--- "title": "Why giving your team ChatGPT licenses isn't an AI strategy" "description": "Licenses give people access to AI. A strategy decides which work changes, who learns what, and what gets built. Here is the difference and a simple way to start." "section": "Concepts" "published": "2026-10-15T00:00:00.000Z" "updated": "2026-10-16T18:46:16.203Z" --- Buying ChatGPT, Claude or Copilot licenses gives your team access to AI, while an AI strategy is the set of decisions about which work should change, what people need to learn to change it, and what has to be built when a chat window isn't enough. Most firms make the first move and assume the rest will follow. It usually doesn't. ## Who this is for Managing partners, COOs and CEOs who have already rolled out AI seats, or are about to, and are now asking a quieter question: why doesn't anything feel different yet? ## The short version - Access is not adoption. A license is an invitation, not a change in how work gets done. - Adoption follows workflows, not tools. People use AI when it is attached to a task they already own and care about. - Some work needs systems, not prompts. Repeated, multi-step work across your tools is an engineering question, not a training one. ## What a license actually buys you A seat gives each person a capable general assistant. That is genuinely useful. Some people will quickly find their own uses: drafting, summarizing, preparing for a meeting. What a seat does not include: - **A shared picture of where AI helps in your firm.** Each person experiments alone, and good discoveries tend to stay with whoever made them. - **Skills.** Giving context, checking output and breaking a task into steps are learned habits. Without them, a couple of disappointing results is often enough to end the experiment. - **Permission and guardrails.** If nobody has said what data can go where, your most careful people may avoid the tool entirely, which is the opposite of what you want. - **Connections to your systems.** A chat window can't see your CRM, data room, shared drive or project files unless someone deliberately sets that up. ## What usage tends to look like without a strategy Without those pieces, a rollout often settles into a familiar shape. A handful of enthusiasts get real value and quietly become informal help desks. A larger group tries the tool a few times, gets mixed results on hard tasks, and drifts back to old habits. Leadership sees a recurring bill and a utilization number, but not a clear story about what changed in the work. ## The three questions a strategy answers A practical AI strategy for a firm doesn't need to be a long document. It needs clear answers to three questions. **1. Where is time and attention quietly slipping away?** Start with the work, not the tool. Look for tasks that are frequent, text-heavy and done by your most expensive people: first drafts, research synthesis, recurring reports, status updates, reconciling information from several places. Rank them by how often they happen and how painful they are. **2. What does each role need to learn?** Training should match the work people actually do. A partner preparing for a board meeting, an analyst building a first-pass model and an operations lead chasing documents need different habits. Generic prompting sessions raise awareness, but they rarely change Monday morning. **3. What should be built rather than prompted?** When the same multi-step process runs every week, pulls from several systems and has a predictable output, asking people to re-prompt it each time is wasteful and inconsistent. That is where agents, connectors and small custom tools earn their keep. ## A quick self-check If you can honestly say yes to most of these, you are already past a license rollout: - We can name the three to five workflows where we expect AI to matter most. - Each role has been trained on its own real tasks, not generic examples. - There is a short written page on which tools are approved and what data can be used. - Good prompts, checklists and examples are saved somewhere shared. - Someone owns AI adoption as part of their job, with time set aside for it. - We review what changed every month or two, in terms of work rather than logins. If most answers are no, that is normal. It simply tells you where to start. ## A simple starting sequence 1. Pick two or three workflows that matter and that people already complain about. 2. Train the people who own those workflows, using their real tasks. 3. Write down what worked as shared prompts, checklists and examples. 4. Decide what deserves a system, and build only that. 5. Review in 30 to 60 days: what changed, what didn't, and what comes next. ## Common mistakes - **Measuring logins instead of outcomes.** Seat utilization says little. Ask which tasks are now faster or better. - **One big kickoff, then silence.** Habits form with repetition and follow-up, not a single session. - **Waiting for the perfect policy.** A short, clear "what's allowed" page unblocks more usage than a long one still in review. - **Jumping straight to custom builds.** Engineering works best once people understand what AI does well in their own work. - **Treating it as an IT project.** Tools are the easy part. The real change is in how people think, decide and execute. ## When you do not need us If your team is small, already uses AI every day on real work, and someone internally has the time and interest to keep sharing what works, you may not need outside help at all. Keep going, write down what works, and revisit when you hit a problem that needs building. ## How Ruby helps We work on all three layers and try to be honest about which ones you actually need: - **AI training:** role-specific, hands-on sessions built around your team's real work, so capability stays inside the firm. - **AI strategy:** finding where AI belongs in your workflows and giving a clear recommendation, including when the answer is "not yet." - **AI engineering:** secure, customized agents and integrations for the work that should run as a system. A good first step is [one practical hour of AI training](https://rubyadvisory.com/book-a-call) for your team. ## Related pages - [Training vs consulting vs engineering: which does your firm need?](/docs/training-vs-consulting-vs-engineering) - [Where should your firm start with AI?](/docs/where-to-start-with-ai) - [Why your team isn't using the AI tools you paid for](/docs/why-ai-adoption-stalls) - [How to run an AI pilot that doesn't stall](/docs/run-an-ai-pilot-that-does-not-stall)