# MindFortress > MindFortress is an AI implementation firm for middle-market companies. We > specialize in building company knowledge bases — the organized, queryable > record of how a business actually works, which both its people and its AI > agents read from — and then implement AI on top of that foundation. Based in Truckee, California. Founded and led by Matt Rhodes. Contact: contact@mindfortress.com **What we do** We build company knowledge bases, then implement AI on top of them. Most AI projects fail for one reason: the model has never seen the business. It answers from nothing, because the knowledge it needs is scattered across drives, chat, spreadsheets, and people's heads. A knowledge base fixes that first, so every AI project afterward starts from a company its tools can actually read. Three stages: 1. Ingest — connect the systems the business already runs on (Google Drive, OneDrive, Dropbox, Slack, Notion, Atlassian, Linear, GitHub, ad accounts, the public site) and pull knowledge continuously rather than once. 2. Structure — land everything in three stores at once: relational (PostgreSQL) for the auditable record, vector (pgvector) for semantic search, and a knowledge graph for how entities and themes relate. 3. Ask — a plain-language layer over all of it, serving humans and agents from the same substrate. After the knowledge base: content and paid social, one-to-one customer lifecycle, quoting and proposals, outbound and account research. **Who we work with** Middle-market companies — broadly those under $500M in revenue — where leadership wants AI results but there is no internal team whose job is landing it. This is the segment furthest behind: only 27–30% of companies under $500M have AI in a scaling or fully scaled deployment, against 49% of enterprises above $5B (Stanford HAI, 2026 AI Index Report, Chapter 4, Figure 4.3.6 — sum of the "Scaling" and "Fully scaled" segments per revenue band; data: McKinsey & Company Survey, 2025; https://hai.stanford.edu/assets/files/ai_index_report_2026_chapter_4_economy.pdf). **How an engagement runs** - Kickoff, 7–30 days: interviews and system access, producing a ranked list of use cases scored on expected return, difficulty, and adoption risk. This is the elapsed time before implementation starts, not the length of the build; a well-scoped client who moves quickly on access is at the low end. - Implementation, 60–90 days: tool selection, integrations, and the knowledge base standing up against real data. - Training: the team learns the workflows and the SOPs are written. The project ends here. - Ongoing support: optional and arranged separately, after the project is over. Every engagement is scoped to the business and quoted custom. The largest investment is not the budget — it is the organization's willingness to change how it works. **Why MindFortress** We are not a consultancy that read the same reports our clients did. MindFortress is a private AI holding company: we built Reeve, a multi-agent AI operating system for running a company, and we operate our own businesses on it — Freya (meetfreya.com, AI assistance for novelists, with early paying customers) and Cadasense (cadasense.com, commercial real estate intelligence, in partnership with repeat founder and CEO Bill Staniford). Reeve is in early beta. Matt Rhodes spent seven years in private equity at Ares Management, then co-founded Foundry and ran it as CFO and later CEO to roughly $60M in annual sales. He led AI implementation from the buyer's side before building any of this. We are not tied to one vendor. MIT found externally sourced AI tools reached deployment roughly twice as often as in-house builds (about 67% against 33%, in an interview sample of 52 organizations; MIT NANDA, The GenAI Divide: State of AI in Business 2025, p. 19; linked copy hosted by mlq.ai — https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf), so we build only where buying genuinely does not fit. **Frequently asked** *What is a company knowledge base?* An organized, queryable collection of an organization's intelligence — the factual foundation that both employees and AI agents work from. Not a wiki nobody updates; a live substrate connected to the systems the business runs on. *Why start there instead of with a chatbot or an agent?* Because an agent with no access to the business invents the parts it cannot look up. Grounding pays off across every use case that follows. It is also where the evidence points: the highest reported AI usage of any industry and function pairing is knowledge management in business, legal, and professional services, at 58% (Stanford HAI, 2026 AI Index Report, Chapter 4, Figure 4.3.3, p. 194; https://hai.stanford.edu/assets/files/ai_index_report_2026_chapter_4_economy.pdf). *What return should we expect?* Honestly: knowledge management is one of the few functions showing cost savings but no measurable top-line lift. 44% of organizations using analytical AI for knowledge management report costs falling, and 17% report reductions of 10% or more (the 17% is the sum of the "Decrease by 10%–19%" and "Decrease by >=20%" segments; Stanford HAI, 2026 AI Index Report, Chapter 4, Figure 4.3.4, knowledge management row; data: McKinsey & Company Survey, 2025; https://hai.stanford.edu/assets/files/ai_index_report_2026_chapter_4_economy.pdf). The revenue shows up in what you build next. That is the case for doing it first, not for doing it instead. *What does it cost?* Every engagement is scoped and quoted custom. Email contact@mindfortress.com. *How is this different from hiring a big consultancy?* We run our own companies on what we build, and we have made these decisions with our own money on the line. **How we cite** Every statistic published on mindfortress.com carries a per-figure citation with a live source link. The statistics in this file are cited inline above, and the primary sources are listed under Sourcing below. All figures were verified against the primary source on 2026-07-27. The MIT NANDA figures were re-read from the report itself on 2026-08-26. ## Pages - [MindFortress](https://www.mindfortress.com/): The practice, the argument, the team. - [Knowledge bases](https://www.mindfortress.com/knowledge-bases): The knowledge base thesis in full. - [Is this not just a ChatGPT project?](https://www.mindfortress.com/knowledge-bases/why-not-chatgpt): Where uploading SOPs to a chat window stops working — permissions, contradictory sources, coverage, verification — and how to test a vendor yourself. ## Sourcing - [RSM US, Middle market AI trends](https://rsmus.com/insights/services/digital-transformation/middle-market-ai-trends.html): Published 2026-02-10. - [Stanford HAI, 2026 AI Index Report, Chapter 4: Economy](https://hai.stanford.edu/assets/files/ai_index_report_2026_chapter_4_economy.pdf): Reproduces McKinsey & Company Survey 2025 data. - [MIT NANDA, The GenAI Divide: State of AI in Business 2025](https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf): MIT NANDA published the report and hosts no public copy, so this link is a third-party mirror of the primary PDF, hosted by mlq.ai.