Duty analyst on shift

UK Cyber Defence

EmilyAI · AI SOC analyst · In service since April 2018

The night has alreadybeen looked at.

EmilyAI is the AI analyst at the core of our security operations centre. She reads the data, forms a view, writes it down as a ticket with the evidence, and hands it to a person who checks her work. What you are buying is a quiet morning.

Begun 2014Live since April 2018Checked by analystsSigned updates every five minutesWith our SOC or inside yours

01What she is

An AI analyst nobody audits is a liability. One audited continuously is leverage.

Emily works the way a competent human analyst works: she forms a view of what an event is, and she writes that view down as a ticket. Around her is a small team of talented engineers who handle everything she cannot, and who do something more consequential than that. They watch her. They check her assertions, correct her where she is wrong, and keep her trained.

Nobody can sell the absence of threat. What Emily and the team provide is the confidence that the night has been looked at by someone whose only job was to look at it, and that anything which mattered has a name, a ticket and a next step waiting.

Reads
All client data in the SIEM
Reviews
Every client, hour by hour, with a structured threat hunt each morning
Writes
Tickets, with the evidence, and the SOC day book
Checked by
Engineers who verify her assertions, correct her and keep her trained
Decides
People. Emily raises, measures, explains and recommends

Emily's quarter, August to October 2026

89%

Of the quarter's investigations were first raised by Emily

570

SOC day book entries written by Emily in its first fortnight, about 40 a day

5,771

Dynamic blocklist updates, more than 60 a day

95

Firewall rule updates, about one a day

1,390

Abuse reports sent to the ISPs and hosting providers behind attacking machines

310

Critical and high vulnerabilities proved fixed by a follow-up scan

  1. Of the quarter's investigations were first raised by Emily: 56 of 63 investigations and threat advisories
  2. SOC day book entries written by Emily in its first fortnight, about 40 a day: 570 of 588 entries, 16 to 29 October 2026
  3. Dynamic blocklist updates, more than 60 a day, Firewall rule updates, about one a day: EmilyAI's own records of the actions it took
  4. Abuse reports sent to the ISPs and hosting providers behind attacking machines: About 15 a day
  5. Critical and high vulnerabilities proved fixed by a follow-up scan: 111 critical and 199 high

The record

She keeps the day book

Every hour Emily writes up what she looked at, what she found, what she could not see, and what she is leaving for the duty analyst. It reads like a handover note from a careful colleague, because that is what it is: posture first, then the reasoning, then the open questions.

Note the judgement calls. Where she cannot prove something she says so, and says why she has not raised a ticket. The analysts read the same book, add to it, and correct it.

An EmilyAI hourly review in the Pulse SOC day book (dark): posture, quiet hosts, tickets, SIEM review. Demonstration data.
One of Emily's hourly reviews in the SOC day book. Demonstration data: client names are replaced with sectors, and every host, user, address and ticket is invented.

02

Defend and disruptAutomated defences

Machine speed where it is clear-cut. People where it is not.

Not every attack needs a person to decide what to do. When an attack is clear-cut, EmilyAI's automated defences block it at the firewall and on dynamic blocklists, and report the attacking machine to the ISP or hosting provider responsible for it, with the evidence that provider needs to take the machine off the internet for everyone.

In September 2026 one client came under an attack generated by AI. It ran at 271 attempts a second and kept that up for six hours, about 5.9 million attempts in all. Emily made eleven automated responses, each one updating the firewall, and the attack was defended. An attack written by a machine moves at machine speed, so the defence has to as well, with people watching over it.

People stay in charge of everything that needs judgement. In an investigation Emily raises, measures, explains and recommends, but she never contains a machine or locks an account herself. An analyst decides, and calls you when the decision needs your knowledge.

Blocks
Firewall rules · Dynamic blocklists
Reports
Abuse reports, with evidence, to ISPs and hosting providers
Never
Contains a machine or locks an account on her own
September 2026
271 attempts a second for six hours, defended with 11 automated responses

Last quarter

What she raised, and how it ended

  1. _0163 investigations and advisories, across 11 organisations46 began with an alarm in a client's own data. The other 17 were threat advisories: a newly published vulnerability checked against the systems the client actually runs.Raised
  2. _02Around three in four had an innocent causeAn administrator's change, a policy update, scheduled maintenance, or a penetration test the client had commissioned. Each client got that answer with the evidence, not a list of red alerts.Explained
  3. _03The real ones were contained, or shown to have failedA password spray, a stolen laptop, hacking tools and an unexpected account. Each was contained, shown to have failed, or sent to the client with the evidence needed to act.Resolved
  4. _04Most answers arrive by morning45 of the 62 investigations closed were closed within 24 hours. Fifteen were raised between 6am and 8am, as her early sweeps and morning hunt reported on the night.Timed

Stories from the quarter

Eight mornings that started with an answer

These stories are anonymised: they name no clients, people or systems. Each ended the way a good SOC engagement should, with a calm morning and a clear answer.

01Financial services · October

The penetration test that met the SOC

A firm commissioned a two-week penetration test, and the SOC picked up the testers at every stage: an unknown machine scanning the network, a dynamic honeypot Emily raised and the testers wandered into, new administrator accounts and changes to multi-factor sign-in, and an attempt to copy every password hash from a domain controller. Each alarm was checked with the client by phone. A penetration test tells you whether an attacker could get in. A SOC tells you whether you would know.

02Engineering · September

64,000 failed logins, no way in

Overnight the SOC spotted a password-spraying campaign against a remote-access VPN: 64,624 failed attempts in 24 hours, against 110 usernames, from 69 rotating addresses. The analysis showed the client's own VPN controls were quietly turning away unknown users, and no login succeeded. By the next morning the client had the full picture.

03Healthcare · October

A ransomware alarm at 7am, answered by breakfast

A rule linked to Blackbit ransomware fired 20 times on one workstation just after 7am. Within about 60 seconds Emily had traced it to a single Defender policy update pushed by the client's own IT team. The client confirmed it, and the ticket was closed before the working day began.

04Pensions · October

A stolen laptop, watched until it was safe

After a staff laptop was stolen overnight, Emily watched for the device and its user's account to reappear. The audit logs showed the password had been reset and old sign-in tokens revoked, and the SOC followed the replacement laptop as it joined the network.

05Financial services · August

A hacking tool on one desktop, and only one

Late on a Saturday afternoon, Defender caught a Windows licence-cracking tool on a single workstation. Within the hour Emily had completed an autonomous threat hunt and shown it was on no other machine, with no sign of an intrusion. The client was told what to do next: confirm the removal, then rebuild the machine.

06Private banking · September

The Tor alarm that wasn't

On a Sunday morning a rule built to spot the Tor browser fired on a bank's laptops. By late morning the SOC had proved the rule was catching two Microsoft programs whose names happen to contain "tor". The SOC told the bank, then Emily rewrote the rule so it would not fire again.

07Construction and property · August

An account nobody asked for

The morning threat hunt found a new local account created on a laptop by a remote-management script. By the evening the same account had appeared on 67 machines and had never been used to sign in. The client did not recognise it, and had the evidence the same day to take up with its IT provider.

08Insurance · October

Admin tool or attacker?

At 10am on a Monday, a remote-administration tool used by IT teams and attackers alike installed itself on one of the firm's machines. The SOC asked the client rather than guessing. It was planned maintenance, and the ticket was closed by early afternoon.

03

The HiveSigned updates and threat intelligence

She does not sit on a shelf going stale.

Every Emily, wherever she runs, calls the Hive every five minutes and takes down cryptographically signed updates, commands and threat intelligence. The Hive is our threat intelligence network: our own honeypots and sensors, ipinsights.io, anonymised telemetry from across our client base, and commercial and open feeds.

That is the single most important thing to understand about buying her on her own. An Emily running on your infrastructure is as current as the one working inside our SOC, and she is never reasoning about your estate in isolation.

Cadence
Every five minutes
Integrity
Updates and commands are cryptographically signed; anything unsigned is ignored
Carries
Updates · Commands · Tuning · Threat intelligence
For your security team
The channel's security architecture is available under NDA

Two routes

With our SOC, or inside yours

EmilyAI can be bought with the SOC365 subscription or without it. If you already run a SOC, with a SIEM you are happy with, a SOAR and analysts on shift, then what you are short of is capacity, and capacity is precisely what Emily is. If you do not run one, or have tired of running one, we supply everything around her.

Neither route is a lesser version of the other, and nothing is wasted if you start with one and move to the other.

Diagram: EmilyAI on her own and EmilyAI with SOC365, showing who supplies each layer, with the Hive above both.
The same Emily on both routes. What differs is who supplies everything around her.
EmilyAI on her ownEmilyAI with SOC365
ForAn organisation that already runs a SOCAn organisation that would rather we ran it
ReportingYour own reporting and case managementPulse, plus the monthly Cyber Risk Report
AnalystsYour analysts verify her findings and decideOur analysts verify her, train her and escalate
ResponseHer own automated actions, plus your SOARHer own automated actions, plus our analysts
EmilyAILicensed to you, on your infrastructure, tuned by usIn our SOC, tuned by us
SIEMYour SIEM; she reads from itOurs, or we read your SIEM or Microsoft Sentinel
Extended visibilityNot includedDeception, OT and IoT, cloud, identity and data loss modules
CollectionYour agents and log feedsWazuh agents we deploy, or your existing feeds
The HiveSigned updates and intelligence every five minutesSigned updates and intelligence every five minutes

Questions

What security teams ask about Emily

Does EmilyAI act on our systems without asking?

Only in one narrow way. When an attack is clear-cut, her automated defences block it at the firewall and on dynamic blocklists, and report the attacking machine to the network that hosts it. Beyond that she takes no action on your estate: she never contains a machine or locks an account herself. Those are decisions for an analyst, taken with your approval or under authority you have agreed in advance.

Who checks her work?

Our engineers, continuously. They check her assertions, correct her where she is wrong and keep her trained, and they fix the detection rules that produce false alarms so the same false alarm does not recur. On the standalone route your own analysts verify her findings and decide.

How long has she been in service?

EmilyAI was begun in 2014 and has been in live service in our SOC since April 2018.

Does an Emily on our own infrastructure fall behind the one in your SOC?

No. Every Emily calls the Hive every five minutes for cryptographically signed updates, commands and threat intelligence, wherever she runs.

What do we need in place to run her ourselves?

A SIEM you are happy with, your own collection, analysts on shift to verify her findings, your own case management and reporting, and a SOAR if you want response orchestration beyond her own automated actions. We provide onboarding and tuning. EmilyAI for your own SOC goes through it.

Can we start with Emily on her own and move to SOC365 later?

Yes. Some organisations start with Emily on their own infrastructure and move to the subscription once the cost of keeping analysts on shift has made its own argument. Nothing is wasted in that move: it is the same Emily and the same Hive throughout.

Can we see how the update channel is secured?

Yes. The security architecture and cryptographic specification of the channel between Emily and the Hive is available to your security team under a non-disclosure agreement. Ask us for it.

Start a conversation

See Emily's working on a real night.

A thirty-minute technical briefing with the SOC lead, using an anonymised incident from last month: what she saw, what she wrote, and what the analyst did next.