Capturing Institutional Memory
A knowledge base covering the firm's edge cases and compliance scenarios — queryable in natural language with firm-specific responses.
Five-week build covering initial content population, followed by a four-week period of supervised use with iterative content additions before the system was treated as the primary reference for junior advisers.
The challenge
The firm's most valuable knowledge was informal and inaccessible — junior advisers were either escalating questions that had already been answered or, worse, guessing.
As the team grew, the asymmetry between what senior advisers knew and what junior advisers could access became operationally expensive in two ways. First, escalation load: junior advisers routinely brought questions to senior advisers that the senior adviser had answered, in some form, before — consuming time that should have been spent on client work. Second, quality risk: in situations where escalation felt awkward or the senior adviser was unavailable, junior advisers made judgment calls based on incomplete information. The specific nature of the firm's client base — Turkish nationals navigating UK property law with complex cross-border financial structures — meant that the gap between a correct answer and a plausible-sounding incorrect one could have real consequences: a missed compliance step, an incorrectly framed offer, a client relationship damaged by advice that contradicted what they had heard from their accountant in Istanbul. General-purpose LLMs were being used as a workaround, but they produced generic UK property guidance that frequently failed to account for the firm-specific protocols, the Turkish-client-specific considerations, or the nuances of cases the firm had actually handled.
What we did
The approach
We designed and built a structured knowledge base seeded from the firm's own case history: documented edge cases, compliance scenario resolutions, client situation patterns, and adviser judgment calls that had been validated over time. The knowledge base is queryable in natural language — a junior adviser can describe a situation in their own words and receive a referenced response drawn from how the firm has handled similar situations, not a generic answer from the open web. Each response includes the source case context, the applicable protocol or principle, and a flag indicating whether the situation requires senior adviser review before action is taken. The system is designed to grow: every new edge case that is resolved by a senior adviser is captured in a structured format and added to the base.
Key findings & actions
Edge case library
structured records of resolved unusual situations, organised by scenario type (visa status, source of funds, family trust structures, tenancy disputes, compliance edge cases)
Protocol and principle layer
documented firm-specific procedures for recurring situation types, written to be actionable rather than aspirational
Natural language query interface
junior advisers describe situations in plain language and receive referenced, firm-specific responses
Confidence and escalation flagging
each response includes a signal indicating whether the situation is within standard protocol (proceed with confidence) or requires senior review before action
Capture workflow
every new resolved edge case is structured and added to the base by the resolving adviser before the case is closed
How we worked
Scope
Case history audit, edge case taxonomy design, knowledge base architecture, initial content population from documented cases, query interface configuration, and escalation logic design.
Timeline
Five-week build covering initial content population, followed by a four-week period of supervised use with iterative content additions before the system was treated as the primary reference for junior advisers.
Operating model
Senior advisers were involved in validating every piece of content before it was added to the base — we did not populate it from case notes alone. The distinction between 'we handled it this way once' and 'this is the firm's approach to this situation' was made explicit for every entry, and only the latter was treated as knowledge base content.
Outcomes
What changed
A knowledge base covering the firm's edge cases and compliance scenarios — queryable in natural language with firm-specific responses.
Senior adviser escalation load reduced by an estimated 60% for routine edge cases within the first 60 days
questions that had been answered before were being answered by the system rather than by interruption
Response consistency improved: junior advisers across the team now give substantively the same answer to the same situation type, rather than responses that vary based on which senior adviser they happened to reach
Quality risk reduced: the escalation flag system means situations that genuinely require senior judgment are identified as such, rather than resolved by junior advisers who didn't know to escalate
Onboarding time for new advisers reduced: the knowledge base functions as a structured induction resource, meaning new hires reach operational competency faster and with less dependency on informal mentoring
Institutional knowledge made durable: the firm's accumulated case intelligence is no longer at risk of loss when a senior adviser leaves
it is captured, structured, and retained in the system
Governance
Trust, collaboration & governance
The system produces referenced responses — every answer links back to the source case or protocol, so advisers know what they are relying on and can exercise their own judgment about its applicability
Escalation flags are calibrated conservatively: when a situation is ambiguous, the system recommends senior review rather than confident resolution — we designed for the cost of false confidence, not the cost of unnecessary escalation
Content governance is owned by the senior advisers, not by us — they decide what goes in, what is updated, and what is retired
Reframe
Senior adviser judgment was invisible and at risk of walking out. Making it queryable is a governance decision.
Across every engagement, the goal is the same: engineer a system that makes better decisions — faster, more consistently, and at scale — than the process it replaces.
Next steps
Related services
Start a discovery
Most engagements begin with a conversation about context.
We do not send a proposal before we understand the problem. Start by telling us about your decision context — we will identify the highest-leverage intervention areas before any scope is agreed.