Background
A 60-person management consulting firm with expertise in operational transformation was sitting on a gold mine of institutional knowledge — 8 years of engagement reports, frameworks, client deliverables, and industry analyses — but couldn't access it efficiently. Consultants spent an average of 6.2 hours per week searching for past work, recreating deliverables that already existed, and asking colleagues for information that was documented somewhere but impossible to find.
The firm had tried SharePoint, Confluence, and a custom-built wiki. Each became a “document graveyard” within 6 months — content was uploaded but never maintained, search was keyword-based and unreliable, and new consultants had no way to discover relevant past work without asking senior partners directly.
The Challenge
The knowledge management problem was costing the firm an estimated $890K annually in lost productivity and duplicated effort:
- Search inefficiency: Consultants averaging 6.2 hours/week searching for information across 4+ systems (SharePoint, email, Slack, local drives). At average billing rates of $275/hour, that's $890K in lost capacity across the firm
- Deliverable duplication: An estimated 35% of new deliverables substantially overlapped with existing work. Teams were recreating frameworks, analyses, and recommendations that had been built for previous clients
- Onboarding drag: New consultants took 4–6 months to become fully productive because institutional knowledge was locked in senior partners' heads and scattered across systems
- Knowledge attrition: When two senior partners left the previous year, their accumulated expertise — client relationships, industry insights, framework refinements — left with them
- Proposal quality: Business development teams couldn't quickly find relevant past engagements, resulting in generic proposals that didn't leverage the firm's actual track record
The Solution
AffixedAI built an AI-powered knowledge management system through a Growth engagement — 3 months at $12,500/month covering system architecture, knowledge ingestion, agent deployment, and team training.
The system has three core components:
1. Knowledge Ingestion Pipeline
Connected to all existing knowledge sources: SharePoint (4,200 documents), Confluence (1,800 pages), email archives (tagged engagement correspondence), Slack channels (industry-specific discussion threads), and a structured database of 340 past engagements. Every document was chunked, embedded, and indexed with automatic metadata extraction (client industry, engagement type, framework used, date range, team members involved).
2. Knowledge Assistant Agent
A natural-language interface that any consultant can query. Instead of keyword search, consultants ask questions like “What frameworks have we used for supply chain optimization in manufacturing?” or “Show me every engagement where we helped a client reduce operational costs by more than 20%.” The agent synthesizes answers from multiple sources and cites every claim to its original document.
3. Proactive Knowledge Agent
Monitors active engagements and proactively surfaces relevant past work. When a consultant starts a new project in the healthcare sector, the agent automatically provides: relevant past healthcare engagements, applicable frameworks, industry benchmarks from previous analyses, and contact information for colleagues with healthcare expertise. Reduces the “cold start” problem for every new engagement.
Implementation Timeline
- Week 1–2: Source audit, API integrations, document pipeline architecture
- Week 3–4: Knowledge ingestion (4,200 SharePoint docs + 1,800 Confluence pages). Quality validation on sample queries
- Week 5–6: Knowledge Assistant deployed to pilot group of 12 consultants. Feedback loop established
- Week 7–8: Proactive Knowledge Agent deployed. Integration with project management system for automatic engagement detection
- Week 9–12: Full firm rollout, training workshops, optimization. Added email and Slack as knowledge sources
Technology Stack
- AI Models: Claude Opus for complex synthesis queries, Claude Sonnet for routine lookups and proactive suggestions
- Knowledge Base: Vector database (Supabase + pgvector) with 180,000+ chunks from 6,000+ documents
- Search: Hybrid retrieval (semantic vector search + keyword BM25), re-ranked by relevance and recency
- Integrations: SharePoint API, Confluence API, Slack, Gmail, custom engagement database
- Channels: Slack bot (primary), web dashboard, email digest
- Infrastructure: $1,200/month (Supabase Pro, AI API costs, integration hosting)
Results
After 90 days of firm-wide deployment:
- Search time reduction: 6.2 hours/week per consultant → 1.4 hours/week (77% reduction). Consultants find relevant past work in under 2 minutes instead of 45+ minutes
- Deliverable reuse: 35% duplication rate → estimated 12%. Consultants now start from existing frameworks and adapt rather than recreating
- New consultant ramp time: 4–6 months → 6–8 weeks. New hires query the knowledge base instead of waiting for senior partners to be available
- Proposal win rate: 31% → 44%. Proposals now reference specific past engagements, relevant metrics, and applicable frameworks — all surfaced automatically
- Knowledge preservation: When a senior consultant left during the pilot, the team reported zero knowledge loss. All of their past analyses, frameworks, and client notes were fully accessible through the system
- Agent usage: 94% weekly active usage across the firm. Average of 8.3 queries per consultant per day. The Slack bot became the most-used integration in the firm's workspace
ROI Analysis
| Item | Amount |
|---|---|
| AffixedAI Growth engagement (3 months) | $37,500 |
| Ongoing advisory (9 months at $5K/mo) | $45,000 |
| Infrastructure costs (12 months) | $14,400 |
| Total Year 1 investment | $96,900 |
| Recovered billable capacity (search time) | $580,000 |
| Reduced deliverable duplication | $165,000 |
| Improved proposal win rate (attributed revenue) | $420,000 |
| Total Year 1 impact | $1,165,000 |
| Year 1 ROI | 1,102% |
What the Client Says
“We had 8 years of brilliant work locked in SharePoint folders that nobody could find. Now any consultant can ask a question in Slack and get a synthesized answer with source citations in 30 seconds. Our proposal win rate went from 31% to 44% because we actually leverage our track record instead of starting from scratch every time. This is the single highest-ROI technology investment we've made.”
— Managing Director, 60-Person Management Consulting Firm