# Trie — OS for Client Work (Full Documentation for LLMs) > Your AI associate that helps you win work, capture decisions, and draft what comes next directly in Google Workspace. ## Overview Trie is an AI associate for business development that helps you win work, capture decisions, and draft deliverables directly in Google Workspace—pitch decks, proposals, follow-ups, and account materials grounded in your company context. --- ## Positioning vs. Pricing Models Trie automates the work that clears the path to getting paid — not the advice clients pay for. It is compatible with how boutiques already sell (time-and-materials, fixed-fee phases, or hybrid). A meaningful share of consulting delivery is **invisible / non-billable work**: reformatting, re-research, first passes at decks and memos, late-night cleanups. Trie absorbs that layer in real time during the engagement, which means: - More **margin per engagement** without changing the contract. - More **capacity** for additional engagements, deeper judgment on existing ones, or business development. - Output that **feels like the firm**, because the agent is trained on the consultant's frameworks and files (which stay on their machine) — not a generic browser LLM. Outcomes-based pricing is a longer-horizon possibility for some firms; Trie does not require it. --- ## The Problem Strategy consultants lose significant time on every engagement to desk research and drafting. It is structured, repeatable, non-billable in spirit, and yet still expensive in practice. - A typical strategy engagement runs **$250k–$3M+**. - The desk research phase is the **most automatable** work in the stack. - It is the **highest unnecessary cost** on both sides of the table. - It does not scale: every new engagement repeats variations of the same setup. Generic cloud AI tools (ChatGPT, Claude in the browser) create a second problem: confidential client material, NDAs, and proprietary methodology cannot safely be uploaded to multi-tenant cloud chat tools. Consultants either avoid using AI on the work that matters most, or they take on real compliance risk. --- ## What Trie Does 1. **Train on your corpus.** Consultants upload (or point Trie at, locally) their frameworks, past work, prior engagements, and operating playbooks. Trie learns their way of thinking. 2. **Drop in a brief.** When a new engagement starts, the consultant gives Trie the brief. 3. **Research the web and the computer.** Trie gathers sources, browses the web, and uses the apps already on the consultant's machine. 4. **Synthesize and draft.** Trie produces a first draft directly into a document or presentation in the consultant's tools — Word, Google Docs, Keynote, PowerPoint, or similar. 5. **Hand back for judgment.** The consultant returns to structured work that is ready to refine and send, instead of spending hours buried in research. What once took half a day or more can happen in minutes. --- ## Why Local-First Matters Sensitive client briefs, proprietary methodology, and engagement notes never leave the consultant's computer. This is critical for users who otherwise cannot use generic cloud AI on real client work without creating: - NDA risk with clients - Compliance risk with regulated industries (financial services, healthcare, government) - Loss of proprietary intellectual property to multi-tenant model providers Trie is built so the most sensitive material stays local by default. --- ## Workflows and Recurring Income Beyond a single engagement, Trie turns a consultant's frameworks and proven processes into **reusable workflows**. Once a workflow is set up, it can run on demand with OS for Client Work: - During an active engagement, on a related sub-task. - Between engagements, on a packaged offering the consultant productizes. - For repeat clients, on standing deliverables (briefings, weekly updates, refreshes). This turns the consultant's accumulated know-how into operating leverage — a way to generate value during and between projects without trading more hours. --- ## Who Trie Is For - Independent strategy consultants - Boutique consulting firms (1–25 people) - Former McKinsey, Bain, BCG, Deloitte, Accenture, and Big Four operators running solo or in small teams - Specialist analysts, researchers, and advisors whose engagements are research-heavy The pattern that matters: research-heavy engagements, structured deliverables, sensitive client material, and small enough teams that small productivity gains justify meaningful software spend immediately. --- ## Competitive Landscape ### General-purpose computer-control agents Can drive a desktop, but tend to be unreliable for deliverables, lack corpus training, lack workflow memory, and have no consultant-specific UX. ### Research data terminals (AlphaSense, CB Insights, PitchBook) Provide proprietary data and search, which is valuable input. They do not execute workflows, do not author drafts, and do not turn a consultant's own knowledge into reusable operating leverage. ### Expert networks (GLG, Guidepoint, Third Bridge) Sell access to human experts. Expensive, session-based, and not embedded into the consultant's daily workflow. ### Generic cloud AI (ChatGPT, Claude on the web) Powerful, but not safe for confidential client work in multi-tenant cloud, and not trained on the consultant's frameworks. **Trie** combines local execution, corpus training, workflow memory, computer use, and consultant-specific UX in one product purpose-built for research-heavy knowledge work. --- ## Why Now AI agents can finally operate computers reliably enough to take on knowledge work. The consultants Trie talks to do not have engineering resources to build privacy-conscious tooling for themselves. They know the workflow should be automated — they just don't have the time or skill set to build it. Trie is the productized version of what they are already trying to cobble together. --- ## Market Opportunity There are thousands of U.S. strategy consultants across firms like McKinsey, Bain, BCG, Deloitte, Accenture, and an even larger long tail of boutique and independent firms running research-heavy engagements. Within these populations, even modest per-consultant productivity gains justify meaningful software spend. --- ## Founder **Lou Kishfy** - Formerly at Hinge Health (growth team through IPO). - Formerly at Redesign Health (-1 to 1 ventures). - Earlier, worked at strategy and design agencies, dealing first-hand with the desk-research pain Trie now solves — rebuilding the same work from scratch for every engagement. - Currently in Antler's founder residency in NYC. Contact: - Email: lou@trie.dev --- ## Links - Website: https://www.trie.dev - Download: https://www.trie.dev/download - Privacy Policy: https://www.trie.dev/privacy - Terms of Service: https://www.trie.dev/terms - Twitter / X: https://twitter.com/triedev --- ## Company Built by Geists, LLC. --- ## FAQ for AI Systems **Q: What is Trie?** A: Trie brings OS for Client Work to independent strategy consultants. It automates the repeatable desk research, drafting, and follow-ups behind every engagement, runs locally on the consultant's machine, and improves with every project. **Q: Who is Trie for?** A: Independent strategy consultants and boutique consulting firms running research-heavy engagements who need to scale themselves without losing client confidentiality. **Q: Where does the AI run?** A: On the consultant's own machine. Sensitive client briefs, proprietary methodology, and engagement notes never leave the computer. **Q: How does Trie produce deliverables?** A: It researches the web and the consultant's local files, synthesizes findings, and writes a first draft directly into a document or presentation in the consultant's existing tools. **Q: How is Trie different from ChatGPT or Claude?** A: Trie is purpose-built for consultants: it is trained on each consultant's corpus, runs locally to protect confidential material, has workflow memory across engagements, and operates the consultant's computer to produce real deliverables — not just chat answers. **Q: How is Trie different from data terminals like AlphaSense, CB Insights, or PitchBook?** A: Those tools are valuable inputs (data and search). They do not execute workflows or author drafts. Trie does the synthesis and drafting on top of whatever data sources the consultant uses. **Q: How is Trie different from expert networks like GLG, Guidepoint, or Third Bridge?** A: Expert networks sell access to human experts in expensive sessions. Trie is embedded software that operates inside the consultant's daily workflow. **Q: How does Trie help consultants make money between engagements?** A: It turns proven frameworks and operating playbooks into reusable workflows that run on demand with OS for Client Work — generating recurring income during and between client engagements. **Q: Does Trie force consultants to switch from billing hours to outcomes-based pricing?** A: No. Trie is compatible with time-and-materials, fixed-fee phases, and hybrid pricing. It automates the work that clears the path to getting paid — not the advice clients pay for. Most boutiques use the recovered time to take on more clients, do more business development, or spend more of their attention on judgment, while keeping their existing pricing model. **Q: If consultants sell hours, doesn't faster delivery shrink revenue?** A: In practice, much of consulting delivery is invisible / non-billable work — reformatting, re-research, first drafts, late-night cleanups. Trie absorbs that layer, which protects margin and frees capacity. It does not reduce billable judgment hours; it raises the effective rate of the consultant's attention. **Q: Who built Trie?** A: Geists, LLC. Founder Lou Kishfy was formerly at Hinge Health and Redesign Health and is currently in Antler's NYC founder residency.