Advanced Modelling Services

Advanced Modelling & AI Services | Paul Faulkner — The Rogue Protocol
Advanced Modelling & AI Services

Thirty years
of models
that had to
be right.

Not a consultant with a chatbot. An operator with three decades of institutional model construction behind him and a purpose-built AI stack in front of him. Banking. Treasury. Payroll. Satellite telemetry. Online gambling reconciliation. Forensic accounting. The domain changes. The standard does not.

Institutional Provenance
The models that ran the institutions
30
Years building models that carried institutional liability
1996
Excel 5. VBA from the beginning. Not adopted — forged.
6/s
Transactions per second. Live financial analysis. Ladbrokes.
4
Cities. JPMorgan FX reporting. One undocumented Access stack.
Domains modelled
Banking mortgages Treasury services FX reporting Payroll Sales compensation MEO satellite telemetry Online gambling reconciliation Forensic accounting Structured credit Cryptocurrency markets Automated reporting Legacy EUC recovery

The models were not
academic exercises.

Every model in this practice’s history carried institutional liability. The mortgage model at Bradford & Bingley was not a prototype — it was the infrastructure running the book. The FX reporting stack at JPMorgan was not a proof of concept — it was the system four cities depended on for month-end close. When a model is wrong in those environments, the board asks questions that cannot be answered with “we’re investigating.” That is the standard this practice builds to. It has not changed.

01
Financial Services & Banking
Mortgage book infrastructure at Bradford & Bingley — the same models that documented the subprime risk before the market understood it existed. FX treasury reporting at JPMorgan across London, Chicago, New York, and Singapore. Institutional finance at SG Kleinwort Hambros and PwC crisis assignments.
Mortgage modelling FX treasury Structured credit Crisis accounting
02
Live Operations & High-Frequency Data
Full financial analysis stack at Ladbrokes processing six transactions per second in a live betting environment. Zero tolerance for error. The reconciliation problem at that frequency is a different class of challenge from end-of-day batch processing. MEO satellite network usage data modelling — applying the same analytical discipline to a completely different data domain.
Real-time reconciliation High-frequency data Satellite telemetry VBA automation
03
Compensation, Reporting & Legacy Recovery
Payroll and sales compensation modelling — actuals and new plan design — where the model determines what people are paid and must survive the scrutiny of every manager whose team it affects. Monthly reporting packs fully automated end to end. Zero-knowledge legacy recovery: inheriting systems with no documentation and no institutional memory.
Payroll models Sales compensation Automated reporting Legacy EUC

Since 1996. Excel 5. VBA from the first version that made it viable. Not adopted as a skill. Forged as the primary tool for institutional model construction across three decades and a dozen domains. The AI stack is the newest layer. The analytical standard underneath it is the oldest thing on this page.

Five capabilities.
One standard.

Every engagement begins with a diagnostic. The service line is determined by what the diagnostic finds — not by what the client assumed they needed when they first made contact. The client describes the problem. The diagnostic determines the entry point.

Service Line 01
Forensic Model Review
The model someone else built. Reviewed to the standard it should have been built to.

Every CFO being handed a DCF by an M&A advisor, every fund manager receiving a valuation model from a research house, every board approving a strategic recommendation built on a financial model — they share the same problem. They cannot evaluate the model themselves. They need someone who can, and who will tell them the truth about it.

Thirty years of building models means knowing exactly what corners get cut, what assumptions get buried, and what the model was constructed to prove rather than test. That pattern recognition cannot be taught. It is the accumulated record of being inside models when they failed.

  • Full assumption audit — every input, every driver, every sensitivity
  • Circular logic and structural error identification
  • Break condition analysis — what combination of inputs causes the model to fail
  • Probability-weighted scenario construction to replace point estimates
  • Independence assessment — was this model built to answer a question or prove a conclusion
  • Board-ready forensic report — findings stated plainly, not buried in caveats
£5,000+
Fixed Fee · Scope Dependent · 1–2 Weeks
Request Review →
Who this is for
Allocator
Told to buy. Never shown the model.
You have been given a price target, a recommendation, or an allocation thesis. You have not been given the conditional framework that supports it. Before the commitment is made, the model should be reviewed by someone who did not build it and has no interest in the conclusion it reaches.
Board / Director
Approving a recommendation built on a model you cannot evaluate.
The strategic recommendation arrived with supporting financial analysis. The analysis looks credible. The question is whether it was built to answer a question or to support a conclusion that was already reached. These require different levels of scrutiny.
CFO / FD
The acquisition model that needs an independent eye.
M&A valuation models are built by people who are paid when the deal completes. The incentive structure is not aligned with surfacing the reasons not to proceed. Independent forensic review is the mechanism that rebalances it.
Service Line 02
Institutional Model Construction
Built to survive scrutiny. Not built to support a narrative.

The Show Me The Model framework — thirteen conditions, eight structural contradictions, four scenarios, probability-weighted expected value — is the public demonstration of what institutional-grade model construction looks like. That is the private standard applied to every bespoke build.

Sectors covered: financial services, cryptocurrency and digital assets, compensation and incentive design, treasury and cash management, operational and commercial modelling, regulatory capital and stress testing.

  • Conditional framework construction — not point estimates, scenario distributions
  • Sensitivity and Monte Carlo analysis where the decision warrants it
  • Falsification conditions built in — not “still early” escape hatches
  • Full documentation: assumptions, sources, methodology, limitations
  • Excel and VBA delivery — institutional-grade, auditable, maintainable
  • Interactive HTML overlay available for board or investor presentation
£8,000+
Fixed Fee · Scoped After Diagnostic · 2–4 Weeks
Initiate Enquiry →
Recent applications
Cryptocurrency & Digital Assets
The model that was never produced.
Show Me The Model is the public version: the Bitcoin $1M thesis modelled to the standard that should have been applied fifteen years ago. The private version is the same framework applied to your specific position — asset, timeframe, conditions, probability-weighted expected value.
Compensation Design
Does the incentive model do what the board thinks it does?
Payroll and sales compensation modelling across actual and proposed plan structures. Every PE-backed business has an incentive architecture that was designed once and never stress-tested against the behaviours it actually produces.
Treasury & Cash Management
The cash position under stress scenarios.
Treasury modelling that maps the cash position under adverse conditions — not just base case. What does the business look like at 70% revenue, at 150% capex, at a 90-day receivables extension? The stress test belongs before the covenant breach, not after.
Service Line 03
Legacy EUC Recovery & Documentation
Zero-knowledge system reconstruction. Ghost logic identified. FCA-ready documentation produced.

The person who built the system has left. The documentation does not exist. The system produces the right numbers — probably — but nobody can prove it, explain it, or modify it without risking something nobody can predict. The FCA is asking questions. The audit function is asking questions. The board is asking questions.

Using Gemini’s two-million-token context window, an entire institutional codebase is ingested in a single analytical session. The AI council maps the logic. Thirty years of experience with these systems identifies what the AI cannot — the ghost rules that evolved over fifteen years of edge cases and were never written down.

  • Full VBA macro audit — every module, every dependency, every edge case
  • MS Access schema mapping — tables, relationships, queries, forms
  • Ghost logic identification — the undocumented rules that are actually running the system
  • FCA EUC compliance documentation — the standard the regulator requires
  • Reconciliation framework — proving the system produces what it claims to produce
  • Migration pathway — stabilise in place, refactor, or transition to Python/SQL/cloud
£15,000+
Fixed Scope · 1–4 Weeks · AI-Accelerated
Request Assessment →
The three scenarios
Regulatory Audit
The FCA has flagged the model as unaudited.
The gap between “it works” and “we can prove it works correctly to the regulator’s standard” is exactly where legacy EUC forensics operates. The documentation that should have been produced when the system was built can still be produced — forensically, from the system itself.
Reporting Failure
The month-end pack is producing numbers nobody can reconcile.
The senior accountable is being asked questions they cannot answer. The person who could answer them left eighteen months ago. The system is producing output but the logic that produces it is invisible to everyone currently in the building.
Migration Risk
The new system doesn’t match the old one’s outputs.
Three months into the modernisation programme, the discrepancy has appeared. Nobody knows why because the ghost logic was never mapped before the migration began. It can still be found — in the legacy system, before the discrepancy finds the regulator.
Service Line 04
Bespoke Interactive Intelligence Tools
The deliverable is not a PDF. It is an instrument.

The eighteen forensic tools on paulfaulkner.com are the public demonstration of what this service produces. Each one takes a specific analytical question — “does this price target require something the world has never done?”, “what does your stop-loss actually advertise to the market?”, “what verdict does your balance sheet actually produce?” — and turns it into an interactive instrument a user can operate themselves.

The same capability is available as a bespoke commissioned service. A client’s specific analytical question, modelled and delivered as an interactive HTML tool — embedded in an investor relations page, a board pack system, a due diligence data room, or a client-facing research platform.

  • Bespoke analytical logic built to the client’s specific question
  • Full Rogue Protocol design system or client brand implementation
  • Live data integration where applicable — market prices, on-chain data, regulatory feeds
  • Embeddable via iframe or delivered as standalone hosted tool
  • Board-ready: operates without technical knowledge, produces defensible output
  • Scenario outputs exportable — the tool produces the evidence, not just the conclusion
£3,500+
Per Tool · Scope Dependent · 1–2 Weeks
Commission Tool →
Applications
Investor Relations
An interactive model that lets investors stress-test the thesis.
Rather than a static PDF with point estimates, an embedded interactive model where the investor can adjust assumptions and see how the output changes. The transparency builds credibility. The interactivity replaces the analyst call.
Due Diligence
A scenario tool embedded in the data room.
A bespoke model built around the specific deal — revenue scenarios, covenant stress tests, exit multiple sensitivity — delivered as an interactive instrument rather than a spreadsheet that requires explanation. The acquirer operates it. The questions answer themselves.
Research Platform
The analysis your clients can run themselves.
A forensic tool built around your research methodology — distributed to clients as an interactive instrument rather than a written report. The client applies your framework to their own inputs. The output is theirs. The methodology is yours.
Service Line 05
Expert Analysis & Litigation Support
Court-ready documentation produced by construction. Primary record from the first handshake.

The Operational Protocol produces primary-record documentation by design. Every engagement generates a timestamped, citable, audio-grounded record of what was said, what was agreed, and what was built. In a dispute, that record is not reconstructed after the fact. It exists from day one.

Expert analysis engagements apply the same forensic standard to financial modelling disputes, structured product mis-selling claims, EUC system failures, and valuation methodology challenges. The Bradford & Bingley provenance — designing the subprime products, documenting the risk, being ignored — is the analytical foundation for structured credit and mortgage litigation support.

  • Financial model dispute analysis — what the model actually produced versus what was represented
  • Structured product and mortgage mis-selling — structured credit expertise from primary construction experience
  • EUC system failure analysis — what the system did, what it should have done, where it diverged
  • Valuation methodology challenge — independent assessment of whether the methodology was appropriate
  • Primary record infrastructure — Operational Protocol documentation from engagement start
  • Court-ready report format — findings stated clearly, methodology documented, limitations acknowledged
£10,000+
Scoped After Initial Consultation · NDA Required
Initiate Confidential Enquiry →
Engagement contexts
Financial Modelling Dispute
The model produced an output that cost someone money.
Whether the question is whether the model was fit for purpose, whether the methodology was appropriate, or whether the output was represented accurately — the forensic analysis starts from the model itself, not from the parties’ accounts of what it did.
Structured Credit
Mis-selling. Risk documentation. What was known and when.
The Bradford & Bingley record is not background context. It is the analytical foundation. Designing the products, modelling the book, documenting the systemic risk before 2008 — that is the experience base for structured credit litigation support.
Regulatory Investigation
Pre-emptive mapping before the first letter arrives.
The most valuable forensic work is done before the regulator makes contact — mapping the exposure, preparing the documentation, identifying what the investigation will find before it finds it. The window between “we think we may have an issue” and “the FCA has written to us” is where this engagement operates.

The deliverable
is not a PDF.

The standard consulting deliverable is a document. It is read once. It is filed. When someone needs to stress-test an assumption six months later, they commission another document. The interactive intelligence tool is a different object entirely. It is an instrument — one that a non-technical user can operate, that produces defensible output from their own inputs, and that does not require the analyst to be in the room.

The Standard Deliverable
A document with a conclusion.
Fixed assumptions. Point estimates. A conclusion derived from inputs the reader cannot change. When the inputs change — and they always do — the document is wrong. A new engagement is required. The analyst is in the loop for every update. The deliverable degrades from the moment it is produced.
The Interactive Intelligence Tool
An instrument with a framework.
Variable inputs. Scenario distributions. A framework that produces outputs from the user’s own assumptions — which they can adjust, stress-test, and interrogate without commissioning a new engagement. The analyst’s logic is embedded in the instrument. The instrument operates independently. The deliverable compounds in value rather than degrading.
01
Forensic Diagnostic Tools
A specific analytical question — a price target, a valuation methodology, a balance sheet stress test — turned into an interactive instrument. The user enters their own inputs. The tool applies the forensic framework. The output is defensible and citable. Live examples: BTC Reality Check, Balance Sheet Suitability Assessment, Show Me The Model companion tools.
02
Scenario & Sensitivity Overlays
A financial model delivered as an interactive overlay rather than a static spreadsheet. Tornado charts, Monte Carlo outputs, sensitivity grids — rendered as interactive HTML instruments embedded in a board pack system, investor relations page, or data room. The board operates the model. The questions answer themselves.
03
Research Platform Instruments
A bespoke tool built around a research methodology — distributed to clients as an interactive instrument rather than a written report. The client enters their own inputs. The framework applies your analytical logic. The output is theirs. The intellectual property is yours. The engagement deepens rather than concluding.
Live Public Examples — paulfaulkner.com/tools/

The eighteen forensic tools currently on paulfaulkner.com are the public demonstration of this service: BTC Cherry Picker (date-range return forensics), MVRV Structural Ceiling Calculator (on-chain cycle positioning), BTC Balance Sheet Suitability Assessment (seven-condition director diagnostic), Hyperbitcoinisation Conditions Stack (six-condition falsification framework), Correlation Collapse (safe-haven claim audit), Cash Erosion Audited (ten-error forensic audit of the 11% claim), and twelve others. Each one takes a specific analytical question and turns it into an instrument a non-specialist can operate. That is the capability available as a bespoke commissioned service.

What the stack adds.
What it cannot replace.

The AI council — Gemini, Claude, DeepSeek, Mistral, with a proprietary Ollama layer — augments every engagement. It does not replace the operator. Twenty-five years of institutional pattern recognition determines which three sentences in a two-hour transcript are the ones that will kill the project in six months. The AI cannot do that. What it does is make the work tractable at a price point that was not viable before.

New Capability
Entire codebase ingested in one session
Gemini’s two-million-token context window means a legacy Access database, a VBA codebase, or an entire engagement history can be ingested in a single analytical session. What previously required weeks of manual audit now takes days. The speed has changed. The standard has not.
New Capability
Regulatory document analysis at scale
FCA consultation papers, company filings, prospectuses, annual reports — ingested in full, cross-referenced against each other, and analysed for the specific exposure the client needs mapped. The forensic work that previously required a team of analysts is now tractable for a single-operator practice at a fraction of the cost.
Structural Constraint
The AI is red-teamed before delivery
Every output the AI council produces is challenged against the primary record before it reaches the client. The transcript is the ground truth. The model’s synthesis is cross-referenced against what was actually said. Contradictions flagged. Gaps identified. Assumptions disconfirmed before they become deliverables. The red team is not a quality check at the end. It is structural.
Data Sovereignty
Client data stays off public APIs
A custom Ollama instance runs open-source models locally. Client data does not touch a third-party API unless explicitly routed and approved. For financial services clients, this is not a nice-to-have. It is the difference between AI-augmented analysis and a regulatory exposure created by the tool used to do the analysis.

▸ The differentiator is not the AI stack. Every consultant has access to the same models. The differentiator is thirty years of institutional pattern recognition directing the stack — knowing which question to ask, which output to challenge, which three sentences in the transcript are the ones that matter. The AI multiplies the operator. It does not replace the judgement. Engagements are accepted on a selective basis. Not every problem is the right fit. Not every client is the right client. That selectivity is itself part of the standard.

Every engagement
begins with the diagnostic.

The diagnostic determines the service line. Not the other way around. The client describes the problem. The diagnostic determines the correct entry point, scope, and price. The diagnostic fee applies as credit toward the subsequent engagement.

Capacity note: Advanced Modelling & AI Services engagements are accepted on a selective basis. The practice is a single-operator model by design — no juniors, no intermediaries, no work produced by a team and signed off by a senior. That is the constraint. It is also the value. Enquiry does not guarantee engagement. The diagnostic determines fit in both directions.

Entry Point — All Engagements
Forensic Diagnostic
£5,000
Fixed Fee · 90 Minutes + Written Brief

A structured diagnostic session followed by a written forensic brief — 10 to 20 pages — identifying what the problem actually is, what the material risks are, and whether a full engagement is warranted. Two weeks of follow-up access. Full fee credited against the subsequent engagement.

Book Diagnostic →
Core Service — Most Engaged
Defined-Scope Analysis
£8,000+
Fixed Fee · 2–4 Weeks · Scope After Diagnostic

The full modelling or forensic review engagement. Forensic model review, institutional model construction, regulatory document analysis, or interactive tool commission. Scope agreed after the diagnostic. Fixed price. No junior intermediaries. Full written deliverable.

Initiate Enquiry →
Ongoing — Retainer Available
Standing Analytical Capacity
£5,000+
Per Month · 30-Day Notice · Tiered

Monthly retainer providing priority access and allocated analytical capacity. Suitable for family offices, active portfolio managers, and institutions with ongoing forensic modelling requirements. Retainer structure detailed on the Services page.

View Retainer Structure →

The model should have
been built to survive
the question it answers.

All enquiries are treated as strictly confidential. NDA available before any substantive discussion. Engagements are accepted on a selective basis — the diagnostic determines the entry point.

NDA Available on Request · Senior Operator Only · No Junior Intermediaries · Engagements Accepted Selectively · nOS v1.0 Deployed