Case Study 6.2: Elena's Consulting Report — The Three-Pass Method

Background

A mid-size technology company — a UK-based SaaS business serving the legal sector — had asked Elena's consulting practice to produce a market entry analysis for entering the US professional services software market. The deliverable was a 25-30 page strategy report for the executive team and board, to be used to make a go/no-go decision on the US market entry.

This was a significant deliverable: complex analysis, high stakes, professional polish expected at board level. Elena had done the primary research herself over three weeks: market sizing data from verified sources, competitive landscape research from SEC filings and industry reports, interviews with five US-based legal technology buyers, and a financial model.

She had her research. She had her conclusions. She needed to produce a polished, boardroom-quality report from this material — typically a two-week process of writing, editing, and revision.

She used her three-pass iteration method to produce the first draft in four days.

The Three-Pass Framework

Elena had developed the three-pass method after several early experiences with AI-assisted consulting documents that failed in specific ways:

First failure: She had used AI to generate content from scratch in her first pass and found that the AI-generated content, while well-written, had weak analytical foundations — it was more speculative than evidenced. She had to substantially rewrite it after doing the research.
Second failure: She had used AI to write specific sections without first validating the overall document structure, and found that sections were redundant, sequencing was illogical, and the document needed major restructuring after content had already been written.
Third failure: She had done structure, then content, but had not applied a systematic quality pass at the end — and a client noticed inconsistencies between the executive summary and the body recommendations that she should have caught.

The three-pass method was designed specifically to address all three failures.

Pass 1: Structure and Argument (Day 1 — 3 Hours)

Elena's first pass was entirely about the architecture of the document. No content was written during this pass.

Session 1.1: Deliverable Scoping

Her first prompt established the project context in full:

I am producing a market entry analysis for a UK-based legal sector SaaS company considering entry into the US professional services software market. The deliverable is a 25-30 page strategy report for an executive team and board.

The report needs to answer three questions: (1) Is the US market large enough and accessible enough to justify entry? (2) What is the right entry mode and go-to-market approach given the company's current capabilities? (3) What are the financial scenarios and what does success look like?

Please propose an analytical framework for this report — not a document outline yet, but the analytical approach: what claims need to be established, in what logical order, and what evidence would make each claim credible?

AI response (summarized):

The AI proposed a claim-evidence framework in three logical groups:

Group 1 — Market Opportunity (must establish before anything else): - Claim: The US legal tech market is large enough to be worth entering - Claim: The segment this company targets (mid-size law firms, 10-100 lawyers) has unsolved problems that this company can address - Claim: Regulatory and compliance barriers to market entry are manageable

Group 2 — Competitive Position (establishes whether entry is viable): - Claim: The current competitive set does not occupy the specific value proposition this company offers - Claim: Existing US players have specific weaknesses this company can exploit - Claim: The company's UK track record provides defensible differentiation

Group 3 — Entry Strategy (establishes how to enter): - Claim: A specific entry mode is optimal given capability and budget constraints - Claim: A specific initial target segment provides the best risk-adjusted path to scale - Claim: Financial scenarios show a viable path to profitability within [timeframe]

Elena's evaluation:

"The framework is right. The three-group structure follows the logical progression I need: establish the market is attractive, establish we can win in it, then specify how. However, I'm missing one critical section: the organizational readiness assessment. The board will need to know not just whether the market is attractive but whether this company is ready to enter it. That needs to be addressed explicitly."

Session 1.2: Document Structure

Elena brought the framework (with her addition) to a document structure conversation:

Based on this analytical framework, please propose a detailed document outline for the 25-30 page report. For each section, include: the section title, a purpose statement (what this section needs to establish), a rough indication of length (pages), and the type of evidence that would make this section credible.

AI response (summarized):

A 9-section outline:

Executive Summary (3 pages) — Synthesizes all key findings and the recommendation
US Legal Tech Market: Size and Dynamics (4 pages) — Market sizing, growth, segment breakdown
Target Segment Analysis (4 pages) — Mid-size law firm profile, buying behavior, unmet needs
Regulatory and Compliance Environment (2 pages) — What a UK SaaS company needs to navigate for US entry
Competitive Landscape (4 pages) — Direct competitors, adjacent players, white space analysis
Strategic Fit Assessment (3 pages) — This company's differentiation and capability against US market requirements
Organizational Readiness (2 pages) — What changes are needed to enter; internal capability assessment
Recommended Entry Strategy (4 pages) — Entry mode, initial segment, GTM approach, timeline
Financial Scenarios (3 pages) — Conservative/base/optimistic scenarios with key assumptions

Elena's evaluation:

"This is essentially correct. Two adjustments: (1) The Regulatory and Compliance section is too short at 2 pages — US legal software has specific compliance requirements (data residency, legal hold, e-discovery compatibility) that are material to market entry feasibility. Expand to 3-4 pages. (2) The Strategic Fit Assessment and the Competitive Landscape should swap positions — I need to establish what the competitive landscape looks like before I can assess our strategic fit against it. The logic is: here's the market, here's the competition, here's how we fit."

Outcome of Pass 1:

A validated 9-section outline with adjusted positions and purpose statements, plus a confirmed analytical framework. Total Pass 1 time: approximately 3 hours (including Elena's own review and adjustment).

What Elena did not do in Pass 1: She did not write a single sentence of content. She resisted the temptation to start drafting when she felt the structure was clear — she wanted full structure validation before content investment.

Pass 2: Content Drafting (Days 2-3 — Two Sessions)

Pass 2 began with Elena's research. She had 47 pages of organized research notes: market data from IBIS World and Gartner, competitive profiles from SEC filings and Crunchbase, interview notes from her five buyer interviews (anonymized), and her own financial model in Excel.

Her approach: section by section, providing verified research as input and asking Claude to draft each section from that research.

Section 2 example — US Legal Tech Market: Size and Dynamics

Elena's prompt for Section 2:

I am drafting Section 2: US Legal Tech Market: Size and Dynamics, for the market entry analysis. Purpose of this section: establish that the US legal tech market is large enough and growing fast enough to justify market entry.

Here is my verified research for this section: [Pasted: market sizing data from IBIS World report, growth figures from two independent analyst sources, segment breakdown from Gartner Legal Technology Market Guide, notes from buyer interview on market dynamics]

Please draft this section (approximately 4 pages / 900-1,100 words) in a professional strategy consulting report style. Use only the data I have provided. Do not supplement with your own knowledge or training data — if any claim requires data I haven't provided, flag it with [NEEDS DATA] rather than generating a substitute figure.

Include: an opening paragraph establishing the market headline, data-driven analysis of market size and growth, segment breakdown with commentary on the mid-size law firm segment specifically, and a closing paragraph that bridges to why this creates an opportunity for our client.

AI response quality (Elena's assessment):

"The section draft was well-structured and used my research accurately. There were two [NEEDS DATA] flags — one for competitive market share estimates (which I hadn't provided) and one for customer acquisition cost benchmarks in the segment (I hadn't collected this). Both were legitimate flags, not AI limitations — I had simply not yet gathered that data."

Elena went back to her research sources for the flagged items: she found a competitive market share estimate in a trade publication she had missed, and found CAC benchmarks in an industry survey from a legal tech association. She brought both back in the next prompt:

Here are the additional data points you flagged: [data]. Please update the draft to incorporate these.

Progression through sections:

Elena worked through all nine sections over two sessions (approximately 6 hours total across both days). Her experience per section:

Sections where research was strong: Draft was close to final quality with minor adjustments. Sections 2, 3, and 5 (market data and competitive landscape) fell into this category.
Sections requiring more human judgment: Sections 6 (Strategic Fit), 7 (Organizational Readiness), and 8 (Recommended Entry Strategy) required more iteration because they involved recommendations, not just analysis. Elena often ran 2-3 rounds per section for these.
The Executive Summary: Elena left this for last. After completing all other sections, she asked Claude to draft the executive summary based on the key findings and recommendation from each section.

Critical rule maintained throughout Pass 2: All specific data, claims, and figures came from Elena's verified research. Claude organized, structured, and articulated; Elena sourced and verified.

Sections requiring mid-pass research trips:

Three times during Pass 2, the [NEEDS DATA] flag revealed a genuine gap in her research. She stopped, gathered the missing data from primary sources, and then continued. This happened with: US employment law complexity for remote SaaS workers, specific state-by-state bar association software certification requirements, and comparable transaction multiples for recent legal tech acquisitions.

Outcome of Pass 2:

A complete first draft of all 9 sections, approximately 27 pages, based entirely on Elena's verified research. Total Pass 2 time: approximately 6 hours across 2 days.

Pass 3: Quality and Polish (Day 4 — 3 Hours)

Elena did a full read-through of the complete draft herself before beginning any AI assistance in Pass 3. She read it as she would read a document she was seeing for the first time.

Her human read produced a list of issues:

Consistency issue: The executive summary recommendation specified "partnership/distribution as primary entry mode" while the Recommended Entry Strategy section recommended "direct sales with selective channel partnerships." These were inconsistent.
Logic gap: Section 5 (Competitive Landscape) established that two specific US competitors had significant weaknesses in mid-size firm UX. Section 6 (Strategic Fit) did not reference these weaknesses as a differentiation opportunity — a logical connection that was made in her research notes but had not been reflected in the draft.
Tone inconsistency in Section 7: The Organizational Readiness section was notably more tentative in tone than the rest of the document — lots of "might need to consider" and "could potentially" where the rest of the document was confident and direct. This section needed a tone adjustment.
Section 8 was too long: The Recommended Entry Strategy had grown to approximately 6 pages in the draft, well beyond the targeted 4. Several paragraphs were redundant.

Session 3.1: Targeted rewrites

Elena addressed issues 2 and 3 with targeted AI rewrites:

Section 6, Strategic Fit Assessment, needs a paragraph added to the Differentiation Analysis subsection. Based on the competitive weaknesses identified in Section 5 — specifically that [Competitor A]'s interface consistently receives below-average UX scores from mid-size firms in independent reviews, and that [Competitor B] lacks native mobile capability — please add a paragraph explaining how our client's UK track record in modern UX and mobile-native design represents a specific, evidence-based differentiation opportunity in this market.

Section 7, Organizational Readiness, is written in a tentative tone that doesn't match the rest of the document. Please rewrite it in the same confident, direct tone as Sections 5 and 6 — same style, same directness. The content stays the same; only the tone changes.

Session 3.2: Consistency check

Elena used AI to check cross-document consistency after making the targeted rewrites:

I've made several targeted rewrites to the draft. Please read the following three elements and check them for consistency:

[Pasted: Executive Summary recommendation paragraph, Section 8 entry mode recommendation, Section 9 financial scenario assumptions]

Do these three sections make the same recommendation, use consistent terminology, and cite consistent assumptions? If they diverge in any way, note the specific divergence.

AI response:

The AI confirmed the entry mode inconsistency (Elena had known about this), confirmed that Section 9's financial scenario base case used a different sales cycle length assumption than Section 8's GTM timeline, and identified a third minor issue: Section 1 (Executive Summary) referenced "two to three strategic partnerships" while Section 8 mentioned "one anchor partnership in year one, with additional partnerships in year two." These were consistent in intent but inconsistent in language.

Elena corrected all three issues directly in the document herself — these were judgment calls she wanted to make as the expert, not hand back to AI.

Section 8 compression:

Section 8 of the draft is approximately 6 pages and needs to be reduced to 4 pages without losing the substantive content. Please identify which paragraphs are least essential to the core recommendation and suggest a condensed version of the section.

The AI identified three paragraphs that were restatements of points made earlier in the section and two that were better placed in an appendix. Elena used this diagnosis to cut the section directly.

Final human read — the 3-pass rule applied:

Before sending the document to the client, Elena applied the 3-pass rule personally:

Pass 1 (Substance): Read for accuracy of all specific claims. She had a checklist of every data point and the source. She spot-checked 15 specific figures. All checked out — the verified-research-only discipline in Pass 2 had held.

Pass 2 (Fit): Read for appropriateness to the specific client context. She adjusted two sections where her language was too general and did not reflect specific things she knew about the client's competitive positioning and risk tolerance from her conversations with them.

Pass 3 (Voice): Read for tone and language consistency. She made approximately 25 small word-level edits — "leverage" replaced with "use," one paragraph shortened because it was repetitive, three instances where passive voice crept back in.

Outcome of Pass 3:

A polished, board-quality 27-page strategy report, ready for client delivery.

Time and Quality Comparison

This project: 12 hours total (3h Pass 1 + 6h Pass 2 + 3h Pass 3) over 4 days.

Previous comparable project (without AI assistance): A similar market entry analysis the previous year had taken Elena 3 weeks to produce at approximately 50-60 hours total, including 2 weeks of writing, editing, and revision after the research was complete.

Quality assessment: Elena's client described the report as "the most clearly argued market entry analysis we've received" — noting specifically the logical progression from market attractiveness to competitive position to entry strategy as being unusually clear and navigable for board members who were not deep in the subject matter.

What the three-pass method specifically enabled:

Pass 1 prevented structural rework. On her previous (non-three-pass) projects, she had averaged 15-20 hours of post-draft structural revision. The Pass 1 investment of 3 hours eliminated most of this.
Pass 2 research discipline prevented content fabrication. By providing all data herself and instructing Claude to flag gaps rather than fill them, every claim in the document was traceable to a verified source. The [NEEDS DATA] flags were the mechanism that enforced this.
Pass 3 systematic consistency checking caught issues human editing typically misses. The entry mode inconsistency and the financial assumption divergence were the kind of internal consistency issues that are easy to miss in a human-only edit of a 27-page document. The structured consistency check prompt caught both.

Key Lessons

The method addresses specific, identified failure modes. Elena's three-pass method was not designed in the abstract — it was designed in response to specific failures she had experienced in earlier AI-assisted projects. Structured iteration methods are most valuable when they are designed to prevent known failure patterns.

Research must precede AI drafting, not the other way around. The Pass 2 discipline — provide verified research, ask AI to draft from it — is the inversion of how most people approach AI-assisted research and writing. Most people ask AI to do the research. Elena provided the research and asked AI to organize and articulate it. This is the correct approach for high-stakes professional deliverables.

Consistency checking is a specific, often-missed quality step. Human editors of long documents tend to catch local errors (within a section) but miss global inconsistencies (across sections). A structured AI consistency check, run on the specific elements most likely to diverge, is a powerful quality assurance step that takes 10 minutes and catches errors that might otherwise reach the client.

The human final read is non-negotiable. Elena made approximately 40 small edits in her final 3-pass read. None of them were individually large, but collectively they made the document authentically hers — in specific word choices, in calibrated hedges that reflected her professional judgment, and in the client-specific adaptations that only she could make because only she knew the client.

AI accelerates execution; human judgment drives quality. Elena's professional value in this project was: identifying the organizational readiness gap the AI missed, evaluating which creative territory to develop (in structure), recognizing inconsistencies (in the human read), making the recommendation (throughout), and knowing this client's risk tolerance and framing preferences (in the fit pass). AI drafted; Elena decided.