Snapshot
The Challenge
- Proposal generation (320 hours/quarter). Each new engagement proposal required assembling capability descriptions, relevant case studies, team bios, pricing models, and methodology sections from past proposals. Consultants spent 4–8 hours per proposal hunting through SharePoint, copying sections from previous proposals, and reformatting. The firm produced 40+ proposals per quarter.
- Knowledge retrieval (280 hours/quarter). Consultants frequently needed: methodology frameworks from past engagements, industry benchmarks, client history, internal policies, and best practices from colleagues’ prior work. The knowledge lived in SharePoint folders, Outlook emails, Teams messages, and the heads of senior partners. Finding the right information required interrupting colleagues or spending hours browsing folders.
- Client deliverable production (240 hours/quarter). Client reports followed predictable structures but required significant formatting: data visualization, executive summaries, recommendation sections, and appendices. Consultants spent more time formatting than analyzing.
- Time entry (160 hours/quarter). The firm billed hourly. Consultants entered time weekly — often on Friday afternoons from memory, resulting in 15–20% of billable time going unrecorded. At $300/hour average, 15% leakage on $32M in revenue represented approximately $4.8M in potential unbilled work, though not all of this was recoverable.
- Internal operations (200 hours/quarter). Client intake processing, resource allocation scheduling, monthly reporting to the partnership, and compliance documentation — all manual, all consuming operations team bandwidth.
Our Approach
Phase 0: AI Readiness Assessment (2 weeks, $12,000)
We interviewed 15 people across all levels: 3 partners, 5 senior consultants, 4 junior consultants, and 3 operations staff. We shadowed consultants through a complete proposal cycle and a complete engagement delivery cycle.
Key finding: The firm’s knowledge was its most valuable asset and its most poorly organized. 8 years of engagement deliverables, proposals, and methodologies existed across 47,000 files in SharePoint — with no consistent naming convention, no tagging, and no search capability beyond filename matching. This single problem underpinned most of the other inefficiencies.
Recommended transformation sequence:
- Knowledge base (foundation for everything else)
- Proposal generation (highest ROI per hour invested)
- Time entry assistant (revenue recovery)
- Deliverable production (quality improvement)
- Operational automation (efficiency)
Phase 1: Knowledge Base + Proposal Assistant (Weeks 1–8)
AI Knowledge Base:
We ingested 12,000 of the most relevant documents (filtered from 47,000 by recency, engagement size, and partner input) into a RAG-powered knowledge base:
- Engagement deliverables and final reports (3,200 documents)
- Proposals and pitch decks (1,800 documents)
- Methodology frameworks and templates (400 documents)
- Industry research and benchmarks (2,100 documents)
- Internal policies and procedures (500 documents)
- Meeting notes and engagement retrospectives (4,000 documents)
Architecture: Documents parsed and chunked with structure-aware chunking (preserving section hierarchy). Embedded with OpenAI text-embedding-3-small. Stored in pgvector. Access control based on the consultant’s practice area and client authorization level. Deployed as a Slack bot and a web interface.
The experience: A consultant types in Slack: “What supply chain optimization methodology did we use for manufacturing clients in 2024?” and receives a sourced answer in 4 seconds, citing the specific engagement deliverables, with links to the full documents.
Proposal AI Assistant:
Built on top of the knowledge base, the proposal assistant generates draft proposals from a structured brief:
- Consultant fills out a brief (client name, industry, problem description, estimated scope, key team members)
- AI searches the knowledge base for relevant case studies, methodology sections, and team bios
- AI generates a draft proposal including: executive summary, understanding of the client’s challenge, proposed methodology (drawn from similar past engagements), team composition with relevant experience, timeline, and fee structure
- Consultant reviews, edits, and polishes
Result: Proposal creation time dropped from 4–8 hours to 45 minutes of review and customization. Proposal quality improved because the AI consistently included the most relevant case studies and the most current methodology — things that hurried consultants often missed.
Phase 2: Time Entry Assistant + Deliverable Tools (Weeks 9–14)
AI Time Entry Assistant: Built an AI that suggests time entries based on observable activity:- Calendar events (meetings, client calls, workshops) → suggested entries with client, duration, and activity description
- Email activity (exchanges with specific clients) → suggested entries
- Document access (SharePoint files associated with client engagements) → suggested entries
- Teams/Slack activity (client channels) → suggested entries
- Report drafting assistant: Given a data set (usually Excel) and a report template, the AI generates: data visualizations (charts, tables), executive summary highlighting key findings, analysis sections with data-supported observations, and recommendation frameworks. The consultant provides the strategic insights and recommendations; the AI handles the structure, formatting, and data presentation.
- Presentation builder: Given a report or analysis document, the AI generates a presentation draft with: key points extracted and organized by slide, data visualizations reformatted for presentation, speaker notes with talking points. The consultant refines the narrative and adds client-specific context.
Phase 3: Operational Automation (Weeks 15–24)
Client intake automation: When a new engagement is signed, an n8n workflow triggers: project setup in the firm’s PM tool, resource allocation based on availability and skill matching (AI-suggested, human-approved), engagement kickoff email to the client with onboarding materials, SharePoint folder structure creation with standard templates, and billing setup in the financial system.
Monthly partnership reporting: AI generates the monthly report for the partnership meeting: revenue by practice area with trend analysis, utilization rates by consultant with benchmarking, pipeline analysis with win probability, client satisfaction scores, and AI-generated narrative highlighting significant trends and anomalies.
Compliance documentation: Automated generation of engagement letters, conflict checks, and NDA documentation — pulling client data from the CRM and inserting it into standard templates with appropriate terms based on engagement type and client tier.
The Results
Measured over the first full quarter (Q4) after all three phases were deployed:
Annualized and accounting for continued optimization: $480,000/year in recovered value — a combination of recovered billable capacity, improved time capture, and operations efficiency.
Qualitative outcomes:
The knowledge base changed how consultants worked. Instead of starting each engagement by reinventing methodology, consultants began each engagement by asking the knowledge base “what have we done before for companies like this?” — accessing the firm’s collective intelligence in seconds.
Junior consultants reported the biggest productivity improvement. Previously, they spent significant time figuring out “how we do things here.” The knowledge base gave them instant access to the firm’s methodologies, templates, and best practices — accelerating their development from months to weeks.
Client Quote
— Managing Partner, [Client]
Investment Summary
Running A Professional Services Firm With Administrative Overhead?
We build AI systems that recover billable capacity for consulting, accounting, and legal firms. From knowledge management to proposal automation to time capture improvement.