Non-life insurance · Actuarial intelligence

The reasoning layer
your reserving
cycle has been missing.

Decisio is an AI copilot trained on actuarial reasoning — drafting reserving models, stress-testing assumptions, and generating regulatory-ready documentation for non-life actuaries.

∼15K
FCAS / FIA Fellows worldwide
4–8 wks
Typical reserve cycle length
>70%
Time lost to mechanical tasks
Productivity multiplier per Fellow
// Chain-ladder LDF selection Bornhuetter-Ferguson a priori benchmarking // Cape Cod method auto-draft Solvency II ORSA generation // Lloyd's SBF documentation Tail factor selection with rationale // Social inflation stress scenarios NAIC Statement of Actuarial Opinion // IBNR triangle assembly & validation Reserve cycle audit trail // Chain-ladder LDF selection Bornhuetter-Ferguson a priori benchmarking // Cape Cod method auto-draft Solvency II ORSA generation // Lloyd's SBF documentation Tail factor selection with rationale // Social inflation stress scenarios NAIC Statement of Actuarial Opinion // IBNR triangle assembly & validation Reserve cycle audit trail

The most expensive talent in insurance
spends its days on spreadsheets.

Qualified Fellows (FCAS, FIA) command £200–600K packages. Their time is the most constrained resource in any P&C operation. Yet the majority of a reserve review cycle is consumed by tasks that require precision, but not judgment.

01
Critical
Triangle Construction
Manual extraction from claims systems, pivot tables, and reconciliation loops. A single data error propagates through all downstream methods — often discovered only at sign-off.
02
Critical
Method Selection
No principled framework for choosing between BF, LDF, Cape Cod per line of business. Selection relies on tribal knowledge and the seniority of whoever last touched the file.
03
High
Assumption Documentation
Actuarial memos written ad-hoc after the model is complete — often incomplete under Solvency II and NAIC requirements. Regulators are asking harder questions than ever.
04
High
Stress Testing
Bespoke scenario scripts written per review cycle. Rarely reproducible. Regulators increasingly mandate formal stress frameworks, yet most teams still script these by hand each quarter.
05
High
Regulatory Submissions
Reformatting reserve reports for Lloyd's, PRA, EIOPA, and NAIC is a 2–3 day clerical exercise per jurisdiction per cycle. The content is identical — the formatting is not.
06
Medium
Peer Review Trail
Second-actuary sign-off is done on printouts. No structured audit trail for assumption changes between cycles. When regulators ask why a tail factor moved, reconstructing the rationale takes days.

Four modules. One complete
reserve review cycle.

Decisio is not a dashboard you visit. It is a workflow-embedded copilot that lives inside your environment — Excel add-in, VS Code extension, or browser workbench — and provides intelligent assistance at every stage.

Triangle Intelligence

Connects to raw claims extracts. Automatically detects development periods, identifies data quality issues — thin cells, negative incrementals, IBNR vs paid splits — and assembles paid, incurred, and case reserve triangles with full provenance metadata.

  • Detects data anomalies before they propagate
  • Splits paid, incurred, and case reserve automatically
  • Flags statistical outliers with natural-language explanations
  • Connects to CSV, Excel, or database extracts
  • Full cell-level audit trail with source references
Paid loss development triangle — Motor TP
AY 12m24m36m48m60m
20194,2187,84110,22011,89012,441
20203,9027,1209,88011,344~12,018↑
20214,5608,21011,102~12,614↑~13,371↑
20224,8919,044~12,198↑~13,880↑~14,713↑
20235,220~9,640↑~12,996↑~14,791↑~15,682↑
Actual
AI-projected
Model Drafting Copilot

Given a triangle and line-of-business context, Decisio drafts a complete reserving model: selects methods with written rationale, selects tail factors, benchmarks against industry development patterns, and writes the first-draft actuarial memo. You review and override — the system learns from every decision.

  • Selects from CL, BF, Cape Cod, Clark LDF with rationale
  • Industry development pattern benchmarking
  • Auto-written actuarial memo explaining every choice
  • Override learning — system calibrates to your style
  • Multi-year comparison with prior cycle assumptions
METHOD SELECTION RATIONALE — PROPERTY DAMAGE
Decisio recommends Bornhuetter-Ferguson for accident years 2021–2023, and volume-weighted chain-ladder for 2018–2020.

Rationale: The 2021–2023 accident years exhibit thin credibility at longer development periods (cell counts <15). BF a priori loss ratios are anchored to the industry Property TP benchmark of 68.2% per CAS 2023 development study. Chain-ladder is appropriate for 2018–2020 where development patterns are fully credible at all periods.

Tail factor: 1.048 selected, consistent with the 5-year weighted average. No structural reason to deviate from industry tail.
Stress Test Engine

A formal stress scenario library — large loss emergence, social inflation shocks, settlement pattern shift, cat loss development — run across the full model in seconds. Produces waterfall charts, sensitivity tables, and narrative summaries identifying which assumptions are the principal drivers of reserve uncertainty.

  • Pre-built scenario library: social inflation, cat, litigation
  • Runs all scenarios simultaneously — results in seconds
  • Identifies primary sensitivity drivers per line
  • Reproducible — same scenario, same output, every cycle
  • Narrative summary for board pack integration
Reserve sensitivity — Base vs stress scenarios
Base case
£ 142.4M
Social inflation +15%
£ 163.8M
Tail elongation +20%
£ 157.2M
Cat adverse development
£ 181.1M
Settlement acceleration
£ 128.9M
Regulatory Document Factory

Transforms the completed model into jurisdiction-specific output: Lloyd's SBF, UK PRA Section 57, NAIC Statement of Actuarial Opinion, Solvency II ORSA. Understands the structural requirements of each format. Actuaries edit the draft rather than write from scratch — a 2-day task becomes 90 minutes.

  • Lloyd's SBF — full market reform format
  • UK PRA / EIOPA Solvency II ORSA
  • NAIC Statement of Actuarial Opinion
  • Regulatory reference citations from live knowledge base
  • Version-tracked across cycles for PRA continuity reviews
Active output — Lloyd's SBF Section 4.3
Lloyd's SBF PRA S57 NAIC SAO SII ORSA
4.3 RESERVE ADEQUACY — CLAIMS PROVISIONS
The Syndicate's carried reserves at 31 December 2024 represent the appointed actuary's best estimate of the ultimate cost of claims incurred to date. Reserves have been set in accordance with Lloyd's Valuation of Liabilities Rules (2024).

[AI draft — review for Syndicate-specific adjustments]
⚡ Generated in 4.2s · 2 jurisdictions · 1 click to export

From raw data extract
to filed report. In days, not weeks.

Step 01
Data Ingestion & Triangle Assembly
Connect claims extract. Decisio assembles paid, incurred, and case reserve triangles with full data quality checks and provenance metadata.
Step 02
Method Selection & Model Draft
AI selects reserving methods per line with written rationale. Drafts the full model including tail factors, benchmarked against industry patterns.
Step 03
Actuary Review & Override
Fellow reviews every decision. Overrides are captured with context. The model learns your team's judgment patterns across cycles — becoming uniquely yours.
Step 04
Stress Test & Sensitivity
Formal scenario library runs in seconds. Results package includes waterfall charts, sensitivity drivers, and board-ready narrative summary.
Step 05
Regulatory Document Generation
One-click transformation to Lloyd's SBF, PRA, NAIC, or ORSA format. Edit the draft — not the blank page. File with confidence.

Four structural advantages
no wrapper can replicate.

Moat 01 — Domain depth
Not a GPT wrapper. A reasoning engine.
Generic AI produces plausible-sounding actuarial text with subtle methodological errors — wrong tail factor logic, misapplied BF a priori selection. A Fellow spots these immediately and the trust is gone. Decisio's numerical kernel runs all triangle math deterministically. The LLM reasons about outputs it did not compute. The gap cannot be closed by a chatbot.
Moat 02 — Override learning
The system becomes uniquely yours over time.
Every override captured with context. After 3–4 reserve cycles, Decisio is calibrated to your team's judgment patterns, house style, and risk appetite in a way no competitor can replicate without your data. Switching cost becomes prohibitive — it would take a year for a replacement system to reach the same standard.
Moat 03 — Regulatory currency
A living knowledge base that keeps pace with regulators.
Lloyd's market reform, PRA SS3/17 updates, NAIC ASOP revisions — regulatory requirements shift constantly. Decisio maintains a continuously updated regulatory knowledge base. Once actuaries rely on it for submission formatting, they cannot switch without rebuilding the entire regulatory mapping from scratch in a competing tool.
Moat 04 — Professional trust network
Credentialed profession. Reputation is everything.
The actuarial community is insular, credentialed, and risk-averse. One CRO at a Lloyd's managing agent who publicly endorses Decisio reaches 40 peers within months. Trust capital in a tiny professional community where reputation is the primary decision criterion is the hardest moat of all to build — and the hardest to destroy once earned.

A small pool. An enormous
value concentration.

The buyer pool is deliberately small — credentialing takes 7–10 years and pass rates on later exams are in the single digits. But these professionals operate at the centre of reserving decisions that touch billions in capital.

Primary buyer
Chief Actuary / Head of Reserving
London Market managing agent, mid-tier UK non-life insurer, or Bermuda reinsurer. Team size 3–12. Running quarterly GAAP and Solvency II reserve reviews. Budget authority $200K–$800K/yr. Evaluated on reserve accuracy and regulatory compliance.
Secondary buyer
Actuarial Consulting Practice
Big 4, Milliman, WTW, Gallagher Re. Running reserve reviews for 10–50 carrier clients annually. Decisio as a practice tool multiplies billable throughput without headcount. Licensing per review, not per seat.
Phase 1 · London Market
Phase 2 · Bermuda + US E&S
Phase 3 · Global
Pricing tiers
Analyst
Small syndicates · startup insurers · up to 5 LoB
$60K / yr
Practice
Mid-tier carriers · consulting firms · unlimited LoB
$180K / yr
Enterprise
Large carriers · reinsurers · on-premise option
$400–700K / yr
Consulting Seat
Per-actuary per-review · scales with practice volume
$12K / seat
Series A milestone
$5.5M ARR
at 20 Practice + 5 Enterprise accounts

London Market first.
Then the world.

Months 0–6 · Design partner phase
Embed with 3–5 actuarial teams
Deep R&D partnerships with influential Fellows. No revenue — pure product build. Goal: a model drafting copilot that passes the "I'd use this" test from an FCAS or FIA. Every override is training data.
Months 6–12 · Credibility launch
CAS or GIRO paper + first paid pilots
Publish a technical paper at CAS or present at GIRO demonstrating methodology and benchmark accuracy. This is the primary trust channel in a credentialed profession. Convert design partners to $30–60K pilots.
Months 12–24 · London concentration
10 managing agent accounts
One actuarial sales specialist — an ex-Fellow who knows the workflow. Target Lloyd's managing agents directly: identical reserving cycle, templated onboarding. Pursue endorsement from a Lloyd's market body.
Year 3+ · Bermuda + US expansion
Global via reinsurance relationships
London managing agents have retrocession and reinsurance relationships with Bermuda and US E&S carriers. Use these as warm referral paths. Localize regulatory factory for NAIC. Pursue Milliman or WTW practice partnership.
Expansion surface
Natural expansion · Pricing
Reserving → Pricing Intelligence
The same triangle data that drives reserving is the foundation of pricing actuarial work. A Pricing Intelligence module doubles the addressable user base without changing the ICP or the technical stack.
Strategic expansion · Capital
Reserve Uncertainty → Capital Models
Solvency II internal models and Lloyd's capital setting both consume reserve uncertainty outputs. Decisio's stress scenarios are a direct input to capital modeling — a pathway to the CRO and CFO as second buyers.

Built by someone who has lived
the problem.

SK
Shivam Kashyap
Founder & CEO · Decisio

Shivam built Decisio from a simple observation: the most credentialed technical professionals in the insurance industry spend the majority of their working hours on mechanical tasks that produce no insight. Triangles assembled by hand. Memos written from blank pages at midnight before sign-off deadlines. Stress scenarios re-scripted every quarter from memory.

His background spans actuarial modeling, machine learning, and the specific institutional knowledge of how the London Market actually operates — the informal networks, the regulatory cadence, the quiet trust that flows between Fellows at GIRO. Decisio is built to earn its place in that world through methodological rigour, not a sales pitch.

Non-life insurance Actuarial modeling ML / fine-tuning London Market Reserve methodology Regulatory compliance
The team we are building
Technical Co-founder
ML Engineer · NLP
Fine-tuning and domain model architecture. Must understand why chain-ladder arithmetic must never be done by an LLM.
Actuarial Co-founder
FCAS or FIA · 8–12 yrs
Product without a credentialed Fellow is immediately distrusted. Needed for CAS/GIRO publishing and professional network access.
First Sales Hire
Ex-Fellow, commercial role
Actuaries do not buy from people who do not understand reserve adequacy. Peer credibility is mandatory for the first sale.
Early access · London Market priority

Your next reserve cycle
starts with Decisio.

We are onboarding design partners from Lloyd's managing agents, UK non-life carriers, and actuarial consulting practices. Limited access. Fellows prioritised.

Data residency
SOC 2 Type II · VPC isolated
On-premise option
Available for Enterprise
Implementation
Onboarded in one reserve cycle