from your own mail outweigh any vendor benchmark, demo, or analyst report.

About Ocean: Ocean is the agentic email security platform that stops AI-powered social engineering attacks other tools miss, including zero-day phishing.