--- "title": "How do I build AI capability that stays inside the company?" "description": "How to build AI skills and know-how that stay inside your firm after any training or consultant engagement ends, using internal champions, shared assets, clear ownership and a steady practice rhythm." "section": "AI Training" "sidebarTitle": "Build in-house AI capability" "published": "2026-10-26T13:30:12.019Z" "updated": "2026-10-26T13:30:12.019Z" --- ## How do I build AI capability that stays inside the company? It depends on where you are starting. If almost nobody uses AI well yet, begin with hands-on training on real work so a few people get genuinely good, then make them your internal champions. If you already have a handful of strong users, the job is to capture what they know in shared prompts, standards and short guides, and give someone clear ownership. If you plan to rely on custom tools, make sure your own staff can run and adjust them. In every case, capability stays when it is written down, owned and practiced. ## Who this is for Founders, managing partners, COOs and heads of people who want AI skills to outlast a single workshop, a vendor contract or one enthusiastic employee. ## The short version - Capability lives in people, shared assets and habits, not in licenses. - Pick a few internal champions and give them time, not just a title. - Write things down: prompt libraries, standards, examples of good work. - Any outside help should leave your team more independent, not less. ## Why AI know-how leaks out of firms Most firms already have some AI skill. It just sits in the wrong places. One analyst has a great set of prompts saved in a personal chat history. A consultant built a useful workflow and then left. A partner figured out how to summarize long documents and never showed anyone. When those people move on or get busy, the know-how goes with them. Buying more seats does not fix this. As we cover in [why licenses are not a strategy](/docs/licenses-are-not-a-strategy), access is the easy part. The harder part is turning individual tricks into the firm's way of working. ## The four pieces of lasting capability ### 1. People who can teach others Choose two to five internal champions across different teams. Good champions are not always the most technical people. They are respected, curious and patient, and they do the real work of the firm every day. Give them protected time each week to help colleagues, test new uses and collect what works. Without time, the role becomes a badge and nothing else. ### 2. Shared assets Capability that only exists in someone's head is fragile. Build a small set of shared assets: - A prompt library organized by task, with examples of good inputs and outputs. - Prompting standards so everyone structures requests the same way. Our playbook on [creating prompting standards](/docs/create-prompting-standards) walks through this. - Short written guides for the five to ten tasks where AI helps most in your firm. - Your [AI acceptable use policy](/docs/write-an-ai-acceptable-use-policy), so people know what data can go where. Keep these somewhere people already look, not in a folder nobody opens. ### 3. Clear ownership Someone needs to own AI adoption as part of their actual job. In a smaller firm that might be the COO or an operations lead. In a larger one it might be a dedicated role. The owner keeps the assets current, tracks usage, decides which tools the firm uses and makes sure new hires get onboarded. If you are weighing whether this should be a new hire, see [hiring an AI lead vs an outside consultant](/docs/hire-ai-lead-vs-consultant). ### 4. A practice rhythm Skills fade without use. Build a light, regular rhythm: - A short monthly session where people share one thing that worked and one that failed. - AI onboarding for every new hire in their first weeks. - A quarterly review of the prompt library and policy, because tools and features change often. ## A simple 90-day plan 1. **Weeks 1 to 3: train on real work.** Run hands-on sessions using the documents and tasks your people handle every week. See [how to train non-technical employees to use AI](/docs/train-non-technical-teams-on-ai). 2. **Weeks 2 to 4: name champions and an owner.** Agree on time commitments in writing. 3. **Weeks 4 to 8: capture what works.** Champions collect the prompts and workflows that are actually saving time and turn them into shared assets. 4. **Weeks 6 to 10: standardize.** Publish prompting standards and short task guides. Fold them into onboarding. 5. **Weeks 10 to 13: measure and adjust.** Check whether people use AI on the agreed tasks and whether quality holds. Our guide on [measuring AI training results](/docs/measure-ai-training-results) explains how. ## If you build custom tools Custom assistants and agents can be a big part of your capability, but only if someone inside the firm understands them. Before anything is built, agree on who will own it, where the documentation lives and what your team can change without outside help. Our guide on [who maintains AI systems after the consultant leaves](/docs/maintaining-ai-systems) covers the questions to settle up front. ## Common mistakes - **Relying on one power user.** If they leave, so does the capability. Spread it across several people. - **Treating training as a one-time event.** A single workshop rarely changes habits on its own. See [AI workshop vs ongoing coaching](/docs/ai-workshop-vs-ongoing-coaching). - **Letting assets go stale.** A prompt library written for last year's tools quietly stops being used. - **Outsourcing the thinking.** If an outside firm does everything, you rent capability instead of building it. ## When you do not need us If you already have a respected internal lead with time to run this, and a few people who use AI well, you can follow the plan above on your own. Outside help is most useful when you are starting from a low base, want to move faster, or need custom tools your team cannot build yet. ## How Ruby helps - **AI training:** hands-on sessions on your firm's real work, plus coaching for the champions who will carry it forward. - **AI strategy:** setting ownership, standards, policies and simple measures so the capability sticks. - **AI engineering:** building assistants and agents with documentation and handover so your team can run them. Our aim is to leave your firm more capable on its own. If that is what you want, [book a call with Ruby](https://rubyadvisory.com/book-a-call). ## Related pages - [Why AI adoption stalls](/docs/why-ai-adoption-stalls) - [What should an AI training program for a small firm include?](/docs/ai-training-program-for-a-small-firm) - [How to create prompting standards for your company](/docs/create-prompting-standards) - [Who maintains AI systems after the consultant leaves?](/docs/maintaining-ai-systems)