1. [Writing](/writing) 2. / 3. The Lock-In Is Moving 2026-05-11 · 3 min read # The Lock-In Is Moving If customers could take their context to a competitor, what would they still choose you for? Scroll to read Read with... Pick an agent - [ChatGPT](https://chatgpt.com/?q=Read%20this%20piece%3A%20The%20Lock-In%20Is%20Moving%0AURL%3A%20https%3A%2F%2Fwww.assemblydigital.com%2Fwriting%2Fthe-lock-in-is-moving%0A%0APage%20context%3A%0ASummary%3A%20AI%20tools%20can%20hold%20the%20reasoning%20behind%20your%20team's%20decisions.%20What%20happens%20when%20you%20switch%2C%20and%20what%20does%20that%20mean%20for%20customer%20retention%3F%0ATags%3A%20decentralized-ai%2C%20context-engineering%2C%20knowledge-graph%2C%20agents%2C%20ai-infrastructure%0A-%20Subhead%3A%20If%20customers%20could%20take%20their%20context%20to%20a%20competitor%2C%20what%20would%20they%20still%20choose%20you%20for%3F%0A-%203%20min%20read%0A%0AReturn%3A%0A1\)%20The%20core%20argument%20in%20one%20sentence.%0A2\)%20Three%20claims%20and%20where%20each%20could%20fail.%0A3\)%20The%20most%20useful%20takeaway%20for%20someone%20applying%20this.%0A4\)%20One%20practical%20next%20step%20to%20test%20this%20week.%0A%0AUse%20the%20page%20as%20the%20source%20of%20truth.%20Keep%20it%20concise%20and%20specific.) - [Claude](https://claude.ai/new?q=Read%20this%20piece%3A%20The%20Lock-In%20Is%20Moving%0AURL%3A%20https%3A%2F%2Fwww.assemblydigital.com%2Fwriting%2Fthe-lock-in-is-moving%0A%0APage%20context%3A%0ASummary%3A%20AI%20tools%20can%20hold%20the%20reasoning%20behind%20your%20team's%20decisions.%20What%20happens%20when%20you%20switch%2C%20and%20what%20does%20that%20mean%20for%20customer%20retention%3F%0ATags%3A%20decentralized-ai%2C%20context-engineering%2C%20knowledge-graph%2C%20agents%2C%20ai-infrastructure%0A-%20Subhead%3A%20If%20customers%20could%20take%20their%20context%20to%20a%20competitor%2C%20what%20would%20they%20still%20choose%20you%20for%3F%0A-%203%20min%20read%0A%0AReturn%3A%0A1\)%20The%20core%20argument%20in%20one%20sentence.%0A2\)%20Three%20claims%20and%20where%20each%20could%20fail.%0A3\)%20The%20most%20useful%20takeaway%20for%20someone%20applying%20this.%0A4\)%20One%20practical%20next%20step%20to%20test%20this%20week.%0A%0AUse%20the%20page%20as%20the%20source%20of%20truth.%20Keep%20it%20concise%20and%20specific.) - [Perplexity](https://www.perplexity.ai/search/new?q=Read%20this%20piece%3A%20The%20Lock-In%20Is%20Moving%0AURL%3A%20https%3A%2F%2Fwww.assemblydigital.com%2Fwriting%2Fthe-lock-in-is-moving%0A%0APage%20context%3A%0ASummary%3A%20AI%20tools%20can%20hold%20the%20reasoning%20behind%20your%20team's%20decisions.%20What%20happens%20when%20you%20switch%2C%20and%20what%20does%20that%20mean%20for%20customer%20retention%3F%0ATags%3A%20decentralized-ai%2C%20context-engineering%2C%20knowledge-graph%2C%20agents%2C%20ai-infrastructure%0A-%20Subhead%3A%20If%20customers%20could%20take%20their%20context%20to%20a%20competitor%2C%20what%20would%20they%20still%20choose%20you%20for%3F%0A-%203%20min%20read%0A%0AReturn%3A%0A1\)%20The%20core%20argument%20in%20one%20sentence.%0A2\)%20Three%20claims%20and%20where%20each%20could%20fail.%0A3\)%20The%20most%20useful%20takeaway%20for%20someone%20applying%20this.%0A4\)%20One%20practical%20next%20step%20to%20test%20this%20week.%0A%0AUse%20the%20page%20as%20the%20source%20of%20truth.%20Keep%20it%20concise%20and%20specific.) - [Your Claw](https://openclaw.ai) - [Gemini](https://gemini.google.com/app) Gemini and Your Claw use copy and paste. I've worked on the growth motions that make products stickier. If a company has three products and customers using two or more are less likely to churn, I want to understand why, then help more customers get value from the second. Lower churn can mean customers are getting more value, that leaving takes more work, or both. The same integration can make a product more useful and more painful to replace. I want to know which of those things we're increasing when we celebrate adoption. Say you're working with an agent to decide where next month's marketing budget should go. You and the team work through why one channel brings in plenty of sign-ups but few customers who stick around, and why another needs more time before you can judge it. Next month, you return to that conversation because it's where those decisions and the reasons behind them live. You can take the campaign reports and budget plan to another AI tool and still leave behind much of what the team learned making those decisions. If that context is buried in old conversations, someone has to go back through them and make it usable again. A better model can still send the team back through problems they've already solved if it doesn't have that context. I'd rather they spend that time learning more about our customers or testing the next idea. Those corrections and decisions are company knowledge, even when they happen in a chat. For tasks several models can handle, I think more of AI's lock-in will come from having to put that knowledge back together before another tool can do useful work. Choosing where a model or agent runs doesn't settle where your company's knowledge lives. [Baseten](https://www.baseten.co/) and [Modal](https://modal.com/docs/guide) offer model serving and compute, while [Blaxel](https://docs.blaxel.ai/Overview) and [Daytona](https://www.daytona.io/) provide sandboxes for agents. [Akash](https://akash.network/docs/providers/getting-started/should-i-run-a-provider/) offers a decentralized compute marketplace, and [Gensyn](https://docs.gensyn.ai/) is building decentralized AI infrastructure that includes verification. In my [GTM knowledge graph](/writing/you-dont-need-better-prompts-you-need-a-knowledge-graph), I keep the source of truth outside any individual AI product. I want to be able to change tools without giving up the work I've put into making the output useful. When a correction goes into those versioned files, I can bring it to the next tool along with the evidence behind it. Anything I've only corrected in a chat still needs to make it back into the graph. A graph doesn't take care of the whole move. You'll still need to reconnect tools, check what the agent can access, and confirm which actions need approval. I'll accept some dependency on a product that saves me and my team time and helps us do better work. I still expect us to be able to take what we've learned with us, including why we moved that budget and what we'd need to see before moving it again. That work should give us a head start on the next decision, even if we make it in a different tool. For a product I'm helping grow, I'd put this question in the next retention review: if customers could take their context to a competitor, what would they still choose us for? Before we push more customers toward product number two, we should have a better answer than how much work it would be to leave. [ Previous post You Don’t Need Better Prompts. You Need a Knowledge Graph. ←](/writing/you-dont-need-better-prompts-you-need-a-knowledge-graph)[ Next post Babe, wake up - it's 3/6/9 weeks now. →](/writing/the-first-90-days-are-now-9-weeks) Read next [01You Don’t Need Better Prompts. You Need a Knowledge Graph.2026-03-09 · 23 min read→](/writing/you-dont-need-better-prompts-you-need-a-knowledge-graph)[02When Your Differentiation Becomes a Release Note2026-02-23 · 11 min read→](/writing/when-your-differentiation-becomes-a-release-note)[03How Small Teams Win Against Giants2026-02-27 · 7 min read→](/writing/how-small-teams-win-against-giants) Decision support ## Fast answers, zero fluff The core framing, audience fit, and time commitment in under a minute. 01Isn't this just product stickiness? Customers can stay because a product is useful, because leaving takes work, or both. With AI tools, some of that work is recovering the reasoning, corrections, and decisions a team has left in conversations. I want to understand how much of our retention comes from continued value and how much comes from the work involved in switching. 02What can get left behind when you switch AI tools? You might take the campaign reports and budget plan with you but leave behind the conversation that explains why you moved the budget. Someone then has to recover that context before the new tool can build on what the team has already learned. 03How does a knowledge graph help? I keep my source of truth in versioned files outside any individual AI product, so I can bring corrections and the evidence behind them to the next tool. Anything corrected only in chat still needs to make it into the graph. Moving workflows also means reconnecting tools and checking access and approvals. 04What would you ask in a retention review? If customers could take their context to a competitor, what would they still choose us for? Before pushing more customers toward a second product, I'd want a better answer than how much work it would be to leave.