SystemSpecialty, E&S & London Market

Cyber Accumulation Management

Cyber accumulation management: which insureds a vendor event exposes, and how much limit, within the hour

Exposed insureds and limit listed the day a vendor event breaks.

See one case, screen by screen ↓
demo12minfrom the vendor’s status page notice to 418 insureds and $340M of limit matched
demo113%of the board’s $300M per-vendor tolerance riding on one cloud region
demo$9.32Mgrossestimated business interruption if the outage lasts 12 hours — $6.52M net of the quota share
target418of 418dependencies shown with their evidence and date
In short

What is Cyber Accumulation Management?

It is knowing how much of a cyber book rides on one shared cause — a cloud region, an identity provider, a payroll vendor — so a single outage or breach does not surprise the board or the reinsurers. Cyber Accumulation Management keeps a standing map of shared vendors and the insureds that depend on each, with the limit and dependent business interruption sublimits exposed.

How to choose AI solutions for insurance →

Updated

The problem

Why the first accumulation number takes days

When a widely used cloud, identity, payroll or managed-IT vendor goes down, the board and the reinsurers ask one question: how much of our limit is sitting on it? The answer is scattered. It sits in application answers (“primary hosting provider”), in outside-in scans and in managed-IT disclosures that nobody has indexed by vendor.

So the team searches applications and calls brokers by hand, and the first figure reaches leadership days later — usually a total limit, not an estimate. Dependent business interruption cover differs insured by insured: waiting periods from 6 to 24 hours, system-failure cover often absent or sublimited, different retentions. Without those terms, the estimate means little, because in a short outage most insureds never pass their waiting period.

estimated≈3daysfrom vendor event to the first exposure figure, by hand
typical3–10%of global cyber premium lost to one shared-vendor outage, July 2024
typical$0.4–1.5bnestimated insured cyber loss from that one outage
Where the hours to a first exposure figure goestimated
By hand72 days
With the solution2.6 days
  • Finding which insureds depend on the vendor30 → 0.2 d
  • Reading limits, sublimits and waiting periods16 → 0.1 d
  • Estimating business interruption12 → 0.3 d
  • Telling each broker who is exposed8 → 1 d
  • Writing and signing the note to the reinsurers6 → 1 d

Estimated split, in hours, for one shared-vendor outage, by hand and with the solution.

How it works

How a vendor event moves

Six agents watch vendors, match every dependency to its evidence and estimate the outage; the head of cyber decides.

What comes in
Evidence in4 sources · advisories, scans, applications, in-force policies
Agents at work
Event watcherstatus pages, advisories
Then
Dependency matcherscan, application, provider
Policy readerlimits, waiting periods
Then
Outage estimatorgross and net
Then
Outreach drafter
Event note writerevery number cited
A person decides
Head of cyberdeclares and signs
What comes out
Exposed insureds with evidence
Limit and BI estimate
Broker emails and event note
One case, step by step

One vendor outage, from the first notice to a signed note

Monday 12 October, 06:12 ET: Nimbara Cloud reports elevated errors for compute and storage in its US-East 3 region. Here is what happens next at Thistlebrook Specialty’s cyber book, screen by screen, in the working solution.

  1. 01Morning

    The whole book around its shared vendors

    Rhea Whitlock · Head of Cyber

    Rhea opens the dependency web: $2.93B of aggregate limit, 2,846 insureds in ten sector wedges, 11 shared vendors in the centre. One vendor is over the $300M tolerance, and the live bar at the top already carries EV-0412, the Nimbara Cloud US-East 3 outage, with its clock running.

    “1 vendor over the $300M tolerance. Choose one to light every insured that depends on it.”

  2. 02One click

    $340M of limit lights up

    Rhea Whitlock · Head of Cyber

    Rhea chooses Nimbara Cloud. A line runs to every insured that depends on it, coloured by how we know: 236 seen in scan, 131 declared on application, 51 through a provider. The rail totals it: 418 insureds, $340M of limit, $181.5M of dependent-BI sublimits, 41 brokers — $40M over the board’s tolerance.

  3. 0306:12 – 06:24

    How it was found, step by step

    Event watcher · Dependency matcher · Policy reader

    The event watcher read the status page notice and matched it to Nimbara Cloud in the vendor register. The dependency matcher found the insureds in scans, applications and provider disclosures. The policy reader read all 418 in-force policies: 63 have no system-failure cover. Each step shows its time, and the match can be re-run at any point.

    “Increased error rates for compute and storage in US-East 3” — the notice is quoted with its time.

  4. 0406:31 · declared

    A watch becomes an event

    Rhea Whitlock · Head of Cyber

    The agents open watches; only the head of cyber declares an event. Rhea declares EV-0412, and outreach and the reinsurer note start. The event workspace keeps one timeline of sources, agent steps and people’s decisions, the exposure by evidence type, and every notice of circumstance as it lands.

    Rhea’s decision, on the record: “Over tolerance — start outreach and the note.”

  5. 05What if

    What if Nimbara is down for 12 hours?

    Outage estimator

    One press and a slider from 2 to 72 hours. Every insured whose waiting period plus recovery is passed turns red on the web. At 12 hours, 258 of 418 pass their waiting period and 186 lose more than their retention: $9.32M estimated business interruption gross, $6.52M net of the 30% quota share — inside the $10M retention of the Vauclair Re event cover.

    An estimate is never a reserve — every figure is labelled an estimate.

  6. 06The largest

    Calloway Freight Systems, from its own policy

    Policy reader · Outage estimator

    Two sources say Calloway depends on Nimbara: the 11 Oct scan found its booking and transport-management hosts on Nimbara US-East 3 addresses, and its April application names “Nimbara Cloud, US-East region” as primary hosting. Its policy: $10M limit, $5M dependent-BI sublimit covering system failure, 8-hour waiting period, $250k retention. 12 hours plus 3 hours of recovery, less the waiting period, is 7 covered hours at $275.6k an hour, less the retention: $1.68M.

    Notice of circumstance received 08:55 from Brightwater Risk Partners; claim file opened by Dana Osei.

  7. 07If it runs longer

    The loss curve, gross to net

    Tomás Ferreira · Portfolio manager

    The scenarios screen draws the estimate across outage lengths, by system failure or malicious attack, with the waiting-period bands and a split by sector. The event cover attaches at $14.3M gross — about 14.5 hours of downtime. At 24 hours the estimate is $40.4M gross and $28.3M net, and Vauclair Re would pay $18.3M.

  8. 0809:12

    Each broker gets its own list

    Kiran Bhatt · Senior cyber underwriter

    The outreach drafter wrote one email for each of the 41 brokers, with that broker’s exposed insureds attached and what to ask them: are operations affected, since when, and how to notify a circumstance. Brightwater Risk Partners holds 126 of the 418. Kiran reads it and presses Send.

    “Nothing goes out until an underwriter presses Send.”

  9. 0909:30 · drafted

    One page for the CUO and the reinsurers

    Event note writer

    Event note 1 is addressed to Martin Adeyemi, Orrin Re and Vauclair Re: what happened, who is exposed, the estimate at 12, 24 and 48 hours, reinsurance, and what the team is doing. Every number carries a numbered citation to its record — the status page, the dependency match, the estimate assumptions, the reinsurance programme, the notices and the outreach.

  10. 10Signed · 09:42

    The head of cyber signs and sends

    Rhea Whitlock · Head of Cyber

    Rhea reads the note and signs it. It goes to Martin Adeyemi, Orrin Re and Vauclair Re as notice of a potential event under both 2026 treaties — three and a half hours after Nimbara’s first notice.

    “Figures are estimates at 12, 24 and 48 hours. Signing confirms you have read them.”

Who it’s for

Built for everyone on the cyber desk when a vendor goes down.

The same event, seen by the five people who carry it — what their day looked like, and what it looks like now.

RW
Rhea WhitlockHead of Cyber, Thistlebrook Specialty
Declares and signs
Before
Faces the board and the reinsurers with a total limit built by hand over days.
Now
Has the exposed insureds, the limit against tolerance and a cited estimate before they ask — and signs the note.
TF
Tomás FerreiraCyber portfolio manager
Owns the assumptions
Before
Treats a dependency without a scan or an application line as a guess.
Now
Sees the evidence behind every dependency and runs the outage scenarios on assumptions approved and versioned in Settings.
KB
Kiran BhattSenior cyber underwriter
Sends outreach
Before
Hears from brokers after their clients start calling.
Now
Tells each broker once, with its own list, and presses Send.
DO
Dana OseiCyber claims lead
Takes the notices
Before
Learns who is exposed when the notices of circumstance arrive.
Now
Knows who is exposed before the notices land, and opens claim files from the exposed list.
MA
Martin AdeyemiChief Underwriting Officer
Reads the note
Before
Waits days for a number to stand behind in front of the reinsurers.
Now
Gets one page with every number cited, signed by the head of cyber.
Built on the engine

6 agents. Each with one job, and hard limits.

Six agents watch vendors, match every dependency to its evidence and estimate the outage; the head of cyber decides.

Event watcher

Reads vendor status pages, security advisories and the known-exploited vulnerabilities list every five minutes, matches the vendor to the shared-vendor register by name, product and region, and opens a watch.

  • Never declares an event — the head of cyber does
  • Quotes the source with its time
Dependency matcher

Finds every insured that depends on the vendor from outside-in scans, application answers and provider disclosures, with the evidence type, the record and its date.

  • No dependency without cited evidence
  • Dependencies through a provider kept separate from direct ones
Policy reader

Reads limit, dependent business-interruption sublimit, waiting period, retention and system-failure cover from each in-force policy and every endorsement in force on the event date.

  • Never assumes a cover the wording does not give — it records it as not covered
Outage estimator

Estimates business interruption insured by insured for an outage length and cause — outage plus recovery, hours past the waiting period, less retention, capped at the sublimit — then gross, ceded and net of reinsurance.

  • An estimate is never a reserve or a claim
  • Every figure labelled an estimate
Outreach drafter

Drafts one email per broker listing its exposed insureds, asking whether operations are affected and reminding how to notify a circumstance.

  • Never sends — an underwriter presses Send
Event note writer

Drafts the one-page event note for the chief underwriting officer and the reinsurers: what happened, who is exposed, the estimate at three outage lengths, reinsurance and next steps.

  • Every number cited to its record
  • Never sends a note unsigned
Head of cyber

Declares the event and signs every note. The agents propose; a named person decides.

Ask in plain words

Ask about any vendor, insured or outage length

Rhea and the cyber team ask in plain words. The answers come from the dependency web, the policies and the event record, with citations.

What if Nimbara stays down for 12 hours?

At 12 hours, 258 of the 418 insureds pass their waiting period once recovery is added, and 186 lose more than their retention. Estimated business interruption is $9.32M gross, $2.8M ceded to Orrin Re under the 30% quota share, $6.52M net — inside the $10M retention of the Vauclair Re event cover. 63 insureds have no system-failure cover and count only if the cause turns out to be malicious.

Why is Calloway Freight counted?

Two independent sources. The 11 Oct scan found book.callowayfreight.com and its transport-management API on Nimbara US-East 3 addresses, and the April 2026 application answers question 14, primary hosting provider, with “Nimbara Cloud, US-East region”. With an 8-hour waiting period and a $250k retention, it is the largest single estimate at 12 hours: $1.68M.

When does the Vauclair Re event cover attach?

At about 14.5 hours of Nimbara downtime. The event cover is $20M excess of $10M on the net after the 30% quota share. At 24 hours the estimate is $40.4M gross, $28.3M net, and Vauclair Re would pay $18.3M. The note to the reinsurers already says so.

Which brokers have the most exposed limit?

Brightwater Risk Partners, by a distance: 126 of the 418 insureds and $102.25M of the $340M. Then Calder & Rowe (37, $44.5M), Hollis Fenwick Brokers (26, $26.25M) and Tidewell Specialty Brokers (14, $22.25M). Kiran Bhatt sent Brightwater’s email at 09:12; the other 40 wait for an underwriter.

Every screen

The working solution, as it ships.

13 screens from the working solution, on its sample data. Pick one to see it large.

The dependency webThe whole cyber book around its 11 shared vendors: insureds in sector wedges, sized by limit, with the live event on top.
One vendor, every exposed insuredChoose a vendor and every dependent insured lights up, coloured by evidence, with the limit against the tolerance.
How it was foundThe largest limits on the vendor and each agent step that found them, with the times — re-run on demand.
What if it stays downA 2-to-72-hour slider: insureds past their waiting period turn red; gross and net estimate in the rail.
The event workspaceTimeline of sources, agent steps and decisions; exposure by evidence; notices of circumstance as they land.
Exposed insuredsAll 418 with limit, dependent-BI sublimit, waiting period, retention, broker and the estimate at 12 hours — filter and export.
One insured, its evidenceThe scan record and application answer behind the dependency, the policy terms, and the estimate line by line.
Outage scenariosLoss by outage length and cause, gross to quota share to net to event cover, by waiting period and by sector.
Broker outreachOne drafted email per broker with its own insureds attached; an underwriter sends one or all.
The event noteOne page for the CUO and the reinsurers, every number numbered to its source record.
Sign and sendThe head of cyber signs; the note goes as notice of a potential event under both treaties.
Shared vendorsEvery vendor more than 50 insureds depend on, the limit riding on it against tolerance, and its 12-month trend.
Your book’s rulesTolerances, the estimate’s assumptions, reinsurance terms, who decides and the feeds the agents watch.
Governance

Built for decisions the board and the reinsurers will read: evidenced, cited, signed.

Every dependency carries its evidenceA scan record, an application answer or a provider disclosure, with its date. An insured with no cited evidence is not counted, and dependencies through a provider are kept apart from direct ones.
Every number in the note is citedEach figure in the event note links to the record it came from — status page, dependency match, estimate assumptions, reinsurance programme, notices and outreach.
Estimates are labelled as estimatesBusiness interruption figures are marked estimate on every screen and in the note — never presented as a reserve or a claim. Closed events show how the first estimate held up against incurred.
People decide what leaves the buildingAgents open watches; the head of cyber declares the event and signs the note. Nothing reaches a broker until an underwriter presses Send, or a reinsurer until the head of cyber signs.
Assumptions owned and versionedRecovery time, the cause rules and the reinsurance terms behind the estimate change only with the portfolio manager’s approval, and each set is versioned — the current one, v3, approved by Tomás Ferreira on 2 Sep 2026.
Every step on the event timelineSources, agent steps and people’s decisions are recorded on the event’s timeline with who, what and when, in ET.
Configuration

Your book’s rules, not ours

Tolerances, the estimate’s assumptions and who decides are settings, not a project.

SettingDefaultChoose from
Limit tolerance per vendor$300M, set by the board$150M to $500M, in $25M steps
Put a vendor on the watch list at90% of tolerance80 · 90 · 95 % of tolerance
New business on an over-tolerance vendorRefer to the head of cyberRefer · Allow, note on file · Decline
Count dependencies found only through a providerOnOn · Off
Recovery after the vendor is back3 hours2 · 3 · 6 hours, when the application gives none
Quota share ceded30%, Orrin Re25 · 30 · 35 %
Event cover$20M xs $10M, Vauclair Re$20M xs $10M · $30M xs $10M
Who declares the event and signs the noteHead of CyberHead of Cyber · Chief Underwriting Officer
Connections

Works with the sources you already have

Vendor status pages and advisoriesevery five minutes, with vendor advisories and national alerts
Known-exploited vulnerabilities listthe public catalogue
Outside-in scansevery insured, refreshed nightly
Cyber applications and provider disclosurestechnology answers, read on bind and renewal
Your policy systemlimits, sublimits, waiting periods, endorsements
Emailbroker outreach and the event note, sent by people
What it changes

The difference, in numbers.

Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.

target
< 1hour
from vendor event to exposed insureds and limit total
By hand≈ 3 days
With agents< 1 hour
target
418of 418
dependencies shown with their evidence and date
418 of 418
with a cited source
typical
3–10%
of global cyber premium lost to one shared-vendor outage, July 2024
LowHigh

“demo” = seen in the working solution, on its sample cyber book · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (CyberCube estimate of the July 2024 outage: $0.4–1.5bn, 3–10% of about $15bn of premium; Munich Re, Cyber Insurance: Risks and Trends 2025; Corvus on dependent business interruption). People, companies, vendors and reinsurers named on this page are fictional — characters and sample data in the working solution.

Questions

What cyber underwriting and portfolio teams ask us.

What is cyber accumulation management?

It is knowing how much of a cyber book rides on one shared cause — a cloud region, an identity provider, a payroll vendor — so a single outage or breach does not surprise the board or the reinsurers. Cyber Accumulation Management keeps a standing map of shared vendors and the insureds that depend on each, with the limit and dependent business interruption sublimits exposed.

How does it know an insured depends on a vendor?

From three kinds of evidence: outside-in scans, answers on the cyber application, and managed-IT or software provider disclosures. Each dependency shows its evidence and date. Dependencies through a provider are kept separate, and you choose whether to count them.

How is the business interruption estimate made?

Interruption is the outage length plus the insured’s recovery time from its application (3 hours when it is silent). Hours up to the policy’s waiting period are not covered; income lost per hour from the application’s worksheet times covered hours, less the retention, is capped at the dependent-BI sublimit. A system failure counts only policies that cover it; a malicious attack counts all.

Does it account for reinsurance?

Yes. The estimate goes from gross to the quota share ceded, to net, to the event excess of loss cover, and shows the outage length at which the event cover attaches. Treaty terms are settings.

Does anything go to brokers or reinsurers automatically?

No. The agents watch, match, read policies, estimate and draft. The head of cyber declares the event and signs the note; nothing reaches a broker until an underwriter presses Send, and nothing reaches a reinsurer unsigned.

Can we set our own tolerances and assumptions?

Yes. The per-vendor limit tolerance, the watch-list threshold, how new business on an over-tolerance vendor is handled, recovery time, quota share and event cover, and who declares and signs are all settings. Assumptions are changed by the portfolio manager and versioned.

Which vendors does it watch?

Every vendor your insureds share — in the working solution, 11: cloud hosting, identity, endpoint security, email, payroll, ERP, managed file transfer, health records, managed IT and point of sale. You can add a vendor, and the dependency matcher looks for it in scans, applications and provider disclosures.

How long does it take to go live?

The Agentic Solution Engine builds and deploys it from your requirements and documents — your applications, scan feed, policy wordings and reinsurance programme — and it goes live once every quality gate has passed. We will walk you through it on your own book first.

See it on
your cyber book.

We’ll map your own shared vendors and run a past vendor event against your in-force book.