# Your company’s AI platform — built into your perimeter. > The advantage isn’t access to AI. It’s what your people can build with it. Wide Moat deploys a secure AI platform inside your perimeter, connects it to your systems, and trains your team to build their own agents and apps. - 1,500 / 2,000 active users - 100 agents built by users - your data never leaves ## Real AI advantage comes from integration: your data, systems, governance, and teams trained to build with it. - **Same models. Different outcomes.** — Everyone can access frontier AI. The advantage is what your company builds around it. - **Integration turns AI into leverage.** — Connect your data, systems and workflows. Then train teams to build on them. - **Every build makes the moat wider.** — Each agent, skill and connector becomes reusable company infrastructure. ## One AI platform, running inside your walls. Portal, gateway, access controls, connectors, agents, app deployment and enablement — assembled into one secure system your people can build on. - **AI portal & workspaces** — One entry point for every employee — chat, documents, voice, web — plus one-click cloud dev environments for builders. - **Unified gateway** — Every request routed, budgeted and governed in one place — security, cost and routing under control. - **Context & knowledge** — Access-aware retrieval and memory on your knowledge bases — people see in AI only what they may already see. - **Connectors to your stack** — Code, tickets, CRM, BI, knowledge, automation — specific to your company, all behind the same gateway. - **Skills factory** — Reusable, governed skills built with the people who do the work. Build one — thousands reuse it. - **Agent & app deployment** — Build agents and internal AI apps, then deploy them inside the same governed ecosystem. - **Analytics & ROI** — Dashboards and a daily digest: who’s adopting, what’s working, where the real return is. - **Training & enablement** — A real curriculum, personal dashboards and a path that moves whole teams from curious to capable. - **Ambassadors & demo days** — The cultural engine — champions inside every team, recurring demo days, momentum that compounds. ## The platform, end to end. Everything flows through one gateway — money, security and routing under control. Actors on top, models and connectors below. ★ marks the pieces unique to Wide Moat. Connectors are illustrative — yours are specific to your company. Tap any block to see what it links to. - **Business users** (ops · legal · sales · finance) — The broad base of knowledge workers — the people whose hours AI moves most. Work through the portal, chat and the cloud workspace; build their own skills and small apps. - **Technical teams** (support → architects · dev · QA · SRE) — Every technical role, from L1 support to architects. The multiplier that builds the platform. Coding agents locally and in the cloud workspace; automate themselves; feed the skills factory. - **Agents** (autonomous · scheduled · in-process) — The third kind of actor — not a person. QA, document and client-facing agents. Use the same skills and connectors as people, with their own access, memory and budgets. - **AI Portal** (the entry point for everyone) — One door for every employee: chat, documents, voice and web — inside the perimeter. Upload a doc → extract. Search the web → draft a report. All skills and connectors, day one. - **Chat & messengers** (assistants where people already are) — Assistants that live inside the messenger your teams use all day. Just write in the chat — convert a statement to a sheet, pull numbers, collect documents. - **Coding agents** (in the IDE · most AI traffic) — IDE-resident agents — the main tool of the technical org; the bulk of AI traffic. Subscriptions + API through the gateway; live next to the code, review, document, debug. - **Cloud AI Workspace** (one-click dev environment) — A virtual IDE plus coding agents in the cloud. One click spins up a workstation. Skills and connectors ready instantly — no env to configure, no packages, no tokens to chase. - **Computer Use** (disposable AI sandboxes) — Isolated, disposable sandboxes where AI actually does work — hundreds of packages inside. Agent requests a sandbox → gets an environment → builds → it is destroyed. The workplace for AI. - **Analytics** (adoption · ROI · daily digest) — The coordination layer — dashboards and a daily digest for leaders and teams. Automatic overview of what worked and what didn’t; good cases get picked up and scaled. - **LLM Gateway** (routing · budgets · guardrails · caching) — The one gateway every system speaks through. All AI traffic passes here. Connectors, per-user/team budgets, rate limits, guardrails and prompt caching — all observable. - **Context engineering** (access-aware · RAG · memory) — The context layer — history, retrieval and memory, always with access rights applied. Checks who may see what, retrieves from the knowledge base, keeps a per-user memory. - **Skills Factory** (reusable skills · build once, reuse everywhere) — A skill = a reusable bundle of prompt, context, tools and data. Build one — thousands reuse it. Actors call skills through the interfaces; skills reach into connectors, knowledge and context. - **Private models** (your GPU · LLM · voice · vision · embeddings) — Your own model server: open-weight LLMs, speech-to-text, TTS, vision and embeddings. The most sensitive paths run here — client data, voice, documents — never leaving the walls. - **Frontier API** (best-in-class via subscription + API) — The leading frontier models for the hardest reasoning and coding work. Proxied through the gateway for people and agents alike, with budgets and caching. - **Second frontier** (diversification) — A second frontier provider for coding and reasoning. Subscriptions + API through the gateway. - **Regional fallback** (jurisdictional · vision) — A regional provider as fallback plus strong vision models for documents and plans. API through the gateway, used when its specific vision models are the right tool. - **Aggregator** (one contract, many models) — A legal aggregator giving access to many external models through a single contract. One billing relationship instead of ten vendor contracts; extra connectors included. - **Issue tracker** (tasks · incidents · docs) — Connector for people and agents — tasks, incidents and documentation. An agent can take a ticket and carry it to done; people get fast access to projects and knowledge. - **Code & Git** (repos · PRs · pipelines) — All the company’s code: repositories, pipelines, CI/CD — for people and agents. Agents read code, open merge requests and review; the skills factory mines commit history. - **CRM** (clients · deals) — Connector to pull client and deal data out of the CRM. Account managers ask AI about their clients — turnover, deals, history — and it assembles answers. - **BI** (dashboards · databases) — Business-intelligence dashboards and the databases behind them. Rights are inherited — everyone sees only their own; AI queries with the asker’s permissions. - **Knowledge DB** (vector stores) — Knowledge in vector form — documents, FAQs, and indexed source code. Separate stores for semantic code search, general knowledge and company-wide code search. - **Search & enrich** (web · company data) — Web search plus company-enrichment and reference-data services. AI checks a counterparty → enriches it → writes conclusions. - **Authorization** (SSO · directory · groups) — One identity system — SSO, directory and groups. On every model and connector call, the gateway checks who this is and what they may access. - **Observability** (alerts · logs) — Monitoring, alerting and the prod/dev logs. Support asks “what’s happening with app X?” — AI reads the alerts and logs and proposes causes. - **Automation** (workflows · webhooks) — A connector to the workflow-automation tool — run a workflow, update something, call a system. People build automations for themselves; agents call those same workflows as connectors. - **Doc parsing** (PDF → text layers) — A utility that turns PDFs into clean text layers. Cheap, simple extraction feeding document pipelines, the knowledge DB and document skills. - **Memory** (graph of people & work) — Long-term memory of people and agents plus knowledge about the company. Stores who did what, when and about what; linked to context engineering and analytics. ## Stop AI spend from running away. Teams burn through AI budgets when every tool, model and subscription is managed separately. Wide Moat puts token limits, model routing, subscription usage and quality analytics behind one governed gateway. - **Token and budget control** — Set limits by team, user, workflow or use case. Route simple tasks to cheaper models and reserve premium models for work that needs them. - **API vs. subscription routing** — Route work through API usage, enterprise subscriptions or internal/open-source models depending on cost, quality and policy. Reduce duplicate seats and keep spend under one gateway. - **Usage and quality analytics** — See who uses AI, for what, at what cost and with what quality. Track adoption, token spend, model performance and business impact together. **Lower cost. Less lock-in.** Wide Moat is built around interoperable components and an open-source core. Teams can use the models and tools that fit their work, while the company keeps control of cost, access, observability and vendor dependency. ## Three phases to a wide moat. First everyone saves their own hours, ad hoc. Then the company aims that effort at revenue and cost. Then it rebuilds how it operates — so deeply that rivals can’t catch up. | | Individual AI (Personal productivity) | Enterprise AI (Company productivity) | Wide Moat (An advantage rivals can’t reach) | |---|---|---|---| | Essence | AI starts as personal leverage — helpful, but fragmented | Work aligns around business goals — AI starts moving company metrics | A new business architecture — processes rebuilt around AI | | Business users | Do today’s work faster & better — − time per task, + quality | Create new things, ship improvements — + revenue, − cost | Do what rivals can’t — more value, better price | | Technical teams | Less toil — copilots, auto-tests, agents in the SDLC | Improve how you build — full-cycle AI development | Redesign how work runs — not code — operating architecture | | Agents | Assist people — on request | Agents run repeatable workflows — at scale, under control | Act on signals — notice, decide, execute, improve | | KPI | Hours & $ saved · active AI users — weekly cuts | Revenue · time-to-market · cost efficiency — P&L | Market position · margin · growth rate — company-level advantage | ## We don’t build the agents. Your people do. Wide Moat AI gives them the platform, guardrails and training to build safely at company scale. A real case study from a 2,000-employee financial company — 75% adoption across the organization. “The win wasn’t the model. It was a whole company that knew how to use it.” ## Built for companies with enough office work to need governed AI — and enough urgency to move. - < 500 office workers — early fit. AI tools can help, but platform infrastructure usually pays off later. - 500 – 5,000 office workers — recommended. The sweet spot: enough knowledge work, systems and security needs for governed AI to move real hours. - 5,000+ office workers — phased rollout. Still a fit — usually best started in one business unit, region or function before scaling. **What makes a fit** - Knowledge-work heavy organizations — where hours live in documents, decisions and workflows. - Sensitive data or regulated workflows that need access controls, auditability and clear boundaries. - Existing systems to connect — CRM, tickets, BI, knowledge bases, code and internal tools. - Leaders who want adoption, not another disconnected pilot. ## It lives in your perimeter. - **Your data never leaves** — Sensitive paths run on your own models and infrastructure — on-prem or your private cloud. - **One governed gateway** — Every request is routed, budgeted, rate-limited and logged. Nothing talks to a model on its own. - **Access-aware by design** — SSO and directory groups flow into every call — people see in AI only what they already can. - **Fully observable** — Every prompt, cost and action is measured — so adoption and risk are both visible. ## Start digging the moat. A 30-minute demo of the platform, the architecture and how we’d roll it out inside your perimeter. contact@widemoat.ai