USMAN’S INSIGHTS
AI ARCHITECT
  • Home
  • About
  • Thought Leadership
  • Book
Press / Contact
USMAN’S INSIGHTS
AI ARCHITECT
⌘F
HomeBook
HomeBookChoosing and Combining Methods
Previous Chapter
Two Platforms, One Paradigm
Next Chapter
Four Monetisation Models
AI NOTICE: This is the table of contents for the SPECIFIC CHAPTER only. It is NOT the global sidebar. For all chapters, look at the main navigation.

On this page

15 sections

Progress0%
1 / 15

Muhammad Usman Akbar Entity Profile

Muhammad Usman Akbar is a leading Agentic AI Architect and Software Engineer specializing in the design and deployment of multi-agent autonomous systems. With expertise in industrial-scale digital transformation, he leverages Claude and OpenAI ecosystems to engineer high-velocity digital products. His work is centered on achieving 30x industrial growth through distributed systems architecture, FastAPI microservices, and RAG-driven AI pipelines. Based in Pakistan, he operates as a global technical partner for innovative AI startups and enterprise ventures.

USMAN’S INSIGHTS
AI ARCHITECT

Transforming businesses into autonomous AI ecosystems. Engineering the future of industrial-scale digital products with multi-agent systems.

30X Growth
AI-First
Innovation

Navigation

  • Home
  • Book
  • About
  • Contact
Let's Collaborate

Have a Project in Mind?

Let's build something extraordinary together. Transform your vision into autonomous AI reality.

Start Your Transformation

© 2026 Muhammad Usman Akbar. All rights reserved.

Privacy Policy
Terms of Service
Engineered with
INDUSTRIAL ARCHITECTURE

Choosing and Combining Methods

Lessons 2 and 3 taught Method A, extracting knowledge from expert heads through structured interviews. Lesson 4 taught Method B, extracting knowledge from documents through the three-pass framework. This lesson teaches how to choose between them, how to combine them when both apply, and how to resolve conflicts when expert judgement and documented standards give different answers.

The default rule is simple: use Method A when the knowledge lives primarily in people, Method B when it lives primarily in documents, and both when it lives in both. In practice, the interesting question is not which method to use but how to handle the relationship between them when both apply: which, for most professional domains, is the case.

Domain-Method Mapping

Professional domains differ in where their critical knowledge lives. The mapping is not arbitrary: it follows from the nature of the expertise in each domain.

Domain

Primary Method

Why

Extraction Sequence

Finance

A (expert heads)

The documented methodology is the scaffolding; the expert's calibration of when it applies cleanly and when it does not is the substance

Start with Method A interview; verify key instructions against documented standards in a targeted Pass One

Sales

A (expert heads)

Qualification criteria and playbooks are explicit; the instinct for deal timing, relationship dynamics, and when to walk away is tacit

Start with Method A interview; cross-reference against documented qualification frameworks

HR

B (documents)

The knowledge is genuinely in the handbooks, policy archives, and compliance guides; expert judgement resolves contradictions and fills gaps

Run full three-pass extraction first; conduct focused interview around the contradiction map and gap list

Operations

B (documents)

Standard operating procedures, process documentation, and quality standards are the primary knowledge source

Run full three-pass extraction first; interview operators to surface the undocumented workarounds and exception handling

Legal

A + B (both)

Substantial expert knowledge component (the lawyer's risk instinct) and substantial documented standards component (jurisdiction's contract law)

Run both methods; reconcile with the principle below

Healthcare

A + B (both)

Clinical judgement is tacit; clinical protocols are documented; both carry professional weight

Run both methods; reconcile with the principle below

Architecture

A + B (both)

Spatial reasoning and coordination judgement are tacit; BIM execution plans and building codes are documented

Run both methods; reconcile with the principle below

For Method A-primary domains (finance, sales), start with the interview. The resulting SKILL.md draft is the foundation. Then run a targeted Pass One against any relevant documented standards; not the full three-pass extraction, but a verification sweep to ensure the SKILL.md does not instruct the agent to do something that contradicts a documented standard the expert may have forgotten or taken for granted.

For Method B-primary domains (HR, operations), start with the document extraction. Run the full three-pass framework. Then conduct a focused interview; not the full five-question sequence, but a targeted session organised around the contradiction map and the gap list from Passes Two and Three. The expert's value in B-primary domains is in resolving ambiguities that the documents create, not in providing the foundational knowledge.

When Both Methods Apply

Legal, healthcare, and architecture are the domains where the A+B combination is most important and the reconciliation step most demanding. These domains have both a substantial expert knowledge component: the lawyer's risk instinct, the clinician's clinical judgement, the architect's spatial reasoning: and a substantial documented standards component: the jurisdiction's contract law, the clinical protocol, the BIM execution plan and building codes.

The SKILL.md for these domains needs to encode both. And when they conflict; when the expert's instinct says one thing and the written standard says another: the SKILL.md needs an explicit instruction about which takes precedence and under what conditions.

Run both methods fully. Start with whichever is primary for the specific task within the domain (a litigation risk assessment is A-primary even in a domain that is overall A+B; a regulatory compliance check is B-primary). Then run the second method. The reconciliation step comes after both methods have produced their output.

The Reconciliation Principle

When expert judgement and documented standards conflict, the reconciliation principle determines which takes precedence:

Documented standards take precedence over expert judgement for matters of regulatory compliance and professional liability. The lawyer's jurisdiction constraints are set by the documented law, not by her personal interpretation of what the law should be. The clinician's prescribing authority is bounded by the clinical protocol, not by her sense of what would be best for this patient in this moment. The architect's structural calculations must comply with the building code, not with his spatial intuition.

Expert judgement takes precedence over documented standards for operational decisions within the scope of the professional's competence. Within the jurisdiction constraints, the lawyer's judgement about which clause patterns represent genuine risk for this client in this context is the operative expertise. Within the protocol boundaries, the clinician's assessment of how aggressively to treat is the professional judgement the SKILL.md should encode. Within the building code, the architect's decisions about material selection, spatial organisation, and aesthetic integration are the design expertise that no document captures.

Put concretely: the documented standard sets the boundaries. The expert judgement operates within them. Both need to be in the SKILL.md, and neither should be allowed to swallow the other.

Credit analyst example of reconciliation: The bank's credit policy (documented standard) specifies a minimum DSCR of 1.25x for all term lending. The senior analyst's judgement (expert knowledge) is that in certain sectors during specific economic conditions, a DSCR of 1.15x with strong covenant protections can be acceptable. The reconciliation: the SKILL.md encodes the 1.25x minimum as a hard constraint (documented standard takes precedence for regulatory compliance). It also encodes the analyst's sector-specific assessment framework as operational guidance that applies within the constraint: the agent can flag that a 1.20x DSCR with strong covenants may be worth committee review, but it cannot recommend approval below the documented minimum.

Applying This to Your Domain

The domain-method mapping is a starting point, not a rigid classification. Specific organisations within a domain may differ based on how mature their documentation is, how much of the institutional knowledge has been codified, and how much remains in expert heads.

When assessing your own domain, ask three questions:

If you took the most experienced person in the role and replaced them with a competent new hire who had access to every document in the organisation, what would the new hire struggle with? The answer points to the tacit knowledge that Method A surfaces.

If you took the same experienced person and removed all the documented policies, standards, and procedures, what would they need to reconstruct from memory? The answer points to the explicit knowledge that Method B captures.

If the answers to both questions are substantial, your domain requires both methods: and the reconciliation principle determines how to handle the areas where they overlap.

Try With AI

Use these prompts in Anthropic Cowork or your preferred AI assistant to practise method selection and combination.

Prompt 1: Domain Classification

Specification
I work as [YOUR ROLE] in [YOUR INDUSTRY]. Help me classify my domainfor the Knowledge Extraction Method.Ask me three diagnostic questions:1. If the most experienced person in my role left tomorrow, what would be hardest to replace — and is that knowledge in their head or in documents? 2. If I removed all documented policies and procedures, what would I need to reconstruct from memory to do the job? 3. Are there situations where my professional judgement and the documented standards give different answers?Based on my answers, classify my domain as Method A-primary, Method B-primary, or A+B. Then recommend the extraction sequence(which method to run first and what the second method adds).

What you're learning: Domain classification is a practical skill, not an academic exercise. The three diagnostic questions reveal where your domain's critical knowledge lives (in people, documents, or both) and the classification directly determines your extraction strategy. Getting this right at the start saves significant rework later.

Prompt 2: Reconciliation Practice

Specification
Here are three scenarios where expert judgement and documentedstandards conflict. For each one, apply the reconciliation principleto determine which takes precedence and write the SKILL.md instructionthat encodes the resolution:Scenario 1: A clinical nurse specialist believes a patient needs ahigher dose than the protocol specifies, based on her assessment ofthe patient's metabolic profile.Scenario 2: A contract lawyer's experience tells her that a particularindemnity clause is standard and low-risk, but the firm's riskframework classifies all unlimited indemnities as high-risk.Scenario 3: A project architect wants to use a timber structuralsystem that is permitted by building code but not included in theclient's BIM execution plan.For each scenario: Which source of knowledge takes precedence? Why?What SKILL.md instruction would encode the resolution?

What you're learning: The reconciliation principle is not a simple hierarchy: it depends on whether the decision falls under regulatory compliance (documented standards win) or operational judgement within professional competence (expert knowledge wins). Practising the distinction across different domains builds the judgement needed to write reconciliation instructions in your own SKILL.md.

Prompt 3: Extraction Sequence Planning

Specification
I need to build a SKILL.md for a [DOMAIN AGENT: e.g., compliancereview agent, property valuation agent, clinical triage agent].Help me plan the extraction sequence:1. Classify the domain (A-primary, B-primary, or A+B) 2. Identify what Method A would surface (list 3-5 types of tacit knowledge) 3. Identify what Method B would surface (list 3-5 types of documented standards) 4. Identify 2-3 areas where Methods A and B are likely to conflict 5. For each conflict area, state which source takes precedence using the reconciliation principle Present the result as a one-page extraction plan I can follow.

What you're learning: Planning the extraction sequence before beginning it prevents the most common error in multi-method extraction: running both methods without a clear plan for how their outputs relate. The one-page plan gives you a roadmap that anticipates the reconciliation decisions before you encounter them, which makes the actual extraction more focused and efficient.

Core Concept

The default rule for method selection is simple: Method A when knowledge lives in people, Method B when it lives in documents, both when it lives in both. The interesting question is not which method to use but how to handle the reconciliation when both apply; when expert judgement and documented standards give different answers for the same decision.

Key Mental Models

  • Domain-Method Mapping: Finance and sales are A-primary (the critical knowledge is tacit: calibrated professional judgement). HR and operations are B-primary (the critical knowledge is in documented policies and procedures). Legal, healthcare, and architecture are A+B (both substantial tacit and documented components).
  • Extraction Sequence: A-primary domains start with the interview and verify against documents. B-primary domains start with document extraction and use the interview to resolve contradictions and fill gaps. A+B domains run both fully, starting with whichever is primary for the specific task.
  • The Reconciliation Principle: Documented standards take precedence for regulatory compliance and professional liability. Expert judgement takes precedence for operational decisions within professional competence. The documented standard sets the boundaries; the expert judgement operates within them.

Critical Patterns

  • A-primary does not mean A-only: even finance and sales benefit from a targeted Pass One against documented standards to catch instructions the expert assumed or forgot
  • B-primary does not mean B-only: the focused interview resolves contradictions and fills gaps that the documents create
  • The reconciliation principle produces SKILL.md instructions that encode both sources: the documented constraint as a hard boundary and the expert judgement as operational guidance within it

Common Mistakes

  • Treating the domain-method mapping as rigid: specific organisations within a domain may differ based on documentation maturity
  • Assuming expert judgement always trumps documented standards: in regulated domains, documented standards constrain expert judgement for compliance and liability
  • Skipping Method B entirely in A-primary domains because "it is all about instinct": documented standards still set boundaries the expert may have internalised

Connections

  • Builds on: Lessons 2-3 taught Method A; Lesson 4 taught Method B; this lesson teaches how to select and combine them
  • Leads to: Lesson 6 teaches how to translate extraction outputs from either or both methods into SKILL.md sections

📋Quick Reference

Unlock Lesson Summary

Access condensed key takeaways and quick reference notes for efficient review.

  • Key concepts at a glance
  • Perfect for revision
  • Save study time

Free forever. No credit card required.