Here is a fact about security operations centers that almost nobody in a security operations center will say out loud: they do not look at most of their alerts. This is not laziness. It is arithmetic. A single analyst can face thirty-plus alerts in a shift, each one a small mystery that might take twenty minutes to unravel and usually turns out to be nothing, and there are not enough minutes in a shift, or a life, to unravel all of them. So the queue grows, the analyst triages the loudest few, and everyone quietly agrees not to mention the rest. This works fine right up until the one alert that mattered was in the pile nobody opened.
Radiant Security, founded in 2021 and headquartered in Pleasanton, California, is a company built entirely around that uncomfortable pile. Its pitch is that the alert problem is not a hiring problem - you cannot hire your way to reading 100% of alerts, because the alerts scale and the humans do not - but a triage problem. And triage, it turns out, is a thing you can automate if you build the right kind of AI.
The founder saw the queue from the inside
The person who decided this is Shahar Ben-Hador, Radiant's co-founder and CEO, who has the useful biographical detail of having actually lived the problem. He spent nearly a decade at Imperva, rising from IT admin all the way to Chief Information Security Officer, then served as CIO at Exabeam, a company that sells the security analytics tools whose alerts pile up in the first place. When someone who has been the CISO getting paged at 3am tells you the SOC needs less noise and not more dashboards, it lands differently than the usual startup origin story.
He co-founded the company with Barry Shteiman, the CTO, another security-industry veteran, and the two assembled a team stocked with former CISOs, threat researchers, and product builders spread across the San Francisco Bay Area, Tel Aviv, and Sao Paulo. The through-line is that these are people who have done the job, which is relevant because the product's core promise is that it will do the job the way an experienced analyst would.
"Radiant is the only AI SOC built to handle every alert type across all your data sources, with no playbooks or pre-configured rules required."
The bet against playbooks
The most interesting thing about Radiant, and the thing that most separates it from a decade of security tooling that came before, is what it refuses to do. The previous generation of security automation - the category was called SOAR, for security orchestration, automation and response - operated on playbooks. You anticipated a threat, you wrote a step-by-step script for how to handle it, and when that exact threat appeared, the machine ran your script. This is great for the threats you anticipated and useless for the ones you did not, which are, inconveniently, the threats that tend to hurt you.
Radiant's platform ships without playbooks. Its agentic AI is designed to investigate an alert it has never seen before, from scratch, reasoning its way through the evidence the way a human analyst would rather than matching against a pre-written rule. It automates triage, investigation, and remediation across 100% of alert types - cloud, network, data loss, endpoint, identity, email, supply chain - regardless of source, complexity, or novelty. Then it does the merciful thing: it escalates only the real threats to a human, who can respond in one click.
The claimed result is a number the company puts front and center: up to 98% alert reduction. An analyst who was drowning in thirty alerts a day is left with two or three escalations that Radiant believes are genuinely worth a human's attention. Whether that number holds across every customer is exactly the sort of thing a skeptic should probe, but the shape of the claim is coherent with the thesis: the machine does the reading, the human does the deciding.
The quiet second business: log storage
There is a less glamorous feature that may matter just as much to the people writing the checks. Security teams pay enormous sums to store their logs, usually inside a SIEM - a security information and event management system - whose pricing has a way of scaling with your data until the storage bill becomes its own line-item crisis. Radiant folded log management directly into the platform, ingesting and compressing logs at roughly 70% efficiency and claiming to cut logging costs by up to 85% while keeping the data searchable.
This is a shrewd bit of product design. It bundles a cost-savings story with a capability story, and it means the AI is investigating alerts against the same log data it is cheaply storing. The company frames it as escaping "the SIEM tax," which is the kind of phrase that makes a budget owner nod.
"Automate investigation across 100% of alert types and escalate only real threats to analysts, who can then respond in one click."
A category of one becomes a crowd
When Radiant started, "AI SOC" was barely a phrase. By 2025 it was a market, with a wave of new entrants - Dropzone, Prophet, Seven AI, Tines, Simbian and others - all pitching some version of the agentic security analyst. A lesser company might treat that as a threat. Radiant, in a blog post titled to the effect of "2025 proved why we built AI SOC before it even had a name," treated it as validation. Being early is lonely, the reasoning goes, right up until it is obvious.
The outside world has been reasonably kind. The Software Analyst Cyber Research group named Radiant "most unique value proposition" in its 2025 AI SOC Market Landscape, singling out the ability to triage any alert type plus the log-management cost savings. The company also collected a "highest triage fidelity" nod and, in early 2026, "Best SOC Automation Solution" at The Hacker News Cybersecurity Stars Awards. Radiant reports protecting more than 30 organizations and over a million users and endpoints, and surpassing 100 integrations, which it argues is the precondition for its 100%-coverage claim - you cannot triage an alert from a tool you cannot connect to.
Why transparency is a feature, not a slogan
The last piece worth noting is philosophical but also intensely practical. Radiant leans hard on AI transparency and explainability, and this is not just for the marketing. If an AI decides which threats reach a human and which get closed, the human has to be able to see the machine's reasoning - because nobody is going to stake an incident response, or their job, on a verdict they cannot inspect. In a domain where the cost of a wrong "this is nothing" can be a breach, showing your work stops being a nicety and becomes the whole basis of trust.
That is the tidy version of Radiant Security: a company that looked at the least sexy fact in cybersecurity - that everyone silently ignores most of their alerts - and decided the fix was not more people or more dashboards, but an AI patient enough to open every one. The $15M Series A that Next47 and Lightspeed led in late 2023 is a bet that this patience scales. The alert queue, at least, is not going to get any shorter on its own.