# Tricky Wombat

> Enterprise AI search platform built on context-first architecture. We engineer the full context pipeline — from query intent classification to retrieval, assembly, generation, and evaluation — so every AI answer is accurate, sourced, and relevant.

## About

- [Homepage](https://www.trickywombat.ai): AI-powered search that finds better answers from your data
- [Small Team Search](https://www.trickywombat.ai/search-small-teams): AI search for teams under 50 people
- [Enterprise Search](https://www.trickywombat.ai/enterprise-search): AI search for organizations with 100-5000+ employees
- [Technical Discovery](https://www.trickywombat.ai/technical-discovery): AI search for engineering teams across repos, wikis, and docs
- [Pricing](https://www.trickywombat.ai/price/calculator): Transparent pricing calculator

## Signals (Articles)

- [What exactly IS prompt engineering?](https://www.trickywombat.ai/signals/what-is-prompt-engineering): Prompt engineering is where AI results start, not where they peak
- [How to roll out AI implementations](https://www.trickywombat.ai/signals/technology-roll-out): Are there techniques that high-performing organizations use to roll out their AI implementations? Lets explore finding a special recipe that all organizations can strive for.
- [The real benefits of adding an AI chatbot to your website.](https://www.trickywombat.ai/signals/benefits-of-ai-chatbot): Companies with strong knowledge bases achieve 75% AI resolution rates and 369% ROI. The 39% that fail almost always have an infrastructure problem, not a model problem.
- [Prompt engineering is only the first step to create great AI](https://www.trickywombat.ai/signals/prompt-engineering): The first five hours of prompt work produce a 35% accuracy gain. The next 60 add only 6%. The best organizations treat prompt engineering as only the first step.
- [Context engineering](https://www.trickywombat.ai/signals/context-engineering): A well engineered pipeline matters. The gap between relevant response and poor answers is not a model problem. A smaller AI with well-engineered context outperforms a much bigger model without it. Context engineering, the discipline of designing what an AI sees before it responds, explains the gap.
- [Your AI support bot isn't stupid](https://www.trickywombat.ai/signals/bot-isnt-stupid): Peer-reviewed research shows the same LLM swings from 50% to 87% accuracy based solely on retrieval pipeline design. Yet 95% of enterprise AI pilots deliver zero P&L impact. The model was never the bottleneck.
- [What is the ROI for AI customer service?](https://www.trickywombat.ai/signals/what-is-the-roi-for-ai-customer-service): Klarna's chatbot projected $40M in profit improvement. That didn't materialize, so they started rehiring human agents within a year.
- [The Real AI Chatbot Implementation Timeline](https://www.trickywombat.ai/signals/chat-implementation-timeline): Gartner found 85% of CS leaders will pilot AI chatbots in 2025 but only 5% have deployed one. The bottleneck isn't the platform, it's the knowledge infrastructure underneath it.
- [Good knowledge = good engineering](https://www.trickywombat.ai/signals/good-engineering-knowledge-pipeline): Most AI pilots fail not because the models are wrong but because engineering data is unclassified, ungoverned, and full of gaps. A problem that naive AI accelerates, not solves.
- [AI in legal practice is broken](https://www.trickywombat.ai/signals/ai-in-legal-practice-is-broken): 74% of lawyers cite AI accuracy as their top concern. Legal AI tools hallucinate 17-33% of the time, even on Westlaw and Lexis. Your data pipeline is the fix.
- [AI Slop is the New Spam](https://www.trickywombat.ai/signals/ai-slop-is-new-spam): AI slop is generic AI output that sounds confident but says nothing. 53% of consumers distrust AI search results. Learn why it happens and how to fix it.

## Technical Articles

- [AI Engineering Disciplines](https://www.trickywombat.ai/technical-articles/ai-engineering-disciplines): AI engineering is a seven-discipline stack that determines AI quality. Prompt engineering is only one. Here's what the other six are and why they matter more.
- [Query intent](https://www.trickywombat.ai/technical-articles/query-intent): Query intent classification cuts RAG hallucination from 40% to under 5% and reduces LLM API costs by up to 95%. Here's how the architecture works.
- [Prompt Rewriting](https://www.trickywombat.ai/technical-articles/prompt-rewriting): Over 60% of RAG errors start before the LLM runs. Query rewriting techniques improve retrieval quality by 15-45% across independent benchmarks.
- [Cognitive Resonance](https://www.trickywombat.ai/technical-articles/cognitive-resonance): Why matching words was never the same as matching meaning, and what the shift to concept-level retrieval changes for teams that depend on their own data
- [User Memory](https://www.trickywombat.ai/technical-articles/user-memory): RAND data shows 80% of AI projects fail from infrastructure gaps. LLM memory layers cut token costs by 90%, but extraction pipelines still hit only 30-70% accuracy.
- [Agglomerative clustering](https://www.trickywombat.ai/technical-articles/agglomerative-clustering): Agglomerative clustering groups related chunks between retrieval and generation, giving LLMs structured context instead of ranked fragments.

## Pages

- [Enterprise Search](https://www.trickywombat.ai/enterprise-search): Knowledge lives in every department. Engineering docs, sales playbooks, HR policies, product specs. Your people need answers that cross those boundaries. Connected Search finds them.
- [Small Team Search](https://www.trickywombat.ai/search-small-teams)
- [Technical Discovery](https://www.trickywombat.ai/technical-discovery)
- [Services](https://www.trickywombat.ai/services)
- [Pricing](https://www.trickywombat.ai/price)
- [What We Do](https://www.trickywombat.ai/what-we-do)
- [Content Creators](https://www.trickywombat.ai/creators)