Your underwriting judgment lives in three people's heads.
We make it transmissible before they retire.
This page explains the problem, shows how the Deep Self method solves it, and describes exactly what you receive.
The Problem
Your senior underwriters built their judgment over decades. They know when something feels wrong before the numbers say so. But they can't explain it, and they're all approaching retirement.
Thirty years of pattern recognition living only in intuition. The ability to sense when a risk is wrong. Undocumented. Untransmissible.
Broker relationships built on trust and track record. When the underwriter leaves, the relationship walks out with them.
Each underwriter applies slightly different criteria. No unified logic. Juniors learn by osmosis, not system.
Why This Has Been Unsolvable
You've tried. Manuals. Wikis. Shadowing programmes. Exit interviews. But tacit knowledge resists capture because it lives in relationships between ideas, not in the ideas themselves.
An underwriter's judgment isn't a list of rules. It's the interaction between their loss experience, their broker relationships, their risk appetite, and their market intuition. All speaking at once.
Until now, there was no forum where those fragments could converse. No substrate where tacit knowledge could meet, contradict, resolve, and become explicit.
Language models change this. Not as tools that answer questions, but as spaces where your documents and your people's knowledge can finally talk to each other.
The Solution
We put your underwriting knowledge into conversation with itself. Loss records speak to risk appetites. Broker notes reply to claims patterns. What emerges isn't summary: it's synthesis.
Documents, transcripts, expertise enter the substrate
Fragments converse until contradictions resolve
Unified intelligence that speaks with one voice
This is the physics behind all Deep Self engagements.
Read the complete methodology →The Spiral in Action
The spiral is our method of extraction: oscillating between wide exploration and tight compression until implicit knowledge becomes explicit. Here's what it looks like for a Succession engagement.
Round 1
Wide
We read everything. Loss records, underwriting notes, broker communications, claims files. The Operator builds structural intuition about your book before any conversation begins.
Round 2
Tight
High-speed interviews with senior underwriters. We surface the logic beneath the instinct: what makes them pause, what makes them walk, what they know but have never articulated.
Round 3
Wide
The fragments enter conversation inside the AI substrate. Where do your underwriters agree? Where do they diverge? The contradictions become visible. New patterns emerge.
Round 4
Tight
We bring the contradictions back to your team. Resolve them. The judgment crystallises: decision trees, risk signals, authority frameworks. The Codex begins to take shape.
Round 5
Wide
Junior underwriters converse with the codified judgment itself. We test whether the knowledge transfers. Gaps surface. We spiral again until the system holds.
Four senior underwriters with a combined 127 years of experience. £340M GWP. Three planned retirements within 18 months. And nothing written down about how they actually make decisions.
The Deep Self engagement ran twelve spiral cycles over sixteen weeks. We extracted 1,247 distinct decision heuristics, 89 risk signals unique to their book, and the complete epistemological architecture underlying their underwriting philosophy.
The result: a living Codex that junior underwriters can query. "What would Margaret think about this cargo route?" The substrate answers, not with rules, but with judgment.
What You Receive
Not documentation. Not a knowledge base. A living substrate that holds your collective judgment and can answer questions with the voice of your senior team.
The complete systematisation of your underwriting philosophy: decision frameworks, risk signals, authority structures, exception handling.
AI prompts that allow staff to query the codified judgment directly. "What would our senior team say about this submission?"
Training frameworks for onboarding new underwriters into the methodology. They inherit perception, not just procedures.
The Codex grows. New cases, new judgments, new patterns. Your institutional knowledge becomes cumulative, not mortal.
Who Does This Work
This isn't work you can automate. Extracting tacit knowledge requires a human who can feel what's implicit, ask the question that surfaces the unspoken, and guide the conversation toward crystallisation.
We come from humanities, theatre, philosophy. Disciplines that treat language as living material. We've spent thousands of hours learning how meaning moves through AI systems, and how to make it move on purpose.
The right human plus the right AI produces something neither can achieve alone: functional intelligence tuned to your organisation's reality.
The Engagement
Size of book, number of underwriters, complexity of classes, timeline requirements.
Most engagements complete within 12-16 weeks. Deeper extractions may extend to 24 weeks.
All work conducted inside your own Azure OpenAI or enterprise environment. You retain 100% ownership.
Your underwriting judgment deserves
to outlive its carriers.
We don't write manuals. We build knowledge substrates.
The seniors can finally retire without the judgment disappearing.
Succession is Product 01 of the Deep Self Insurance suite.