AI Proposal Copilot

Product design and front-end prototype for an AI-powered proposal authoring platform at McKinsey. The tool reimagines how consulting teams build Letters of Proposal (LOPs) — transforming a fragmented, manual process into a guided, AI-assisted workflow. From RFP parsing and project setup through structured storyline generation, intelligent slide matching from the firm's repository, deck assembly with real-time editing, and governed project close with sanitization for reuse. Built as a high-fidelity React prototype to validate the end-to-end experience before engineering investment.

Building a consulting proposal is one of the highest-stakes activities at a firm — yet the process is almost entirely manual. Teams start from scratch or copy-paste from old decks, spending days assembling slides that may already exist somewhere in the firm's repository. There is no systematic way to find, reuse, or remix proven proposal content.

The proposal lifecycle is fragmented across disconnected tools: SharePoint for document storage, PowerPoint for deck building, email for collaboration, and personal knowledge for slide curation. No single surface connects the intake (RFP), the strategy, the narrative structure, the slide library, and the final deliverable.

Institutional knowledge walks out the door with every proposal. Winning decks are rarely sanitized and stored for reuse. Teams in different offices unknowingly rebuild the same slides for similar engagements. The firm's collective proposal intelligence is locked in individual hard drives and email threads.

Mapped the end-to-end proposal lifecycle — from the moment a team receives an RFP through strategy selection, storyline development, slide sourcing, deck assembly, stakeholder review, and project close. Identified six distinct phases, each with unique UX needs and AI intervention opportunities.

Designed an AI copilot interaction model where the assistant is present at every phase but never takes over. The system observes context (RFP content, taxonomy, strategy choice) and proactively surfaces relevant suggestions — storyline structures based on engagement type, repository slides ranked by relevance, and activation resources (advisors, references, competitive intelligence) at the right moment.

Prototyped a progressive disclosure flow: single-column conversational setup (project creation, collaboration, taxonomy) transitions to split-pane workspace (chat + structured content) as complexity increases. Each phase builds on the previous — project setup feeds storyline, storyline drives slide matching, matched slides assemble into a reorderable deck.

Validated the taxonomy and storyline structure against actual PDP (Product Development & Procurement) engagement frameworks — ensuring the AI suggestions map to real proposal patterns consultants recognize and trust.

Conversational project setup: describe the engagement in natural language, attach an RFP, and the system creates a SharePoint project, parses the scope, and suggests a proposal strategy. Collaborators are invited with scoped access. Taxonomy tagging (industry, capability, geography) enables intelligent matching downstream.

AI-generated storyline: structured narrative outline based on the engagement type and strategy, with expandable sections mapping to the firm's proposal frameworks. Consultants review, refine via inline chat, and approve — turning an unstructured brainstorm into a governed document in minutes instead of days.

Repository-powered slide finder: AI searches the firm's proposal repository and surfaces the most relevant slides for each storyline section, with relevance scoring and source attribution. Multi-select, swap, and refine through conversation. The system learns from selections to improve future matching.

Deck assembly and governed close: selected slides are stitched into a reorderable PPT preview with a filmstrip rail for navigation. After assembly, activation resources surface (senior advisors, named references, competitive intelligence, pricing guides). Project close includes an approval gate for deck sanitization and repository submission — closing the loop so future teams benefit from this proposal.

High-fidelity prototype validated the end-to-end workflow across six distinct phases — demonstrating feasibility of AI-assisted proposal authoring before engineering investment.

Established the interaction model for AI copilot integration in consulting workflows: conversational where appropriate, structured where precision matters, and always under human control.

Designed the closed-loop knowledge system: every completed proposal can be sanitized and fed back into the repository, compounding the firm's collective proposal intelligence over time.