The Unbounded SOC

Beyond the single pane of glass.

Solve the problems you see. Find the ones you don’t.

Alerts answer the problems you can see. The AI SOC closes the gaps you can’t:

  • Coverage: systems that were never sending the data you need.
  • Intelligence: attacks that have moved faster than what you know.
  • Detection: rules written for someone else’s network and tools, not tuned to yours.

Fluency finds and closes each one from your own logs, on your own systems, and keeps doing it as attackers change. From investigating every case to CISO-level reporting on health, gap analysis and certification, it’s the whole operation, not a view of it.


From $7 per user a month, one year of retention included →

One incident in. Coverage for every tenant out.

the loopeach gap seedsthe next case1incident2groundtruth3intel4test5huntfleet6rulegaps
An AI agent does itA person decides

From firehose to five

Twelve million events in. One decision out.

12,000,000events ingested today

  1. Every log, flow, alert and API call from endpoint, identity, cloud and network sources, normalized on arrival.

    SyslogCloud APIsEDRIdentityNetwork

  2. About 65% is operational telemetry. It goes straight to Parquet datastores, or becomes Prometheus metrics, so it stays searchable without being analyzed as security events.

    ParquetPrometheus metricsFull retention

  3. Streaming analysis enriches and scores the rest. One event in ten is worth a second look.

    EnrichmentScoringStreaming

  4. Detection signatures, mapped to MITRE ATT&CK, fire on the notable events. Thousands of hits, not millions.

    SignaturesMITRE ATT&CKThreat intel

  5. Behavior models catch what no signature describes: first-time access, impossible travel, unusual process chains.

    BaselinesFirst-seenAnomaly

  6. Hits and behaviors are grouped by user and entity into case summaries, each with its full story attached.

    Per userPer entityTimeline

  7. AI analysts investigate every case and close about 95% with evidence attached. Five need a person.

    Investigates every caseEvidence attachedExplains why

  8. Occasionally one is critical. It reaches an analyst immediately, already investigated, with the recommended response.

    ImmediateResponse ready

ParquetPrometheus metricsevery eventneeds a person now
Illustrative day0% removed

Video series · 5 parts

What an AI SOC does, and how it runs: the structure underneath it, how it investigates a case, how every case makes detection better, and what it reports to the CISO.

Part 1 of 5 · 2:24

Overview

What an AI SOC does

The features and functions of an AI SOC, and how one actually operates, in four steps: build the structure, investigate with it, make detection better, and report to the CISO.

Recognized by G2 · Fall 2026

High performer. Best support.

Customer satisfaction and support quality, recognized in G2's Fall 2026 reports.

G2 Best Support Fall 2026 badgeG2 High Performer Fall 2026 badge

01 · Investigate and respond

Root cause, response and mitigation

AI works every case the way a SOC does, tier by tier.

Alerts arrive already grouped into one scored case per entity per day. AI then validates and scopes each case, reconstructs its root cause, drives the response and runs the hot wash, searching every relevant data source in parallel. Whatever the hot wash finds missing becomes the next turn of the learning loop.

  1. Tiers 1–2

    Validate and scope

    Is it real? Is it actionable? Do we see it anywhere else?

  2. Root cause

    Reconstruct

    Reach back through the data to find out why it happened, not just that it happened.

  3. Tier 3

    Respond

    Use all of that analysis to deal with the immediate danger and the long-term one.

  4. Tier 4

    Hot wash

    What did we miss? Sometimes all we see is the preparation and the after-effects of an attack.

A real case

The investigation found what the first pass couldn’t see.

A user clicked a phishing link in Teams, the first time that alert had fired anywhere in the organization: 8,800, Critical. The investigation found Microsoft removed the message 14 seconds after the click, and closed with a quantified verdict, about 80% likely to be real, and the one check left to settle it.

When the evidence shows a stolen Microsoft 365 session being replayed, Fluency AiTM Module revokes the user’s sessions and tokens within seconds. Every other response is recommended with its evidence, for your team or your MSSP to carry out.

A quantified verdict: about 80% real issue, 15% benign, 5% indeterminate, with reasons on both sides
A quantified verdict with the case for both sides, and the one check that would settle it.Expo demonstration tenant · synthetic identities

Reports · 1

  • Case Investigation See a sample →

    One case from alert to evidence, root cause, verdict and actions.

Technical white paper

AI Workflow versus Swarm AI

Not all AI investigations are the same. Compare an AI-executed playbook with parallel, schema-specific analysis across data lakes. Real security cases show where the verdicts, categorization, and proposed responses differed.

Download the white paper PDF · 8 pages · no form

Or just watch the video

02 · Detection engineering

Signatures, behaviors and code

Detections tuned to your network, tested before they ship.

Rules written for someone else’s tools rarely fit yours. Fluency’s behaviors learn each identity’s own history, its countries, networks, devices and actions, so detection fits your environment. Every rule runs in the stream, before anything is written to the database. AI writes and tests the rules, and a person approves every release.

If you can match it

Signature

A known pattern in a single record: an error code, a known-bad indicator, a vendor alert.

If the entity changed

Behavior

Something new for this person or device: a first-time country, network, action or alert.

If it needs a sequence

Code

A different pivot, a chain of steps or real logic, like one app that is created, granted rights, given a secret and then acts.

Three tests, then a staged rollout.

  1. 1The real incident, sanitized, must fire the rule.
  2. 2Made-up variants of the attack must fire it too.
  3. 3Normal admin work must not.
  4. 4A changed processor is replayed against the old one, and every difference has to be explained.
  5. 5Mirror sites that copy real traffic first, then one real tenant for a day, then the fleet.

A person decides: releases reach tenants only after someone reviews the tests and pushes them.

Scored by evidence that agrees.

A behavior is a signature plus history: a first-time country, network, action or alert for this person is a far stronger signal than something that happens every day. A case’s score rises only when separate evidence points the same way, so twenty bad logins and a thousand are the same evidence.

A Fluency case for one user: risk score 14,400, Critical, built from two behavior rules scored 4,400 and 2,800
Two behavior rules, 4,400 and 2,800, agree, so the case scores 14,400, Critical.Expo demonstration tenant · synthetic identities

Reports · 1

  • MITRE ATT&CK Coverage See a sample →

    Which techniques your detections cover, where the gaps are and what to add.

03 · Threat intelligence

New threat data becomes coverage

Keep pace with attackers, not with yesterday’s indicators.

New threat reporting is set against the ground truth in our own logs: is this a known actor, which reported steps did we see, and which did we not? Each step is mapped to the record that would show it, and indicators that will not hold are flagged before anyone builds on them.

Case C · Storm-3119

Microsoft’s write-up named a “PnP Management Shell” app. The retired app had a fixed ID, but since 2024 each registration gets a random one. A watchlist of that ID would have missed every new victim, so the detection keys on behavior instead: create, grant, secret, act.

Once a pattern is understood, it is hunted across every tenant’s history and checked against every rule we run.

04 · System health

Coverage: every site, every source, every agent

See what you’re not collecting.

Detection can only work on the data you collect. Fluency checks that your systems send the logs that matter: sources arriving on time and complete, endpoint agents reporting, protected and current, and the machines that are not reporting at all. Each gap gets an owner, and a source that never reports is shown as a collection gap, never as a passing control.

Every Monday at 6:00

“Run the health status report for all my tenants. If any data source is broken or stale, message me the list on Telegram. Otherwise, do nothing.”

From the same conversation, your team can onboard new data sources and close cases. Every change is previewed and confirmed first, and credentials go through a secure intake page, never through chat.

Reports · 4

  • Data-Source Health and Status See a sample →

    Whether every source is arriving on time and complete, across every tenant.

  • CrowdStrike Falcon Endpoint Health

    Protection, reporting and OS currency, and the hosts where risks compound.

  • SentinelOne Hygiene

    Agent check-in, scan policy and the Windows endpoints falling behind.

  • Microsoft Defender Hygiene

    Coverage, antivirus health and configuration breaks, with an owner for each fix.

05 · Posture and exposure

Configuration, identity, data and attack surface

Not only what happened. What things look like.

Alongside the activity, Fluency keeps configuration snapshots at least daily: Office 365 users, devices, groups and apps; endpoint agents and OS versions; Defender vulnerabilities and Secure Score; firewall configuration; the latest CVEs. So the gaps in configuration, identity and exposure are found before an attacker finds them.

“Are things properly configured?”
“Who holds privileged access, and how did they get it?”
“Which vulnerabilities are open on our devices?”
“Is MFA enforced everywhere it should be?”
“What can a stranger see from the internet?”
“Are our endpoint agents protected and current?”

Reports · 3

  • Endpoint and Identity Security Posture

    Endpoint protection combined with Microsoft 365 and Entra ID identity and device posture.

  • Microsoft 365 Identity and Access Posture

    MFA, privileged roles, OAuth consent, licenses and device operating systems.

  • External Attack-Surface Footprint

    Public assets, open ports, lookalike domains and DNS, mail and web hygiene.

06 · Gap analysis and certification

Know what your evidence actually supports

Readiness built from live evidence, with the gaps named.

Fluency maps live evidence onto NIST SP 800-53 Rev. 5 and CIS Controls. Each control is marked supported, partial or not observed, with a roadmap to close the gaps.

Fluency supports readiness. Certification decisions stay with your assessor, and it works alongside platforms such as Vanta and Drata.

NIST SP 800-53 Rev. 5 readiness review with evidence coverage
A NIST 800-53 readiness review built from live telemetry, with gaps named.Expo demonstration tenant · synthetic identities

Reports · 1

  • CIS Evidence Readiness

    How much of a CIS Controls assessment your live evidence supports.

07 · CISO and board reporting

Analyze · forecast · compare

“Build the monthly SOC report for August.”

That’s the whole request. The report skill confirms the customer and the month, asks the data 8 to 16 questions about cases, response times, repeat actors and data-source health, and writes an eight-chapter report. Where the data can’t show something, the report says so. Halfway through the next month, ask it to compare: it forecasts the month to its end, analyzes it, and sets it against the last.

“Compare it with July and add recommendations for the board.”

More analysis, for a new audience.

“Format this as a PDF.”

A 16-page, print-ready PDF, checked page by page.

“Post it to Companion.”

Filed in Monthlies for August. Last month’s moves to the archive.

Monthly SOC report: headline numbers compared with July and cases per day for August
From the August report: headline numbers against July, and every chart built from this month’s data.Expo demonstration tenant · synthetic identities

Reports · 1

  • Monthly SOC Report

    An eight-chapter monthly story for the CISO: what happened, the risks, owners and dates.

Why it works

Unbounded by design, run by the AI you already use.

Fluency prepares your data and teaches the agent the job, so Claude, Codex or any MCP client can run your operation, on your systems, with exactly your role and scope.

Structure

Cases, not alerts

Every record is enriched once and detected in the stream, then grouped into one scored case per entity per day, alongside daily configuration snapshots.

Context

Context windowing

Cheat sheets work out which operation you are in and put only its tools and skills in play, so the agent works from current rules instead of guesses.

Action

Skills and tools

Skills are recipes for whole jobs, like every report on this page. Tools are governed functions with the data, state and safety built in.

Five states of data: raw data, events, notable events, alerts and cases
Raw data → events → notable events → alerts → cases. Each step makes the data smaller and more meaningful.

Connect

One address, one sign-in, once per client.

Every Companion site has its own MCP address. Add it to your agent, sign in with the same Microsoft or Google identity you use for the Companion, and approve. The agent gets exactly your role and scope, nothing more.

Claude Code

claude mcp add --transport http fluency https://<your-site>/mcp

Then run /mcp, choose fluency and Authenticate.

Codex

codex mcp add fluency --url https://<your-site>/mcp

Then codex mcp login fluency to sign in.

Claude Desktop and Claude.ai

Settings → Connectors → Add custom connector, paste the /mcp address, then sign in and approve.

Any MCP client

Add a remote (streamable HTTP) MCP server with the /mcp address and complete the browser sign-in.

Read-only by default

Reports, searches and case reviews change nothing in Fluency.

Changes are confirmed

Anything that changes state is previewed first, needs your explicit confirmation, and is recorded.

Your scope, every call

Role and scope are checked on every request. One customer or the whole fleet, never mixed in one answer.

Secrets stay out of chat

Credentials go through a secure intake page. The agent never sees your connector secrets.

Technical white paper

Why AI belongs in front of your SIEM, not behind it.

A SIEM behind a fixed interface takes away what AI does best: analyze, forecast and compare. This paper shows how Fluency's MCP teaches Claude or Codex to run the SIEM with cheat sheets, governed tools and skills, through a real Critical phishing investigation and an eight-chapter monthly SOC report.

Download the white paper PDF · 13 pages · no form

Or just watch the video · 6:40

How Fluency compares

AI investigation is the baseline, not the differentiator.

We compared ten AI SOC platforms on the six capabilities around the shared promise: the evidence a platform keeps, the environment it understands, the reports it delivers and the actions it takes.

CapabilityFluencyWhere it stands
SIEM foundationHighest tier, shared with Exaforce: a full data foundation, not an overlay on someone else’s SIEM.
Raw telemetry historyHighest tier: 12 months of raw history by default, not just preserved case summaries.
Configuration intelligenceHighest tier: identities, apps, devices, Defender vulnerabilities and Secure Score, firewall and audit changes.
Customer reportingThe only platform at the highest tier, with a full assessment suite: incident, hygiene, compliance and gap reports.
Hygiene and exposureSubstantial: posture and hygiene reviews across endpoint and identity.
Response automationBy design, findings and recommendations for your team or MSSP to carry out, plus AiTM session revocation.

Scored 1 to 3 for documented breadth, from published material as of 1 October 2026. A comparison of scope prepared for Fluency, not a laboratory ranking of security effectiveness.

For MSSPs and MSPs

One customer or the whole fleet. Never both in one answer.

Access is checked two ways on every request: role decides what someone can do (Admin, Operator, CISO, Analyst), and scope decides which tenants they can reach. An MSSP sees every customer under its connectors, an MSP sees a hand-picked set, and a customer sees only itself. Analysts and AI clients inherit exactly that access.

The company behind the product

A team that helps you put security intelligence to work.

Fluency combines its SIEM with hands-on deployment and ongoing support. We help customers overcome integration obstacles, establish useful workflows, and build operational capability.

Deploy with a practical plan

Agree on the first sources, access requirements, and questions to answer. Verify data flow and usable results as onboarding progresses.

Operate across customers

For MSSPs, multi-tenant workflows support customer onboarding, health reviews, investigations, and reporting within the appropriate customer scope.

Keep improving with support

Work with a team that understands the product and the operational obstacles. Fluency was recognized for Best Support in G2’s Fall 2026 reports.

Meet the team behind Fluency

Bring your own data

Run your whole SOC on your own data.

Start with Microsoft 365. Free for 30 days, no credit card. After the trial, SMB pricing starts at $7 per user a month with a $75 monthly minimum and one year of retention.