Meet Sarah
Sarah is a strategy consultant at a mid-sized consultancy with a focus on organisational strategy and commercial growth across the professional services, retail, and industrial sectors. She has been in the profession for seven years, holds a senior consultant title, and works within a team of eight, reporting to a partner who manages overall client relationships while Sarah leads day-to-day analytical and deliverable work on most engagements.
At any given point in her working month, Sarah is managing between three and five active client engagements simultaneously. These vary in scale and intensity: one engagement might be in its early research and diagnostic phase, requiring heavy reading and synthesis; another might be in its recommendation development phase, requiring intensive analytical and writing work; a third might be in its implementation support phase, involving regular client contact, progress tracking, and communication. The rhythm of consulting work means that Sarah rarely has a working week where all engagements are in the same phase simultaneously, which creates a demanding context-switching challenge: she must be able to move between client situations that are genuinely different from one another, requiring different knowledge, different communication styles, and different analytical frames, often within the same working day.
Her primary deliverables are slide decks for client presentations, written strategic reports and recommendation documents, and the range of supporting analytical and research material that underlies these outputs. She also manages significant volumes of client communication, internal team coordination, and the administrative work of proposal development and engagement management that runs alongside client delivery.
The specific pain points that characterise Sarah's working experience before building an AI practice are three. First, the cognitive cost of maintaining accurate context across multiple simultaneous client engagements. Returning to a client after several days focused elsewhere requires re-familiarisation with the client's situation, the current state of the engagement, the decisions already made, and the specific sensitivities and preferences that should inform her work. This re-familiarisation is entirely manual and entirely dependent on the quality of the notes and documents she has produced during previous sessions. Second, the time cost of research synthesis. Consulting analysis requires substantial secondary research across industry sources, competitor data, academic and practitioner literature, and market intelligence, and synthesising this material into a coherent, client-specific analytical foundation is one of the most time-consuming elements of the early phase of any engagement. Third, the practical frustration of version management across presentation files. Consulting deliverables evolve through many iterations across a client engagement and are frequently adapted from previous work on similar engagements. Locating the right version of a file, identifying which slide from a previous deck is the most relevant starting point for a new one, and maintaining a clean version history across multiple deliverable iterations are sources of significant friction.
The Before State
Before establishing an AI practice, Sarah's file management is characterised by inconsistency that has accumulated gradually over years of busy engagement work. Client folders exist at the top level of her Google Drive, but within those folders the organisation is variable: some projects have clear subfolder structures, others have everything at a single level, and others have folders whose names reflect abbreviations and shorthand that made sense at the time of creation but require memory rather than inference to interpret. File names vary between projects and between periods within the same project, reflecting the different naming habits of different moments in her career and the influence of file names inherited from colleagues whose conventions differ from her own.
The practical consequence of this accumulated inconsistency is that beginning a new working session on a client requires locating relevant files through a combination of search and memory. Sarah estimates that she spends approximately twenty minutes at the start of each client session re-familiarising herself with the current state of the engagement: finding the most recent version of the primary deliverable, locating the research notes from previous sessions, recalling where the client last expressed preferences or constraints, and reconstructing the current status of outstanding questions. Across three to five active clients, each of which she returns to several times per week, this re-familiarisation time accumulates to several hours of unproductive navigation per week.
Her approach to research synthesis before AI assistance is entirely manual. Sarah reads source material, takes notes in a separate document, and constructs her synthesis progressively through reading and re-reading, a process that is thorough but slow. A comprehensive industry background brief for a new engagement typically requires two to three full working days of reading and synthesis work before it reaches a quality that she is comfortable presenting to a partner or using as the foundation for client-facing analysis.
Presentation file management is a particular source of frustration. Previous client decks contain slides that would be directly reusable or adaptable for new engagements, but finding the right slide requires opening multiple files and browsing their contents, because the file names and folder structures do not provide enough information to identify which file contains the relevant material. Sarah maintains a rough mental catalogue of where useful slides are located, but this catalogue is imperfect and degrades over time as new files accumulate.
Knowledge Base Setup
Sarah's knowledge base is built on the client-primary folder structure recommended in Module 4.1, reflecting the fact that long-term client relationships are the most stable organisational unit in her professional context. The structure is:
Clients / ClientName / ProjectName / Deliverables, Research, Admin, Archive
At the top level, each folder bears the client name in full, without abbreviation. Beneath the client folder, each discrete engagement has its own project folder, named with the project type and year: for example, GrowthStrategy_2024 or OperatingModelReview_2025. Beneath each project folder, four subfolders address the fundamental categories of material in consulting work.
Deliverables holds all client-facing outputs: slide decks, written reports, executive summaries, and supporting exhibits. Only finalised or formally submitted versions are held in this folder. Versions in active development are held in a separate Working folder within Deliverables, which prevents the accumulation of draft files alongside finished outputs and makes it immediately clear to any user of the folder, including AI tools and human colleagues, which documents represent the current state of the engagement's outputs.
Research holds all secondary research material: industry reports, competitor analyses, market data downloads, and the research notes and synthesis documents produced from this material. Subfolders within Research are organised by research theme rather than by source, reflecting the way research is used in consulting work: a user looking for material on market sizing looks for it under Market, not under McKinsey or Statista.
Admin holds engagement management documents: the signed engagement letter, project plan and timeline, budget and fee tracking, and internal meeting notes. This material is operationally important but should not influence the content of deliverables, and holding it in a separate folder prevents AI tools searching for analytical content from surfacing administrative documents alongside research and deliverable material.
Archive holds completed deliverables and superseded research from earlier phases of the engagement. When a phase closes, the relevant documents are moved to Archive rather than deleted, preserving them for reference while keeping the active folders focused on current work.
Sarah's file naming convention follows the format: ClientName_ProjectName_DocType_YYYY-MM-DD. The client name is abbreviated to a three or four letter code established at the start of each client relationship and applied consistently throughout. The project name is abbreviated similarly. The document type field uses a controlled vocabulary: Deck for slide presentations, Report for written documents, Brief for research briefs, Notes for meeting or session notes, Analysis for quantitative analytical documents, and Outline for structural planning documents. The date is the date of the most recent substantive revision.
Within each project folder, Sarah maintains three context documents, each stored at the top level of the project folder so that AI tools searching within the project have immediate access to them.
The Client Background Document is written at the start of each new client relationship and updated at significant intervals, typically quarterly for ongoing relationships and when significant changes occur. It covers the client's industry and competitive position, the organisation's structure and key decision-makers, the history of the relationship with the consultancy, the client's communication preferences and working style, and any sensitivities that should inform how Sarah approaches her work with them. This document is typically two to three pages and represents the institutional memory of the client relationship. When Sarah returns to a client after working on other engagements, reading the client background document replaces the twenty minutes of re-familiarisation that previously characterised her return to each client.
The Project Scope and Objectives Document is written at the start of each engagement and updated when the scope changes. It records the agreed objectives of the engagement, the specific deliverables that have been committed to, the timeline and key milestones, the constraints that govern the approach, the definition of success that the client and the firm have agreed upon, and any boundaries around the scope: things that are explicitly outside the current engagement and should not appear in recommendations. This document ensures that AI-assisted analysis and drafting remains within the agreed scope rather than producing recommendations that exceed the mandate or address issues the engagement was not designed to resolve.
The Key Findings and Decisions Log is updated at least weekly throughout the engagement. It records the significant analytical findings from each week of work, the decisions that have been made about the direction and framing of the analysis, and the specific conclusions that have been reached and approved. This document serves the function of the decision log described in Module 4.1 and addresses a specific consulting challenge: the risk that AI tools will surface alternative analytical approaches or reopen questions that have already been resolved through careful analysis and client discussion. When the AI tool has access to a record of what has been concluded and why, it does not repeatedly suggest approaches that the analytical process has already considered and set aside.
Model and Tool Selection
Sarah's primary AI tool is Claude, selected on the basis of the task-type matching framework from Module 4.2. Her most demanding AI-assisted tasks involve the processing of long research documents, the production of nuanced analytical writing that must be calibrated to a specific client context, and the careful instruction-following required to produce deliverable content that stays within agreed scope boundaries. Claude's demonstrated strength on each of these task types, combined with its large context window that allows the simultaneous processing of multiple long source documents alongside context documents, makes it the most appropriate primary model for the majority of her AI-assisted work.
For file search and document retrieval across her Google Drive, Sarah uses Google Drive's native AI search features. This integration provides content-based search within her file system without requiring her to leave the Google Workspace environment or to manage a separate connected integration. The limitation of this integration is that its AI capability is constrained to search and retrieval rather than analysis and drafting: it excels at finding relevant documents but the analytical and drafting work is then performed in Claude, where the depth of reasoning and instruction-following is more consistent.
For client communication, Sarah uses Claude for first-draft generation of structured correspondence, working manually with the copy and paste approach described in Section 1 of Module 4.3. She made this choice deliberately rather than configuring an email integration: the sensitivity and relationship-criticality of her client communications means she wants to control exactly what is submitted to the AI tool in each instance, and the re-familiarisation time previously spent at the start of each client session has been largely eliminated by the client background document, making the manual workflow efficient enough that the setup and governance overhead of an email integration is not justified.
Presentation creation is performed entirely without AI assistance for the slide construction and formatting work. Sarah made this decision after testing AI-assisted slide generation and finding that the formatting, visual consistency, and professional presentation standards required for client-facing consulting deliverables were not reliably met by AI-generated slides. The time saved in initial generation was consistently exceeded by the time required to correct formatting inconsistencies, adjust visual hierarchy, and ensure the deck met the quality standards the firm's clients expect. The AI assistance that applies to presentation work is in the research and analysis that underlies the slide content, not in the slide construction itself.
Workflow One: Starting a New Client Engagement
The engagement start workflow is the most important workflow in Sarah's AI practice because it establishes the knowledge base foundation that determines the quality of all subsequent AI assistance throughout the engagement. A thirty-minute investment in context documents at the start of an engagement produces returns that compound over the weeks of active work that follow.
Step one: Build the folder structure. Sarah creates the full folder structure for the new engagement before any documents are produced. This means creating the project folder within the correct client folder, creating the four subfolders, and creating the Working subfolder within Deliverables. The structure is built empty rather than created progressively as documents accumulate, because an empty but correctly structured folder is easier to populate consistently than a partially structured one that gets amended under the time pressure of active work.
Step two: Write the Client Background Document. Where a client relationship is genuinely new, Sarah writes the client background document from scratch, drawing on the information available from the engagement intake process, the initial client meetings, and any publicly available information about the client's business. For a returning client, she reviews and updates the existing document to reflect any changes since the previous engagement. She invests approximately thirty minutes in this document and treats this time as engagement setup rather than as a discretionary activity that can be deferred. The document is written in plain prose rather than bullet points, because prose contextualises the information in ways that AI tools find easier to interpret than disconnected bullet items.
Step three: Write the Project Scope and Objectives Document. Using the signed engagement letter, the proposal, and the notes from the initial scoping meetings, Sarah writes the scope document covering the agreed objectives, deliverables, timeline, constraints, and scope boundaries. This document is typically shorter than the client background document: one to two pages that record the essential parameters of the engagement with precision. Any ambiguity in the agreed scope that surfaces during this writing is noted explicitly as an open question to be clarified with the client or partner, because ambiguity in the scope document will propagate into ambiguity in AI-assisted analysis throughout the engagement.
Step four: Conduct industry and competitive research with AI assistance. Using the scope and client background documents as context, Sarah prompts Claude to produce a structured overview of the key trends, competitive dynamics, and relevant developments in the client's industry. She is explicit in her prompt about the client's specific situation: the industry segment, the geographic scope, the strategic question the engagement is addressing, and the level of detail and sophistication appropriate for the audience. She reviews the AI output critically, identifies gaps and areas requiring additional depth, and uses targeted additional research to address those gaps. The AI output is a research scaffold, not a finished research product.
Step five: Synthesise research into a background brief. Sarah uses Claude to help synthesise the research material into a structured analytical brief that will serve as the primary research reference for the engagement. She submits the research material, the client background document, and the scope document together, and asks for a synthesised brief that addresses specific analytical questions relevant to the engagement's objectives. She reviews the brief against the source materials, verifying factual claims and ensuring that the synthesis accurately represents the sources it draws on. She edits the AI draft substantially, incorporating the analytical judgment and client-specific framing that the AI output lacks, and saves the finished brief in the Research folder as the engagement's primary research reference.
Step six: Create the initial deliverable outline. Using the scope document and the research brief, Sarah works with Claude to produce an initial structural outline for the primary deliverable. She provides explicit guidance about the narrative logic of the deliverable: the key message, the sequence in which supporting points should be developed, and the way the recommendation should be introduced and justified. The AI output is a first draft of the outline structure, which Sarah reviews against the client context and the scope document before using it as the working framework for deliverable development.
Workflow Two: The Weekly Client Update Email
Client update emails are a high-frequency, high-value communication task that follows a consistent structure across engagements. The structure is stable enough to benefit from AI drafting assistance, and the relationship-specific personalisation that distinguishes a useful update from a generic one is easily applied to an AI draft in the final editing step.
Step one: Establish the week's content. Sarah reviews the project plan and her notes from the week's work, identifying the key accomplishments, the current status of active workstreams, any blockers or risks that have emerged, and the planned activities for the following week. This content review takes five to ten minutes and produces the raw material for the email. The review is performed by Sarah rather than delegated to an AI tool because it requires the exercise of judgment about what is significant and what the client most needs to know, which depends on knowledge of the client relationship and the dynamics of the engagement that only Sarah has.
Step two: Draft with AI assistance. Sarah opens the client background document and the current project scope document to have them ready as context material, then constructs a prompt that specifies the email's purpose, the relationship context, the content to be covered, and the tone required. A representative prompt structure is: "Draft a client update email from me to [contact name]. Context about the client and our relationship: [paste relevant section of client background document]. This week we achieved: [bullets]. Current status: [status note]. Next week we will: [bullets]. The tone should be professional and forward-looking, appropriate for a senior finance director who prefers concise communications." The AI draft is reviewed and then edited for relationship nuances: specific phrasing that reflects the history with this client, a reference to a conversation from the most recent meeting, or an adjustment of emphasis that reflects Sarah's read of what this particular client will find most reassuring or most useful.
Step three: Review and send. The final email is read in full before sending with specific attention to accuracy of content, appropriateness of tone, and the absence of any generic phrasing that would make the email feel like a template rather than a personal communication. Any section that reads as generic or that makes a factual claim Sarah has not verified against the actual week's progress is revised before the email is sent.
Workflow Three: Developing a Deliverable Section
The development of substantive deliverable content is where AI assistance delivers the highest value in Sarah's practice, and also where the verification requirements are most demanding. AI-assisted deliverable development follows a consistent process that maintains the human analytical judgment at the centre of the work while using AI assistance to accelerate the structural and drafting components.
Step one: Assemble the context. Before beginning any section of a deliverable, Sarah assembles the relevant context documents: the client background document, the project scope document, the key findings log, and the specific research material relevant to the section being drafted. For a section on market trends affecting the client's sector, this means the relevant section of the research brief, any specific industry data sources, and the analytical questions the section is meant to answer as defined in the deliverable outline.
Step two: Generate an initial draft. Sarah constructs a prompt that provides the assembled context material, specifies the section's analytical objective, identifies the key claims the section should make, and sets explicit constraints on scope and framing. A representative prompt structure for a market trends section might be: "Based on the research material and client context provided, draft a section on market trends affecting [client's sector]. The section should address: [specific trends identified in research]. The analysis should be framed around their implications for [client's strategic situation as described in scope document]. The tone should be analytical and direct. Length approximately 400 words. Do not include recommendations, which will appear in a separate section." The specificity of the prompt is deliberate: vague prompts produce generic outputs that require more extensive editing than specific prompts that constrain the AI's output to the actual analytical territory required.
Step three: Verify every claim against source material. Every factual claim in the AI draft, every market statistic, every characterisation of competitive dynamics, every assertion about industry trends, is verified against the specific source material from which it should have been derived. Where a claim cannot be traced to a specific source, it is either removed or independently verified. This verification is not a cursory read of the draft but a systematic check that treats every substantive claim as requiring confirmation. The verification step is budgeted as a separate activity with its own time allocation, not as a quick review appended to the drafting step.
Step four: Revise for analytical quality and client fit. The verified draft is then revised for the two qualities that AI assistance cannot provide reliably: analytical quality and client fit. Analytical quality means that the claims in the section support the overall argument of the deliverable coherently, that the logical connections between points are sound, and that the section does not overreach beyond what the evidence supports. Client fit means that the framing, the emphasis, and the specific language used reflect the client's particular situation and the way the partner and client have discussed the issues rather than a generic analytical treatment of the subject. This revision step is performed by Sarah and is where her professional judgment and knowledge of the client makes the most distinctive contribution to the quality of the finished section.
Step five: Integrate into the deliverable. The revised section is placed in the appropriate position in the deliverable structure, reviewed for consistency with adjacent sections in terms of framing, terminology, and logical flow, and flagged for the partner's review at the next internal session. Sections that depend on data or analysis from other sections are cross-checked for consistency before the full draft is circulated.
Quality Control Checklist
Sarah applies a consistent quality control checklist to AI outputs before they are incorporated into deliverables, sent as client communications, or shared internally. The checklist is not a bureaucratic formality. It is a professional verification standard that reflects the specific risk categories present in consulting deliverable work.
Can every factual claim be traced to a verifiable source? Consulting deliverables are relied upon by clients making significant strategic and commercial decisions. Any factual claim that cannot be traced to a specific, verifiable source has no place in a consulting deliverable, regardless of how plausible it sounds. This check is applied to every data point, every market statistic, every claim about competitive positioning, and every assertion about industry trends.
Is the analysis consistent with the client's specific constraints and context? AI tools will produce plausible analyses that are generically correct but contextually wrong: recommendations that are sound in principle but unavailable to this client because of their specific regulatory environment, capital position, operational capacity, or strategic history. Every AI-produced analytical output is checked against the scope document and client background document to confirm that the analysis is applicable to this client's specific situation.
Would this client recognise their own business in this analysis? A higher-order quality test that goes beyond factual accuracy to assess whether the analysis reflects the actual character of the client's situation. Generic consulting analysis can be factually accurate and logically coherent while completely failing to capture the specific texture of a client's competitive situation, the particular challenges they face, or the specific strategic choices available to them. If the client could not recognise their own business in the analysis, it requires more client-specific framing regardless of its generic quality.
Are all data points current and from appropriate sources? Research materials used in AI-assisted analysis have dates of publication. Market data, competitive intelligence, and industry statistics can become outdated quickly, and an AI tool asked to synthesise research will use the material provided without assessing whether it remains current. Every data point in a deliverable is checked for currency before the deliverable is finalised.
Does the tone and register match this client's expectations? Client communication in consulting spans a wide range of formality, technical depth, and communication style. A client who prefers direct, data-driven communication requires different framing from one who values detailed qualitative analysis. A board presentation requires different language from an operational briefing to a senior management team. AI drafts are checked for tone alignment with the specific client and specific audience before any communication is sent.
The After State
After twelve weeks of consistent AI practice, the most significant change in Sarah's working experience is the elimination of the re-familiarisation cost that previously consumed the first twenty minutes of every client session. Client background documents and key findings logs bring Sarah to current context in two to three minutes of reading, which represents a recovery of approximately four to five hours per week across her three to five active clients. This recovery is not theoretical: it is the direct result of the time previously spent on re-familiarisation being replaced by a brief, focused document review.
Research synthesis is the second area of significant change. The combination of AI-assisted initial research and structured synthesis, applied to the research brief workflow, has reduced the time required to produce a comprehensive engagement background brief from two to three full working days to approximately one day of work that includes careful verification. The AI does not replace the analytical judgment required to identify what is significant in the research and how it bears on the client's specific situation, but it substantially accelerates the extraction and initial structuring of the relevant material from a large volume of source text.
Client communication drafting has been improved in consistency rather than primarily in speed. The AI-drafted first pass, calibrated with the client background document, produces communications that require less extensive revision than communications drafted entirely from scratch, and the prompting practice of specifying tone and framing explicitly has made Sarah more deliberate about these qualities than she was when drafting without AI assistance.
Presentation work remains outside the AI practice entirely for slide construction. Sarah has tested AI presentation generation tools on multiple occasions since establishing her AI practice and continues to find that the formatting and visual quality requirements of consulting client presentations are not reliably met by current AI generation tools. The research, analysis, and written content that populates the slides is substantially AI-assisted through the deliverable development workflow, but the slide construction itself is performed manually.
Common Mistakes for Management Consultants
Allowing AI to write strategy without client-specific constraints. The single most consequential failure mode in AI-assisted consulting work is the production of analytical outputs that are generically plausible but not specifically applicable to the client. Strategy consulting adds value precisely because it addresses the specific situation of a specific client rather than applying generic frameworks to generic problems. An AI tool operating without comprehensive client context produces outputs that describe what a typical firm in the client's industry might do, which is categorically different from the analysis of what this firm, with its specific history, specific capabilities, and specific constraints, should do. The context documents described in this walkthrough are specifically designed to prevent this failure mode, but their value depends entirely on them being current, comprehensive, and actively used in every AI interaction.
Allowing context documents to become outdated. The value of context documents is proportional to their accuracy, and their accuracy degrades as the engagement progresses and changes accumulate without being recorded. A key findings log that is two weeks out of date does not prevent the AI tool from reopening questions already resolved. A client background document that does not reflect a recent change in the client's leadership structure will produce communications that reference the wrong people. The weekly knowledge maintenance habit from Module 4.1 is not optional in consulting AI practice: it is the activity that preserves the quality of AI assistance as the engagement develops.
Over-relying on AI for competitive analysis without independent verification. Competitive analysis is a category of consulting work where AI hallucination risk is particularly acute. AI tools trained on general text data may produce market share figures, competitive positioning claims, or characterisations of competitor strategies that are plausible-sounding but inaccurate. In a consulting deliverable that will be relied upon by a client making strategic decisions, an inaccurate competitive analysis claim is not a minor error. It is a reputational and professional risk. Every competitive claim produced through AI assistance must be verified against primary sources, and market data from AI outputs should never be cited without independent confirmation.
Treating AI deliverable drafts as polished outputs. The efficiency of AI-assisted deliverable development can create a temptation to apply lighter review and revision to AI drafts than the quality standards of client-facing consulting work require. AI drafts in consulting work are first drafts: they require the same analytical review, logical assessment, and client-specific revision that any other first draft would receive. The verification and revision steps in the workflow described above are not optional refinements to an already-complete product. They are the steps that transform a plausible AI draft into a consulting deliverable of the quality the firm's clients expect.