If your team moved to a different AI tool next quarter, what would you actually still have?<\/p>\n
Most teams answer from the existence of an export option: the data is ours, the files can be downloaded, so we could leave. That answer is a guess dressed as a fact, and the only way to convert it into a fact is to try while the stakes are zero.<\/p>\n
What follows is a constructed rehearsal, written for this article. A content team of four has spent two years working inside one AI workspace. They set aside an afternoon to attempt the move without making it.<\/p>\n
The rules are simple. Pick one live workflow \u2014 here, the weekly production of a client newsletter. Export whatever the tool offers. Then open the export somewhere else and try to produce next week’s newsletter to the same standard, using only what came out.<\/p>\n
No system is switched off and no contract is cancelled. That matters, because a rehearsal held under real pressure becomes a migration, and a migration under pressure teaches you the same lessons at a much worse price.<\/p>\n
They stop only when they can either reproduce the newsletter or identify exactly what prevents them.<\/p>\n
The export is generous. It contains every conversation, every uploaded file, the full history in a structured format, and a folder of the prompts the team had saved deliberately.<\/p>\n
The files move perfectly. Brand guidelines, past newsletters, the tone document someone wrote in the first month \u2014 these are documents, they were documents before the tool existed, and they open anywhere.<\/p>\n
The saved prompts move too, and here the team’s first assumption breaks. The prompts open perfectly and produce weaker drafts in the test environment. They had been written against a workspace that already held the brand guidelines, the last six newsletters and a running memory of what the client rejected. Read in isolation, a prompt like draft this week’s opener in our usual voice<\/em> is a sentence with a hole in it where the context used to be.<\/p>\n The conversation history exports completely, yet it is difficult to use for the newsletter workflow in its exported form.<\/p>\n Two years of exchanges arrive as a wall of text. Somewhere inside it are the relatively few moments that actually shaped how this newsletter gets written: the time the client said the intro was too formal, the decision to stop using bullet lists, the correction that fixed how product names are capitalised. Those moments are present. They are also indistinguishable from the ordinary traffic around them, and locating one means searching through large stretches of ordinary conversation.<\/p>\n The archive is technically complete and machine-readable. That does not make it useful to the person trying to rebuild the work. A compliant export can preserve every word while losing the structure that tells an operator which words matter.<\/p>\n This is the gap between exportable and useful. The material left the tool intact and arrived somewhere else in a form where the effort to make it work again is measured in days.<\/p>\n The approved newsletters reveal the same problem in a quieter form. The files move cleanly, but the judgement attached to them does not. A new colleague can see which versions the client accepted without knowing whether the decisive factor was the opening, the length, the restrained sales language or something discussed elsewhere. The example travelled; the reason it became an example still needs reconstruction.<\/p>\n The last hour of the rehearsal is the uncomfortable one, because it deals with material that no export could have contained.<\/p>\n The team discovers, by trying to write the newsletter without it, that a good deal of how the work runs lives in one person’s head. Which client comments were preferences and which were rules. How much the tone document is actually followed versus quietly ignored. Which of the saved prompts is the current one and which three are abandoned drafts kept for safety.<\/p>\n None of this existed as an artifact anywhere in the tool, so no export could have carried it. It is genuinely the team’s own knowledge \u2014 and it was invisible to them until the rehearsal forced them to work without the environment that had been silently supplying it.<\/p>\n By the end of the afternoon the team’s material has sorted itself into three piles, and the sorting came out of attempting the work itself.<\/p>\n Moves cleanly.<\/strong> Self-contained documents that arrive usable without relying on hidden workspace context.<\/p>\n Needs reconstruction.<\/strong> Material that exports completely and arrives without whatever made it legible \u2014 prompts stripped of the context they assumed, history with the important moments buried in the ordinary ones. Recoverable, at a cost you can estimate once you have tried.<\/p>\n Stays trapped.<\/strong> Material that was never turned into an artifact: tacit operating knowledge, unwritten conventions and the difference between a rule and a preference. It remains trapped until the team makes it explicit.<\/p>\n That is the Exit Drill. It is a rehearsal, and the distinction from an audit matters: an audit tells you what exists, while a rehearsal tells you what survives being used somewhere else. The three piles are the output, and the middle one is where the useful number lives.<\/p>\n Exportable is not the same as portable. A tool passes the first test by handing your material back. It passes the second only when another person or system can reconstruct how the work was actually done.<\/p>\n An export feature may satisfy the first test while leaving the second entirely unanswered. The second test is the one that determines whether leaving is a manageable transfer or a disruptive reconstruction project, and it is rarely run until leaving is already necessary.<\/p>\n It is worth saying plainly that some switching cost is normal and does not, by itself, indicate harmful lock-in. Any tool a team uses well accumulates fit \u2014 shortcuts, conventions, a shape the work has grown into. That is the ordinary price of using something properly, and it appears wherever people get good at anything.<\/p>\n The cost worth worrying about is different: it is the part where the material exists, exports, and still needs its understanding rebuilt before anyone can use it. That gap is measurable, it grows quietly, and it is the thing a drill exposes.<\/p>\n Take one live workflow \u2014 the one you would be most reluctant to rebuild \u2014 and give it an afternoon.<\/p>\n Export what the tool offers, open it somewhere else, and try to produce one real piece of work using only what came out. Then sort what you learned into the three piles and write down, for the middle pile, roughly how long reconstruction would take.<\/p>\n That number is the useful output, and it is worth writing down even when it turns out to be small. It converts a vague unease about being locked in into a figure you can weigh against a price difference, a feature, or a contract renewal. And it tends to move the third pile too, because the fastest way to discover which operating knowledge exists only in someone’s head<\/a> is to try working without them in the room. The same folder has a second problem worth an afternoon: two prompts that do one job<\/a>. And if you want the gap priced rather than guessed, the diagnostic<\/a> does that in writing.<\/p>\n","protected":false},"excerpt":{"rendered":" If your team moved to a different AI tool next quarter, what would you actually still have? Most teams answer from the existence of an export option: the data is ours, the files can be downloaded, so we could leave. That answer is a guess dressed as a fact, and the only way to convert it into a fact is to try while the stakes are zero. What follows is a constructed rehearsal, written for this article. A content team of four has spent two years working inside one AI workspace. They set aside an afternoon to attempt the move without making it. The afternoon, and the rules of the exercise The rules are simple. Pick one live workflow \u2014 here, the weekly production of a client newsletter. Export whatever the tool offers. Then open the export somewhere else and try to produce next week’s newsletter to the same standard, using only what came out. No system is switched off and no contract is cancelled. That matters, because a rehearsal held under real pressure becomes a migration, and a migration under pressure teaches you the same lessons at a much worse price. They stop only when they can either reproduce the newsletter or identify exactly what prevents them. What came out of the export The export is generous. It contains every conversation, every uploaded file, the full history in a structured format, and a folder of the prompts the team had saved deliberately. The files move perfectly. Brand guidelines, past newsletters, the tone document someone wrote in the first month \u2014 these are documents, they were documents before the tool existed, and they open anywhere. The saved prompts move too, and here the team’s first assumption breaks. The prompts open perfectly and produce weaker drafts in the test environment. They had been written against a workspace that already held the brand guidelines, the last six newsletters and a running memory of what the client rejected. Read in isolation, a prompt like draft this week’s opener in our usual voice is a sentence with a hole in it where the context used to be. What the export contained but could not explain The conversation history exports completely, yet it is difficult to use for the newsletter workflow in its exported form. Two years of exchanges arrive as a wall of text. Somewhere inside it are the relatively few moments that actually shaped how this newsletter gets written: the time the client said the intro was too formal, the decision to stop using bullet lists, the correction that fixed how product names are capitalised. Those moments are present. They are also indistinguishable from the ordinary traffic around them, and locating one means searching through large stretches of ordinary conversation. The archive is technically complete and machine-readable. That does not make it useful to the person trying to rebuild the work. A compliant export can preserve every word while losing the structure that tells an operator which words matter. This is the gap between exportable and useful. The material left the tool intact and arrived somewhere else in a form where the effort to make it work again is measured in days. Approved examples travel without their reasons The approved newsletters reveal the same problem in a quieter form. The files move cleanly, but the judgement attached to them does not. A new colleague can see which versions the client accepted without knowing whether the decisive factor was the opening, the length, the restrained sales language or something discussed elsewhere. The example travelled; the reason it became an example still needs reconstruction. The material that lived only in people The last hour of the rehearsal is the uncomfortable one, because it deals with material that no export could have contained. The team discovers, by trying to write the newsletter without it, that a good deal of how the work runs lives in one person’s head. Which client comments were preferences and which were rules. How much the tone document is actually followed versus quietly ignored. Which of the saved prompts is the current one and which three are abandoned drafts kept for safety. None of this existed as an artifact anywhere in the tool, so no export could have carried it. It is genuinely the team’s own knowledge \u2014 and it was invisible to them until the rehearsal forced them to work without the environment that had been silently supplying it. The Exit Drill By the end of the afternoon the team’s material has sorted itself into three piles, and the sorting came out of attempting the work itself. Moves cleanly. Self-contained documents that arrive usable without relying on hidden workspace context. Needs reconstruction. Material that exports completely and arrives without whatever made it legible \u2014 prompts stripped of the context they assumed, history with the important moments buried in the ordinary ones. Recoverable, at a cost you can estimate once you have tried. Stays trapped. Material that was never turned into an artifact: tacit operating knowledge, unwritten conventions and the difference between a rule and a preference. It remains trapped until the team makes it explicit. That is the Exit Drill. It is a rehearsal, and the distinction from an audit matters: an audit tells you what exists, while a rehearsal tells you what survives being used somewhere else. The three piles are the output, and the middle one is where the useful number lives. Exportable and portable are different tests Exportable is not the same as portable. A tool passes the first test by handing your material back. It passes the second only when another person or system can reconstruct how the work was actually done. An export feature may satisfy the first test while leaving the second entirely unanswered. The second test is the one that determines whether leaving is a manageable transfer or a disruptive reconstruction project, and it is rarely run until leaving is already necessary. Normal switching cost, and the kind worth worrying about It is worth saying plainly that some switching cost is normal and does not, by itself, indicate harmful lock-in. Any tool a team uses well accumulates fit \u2014 shortcuts, conventions, a shape the work has grown into. That is the ordinary price of using something properly, and it appears wherever people get good at anything. The cost worth worrying about is different: it is the part where the material exists, exports, and still needs its understanding rebuilt before anyone can use it. That gap is measurable, it grows quietly, and it is the thing a drill exposes. Run the drill while it is still a rehearsal Take one live workflow \u2014 the one you would be most reluctant to rebuild \u2014 and give it an afternoon. Export what the tool offers, open it somewhere else, and try to produce one real piece of work using only what came out. Then sort what you learned into the three piles and write down, for the middle pile, roughly how long reconstruction would take. That number is the useful output, and it is worth writing down even when it turns out to be small. It converts a vague unease about being locked in into a figure you can weigh against a price difference, a feature, or a contract renewal. And it tends to move the third pile too, because the fastest way to discover which operating knowledge exists only in someone’s head is to try working without them in the room. The same folder has a second problem worth an afternoon: two prompts that do one job. And if you want the gap priced rather than guessed, the diagnostic does that in writing.<\/p>\n","protected":false},"author":113,"featured_media":34805,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"drb_seo_title":"Exportable Is Not Portable: The AI Exit Drill","drb_seo_desc":"An export button returns your material. It does not return the understanding that made it usable. Run one workflow elsewhere and price the gap in days.","footnotes":""},"categories":[1629],"tags":[],"class_list":["post-34190","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-systems-operations"],"_links":{"self":[{"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/posts\/34190","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/users\/113"}],"replies":[{"embeddable":true,"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/comments?post=34190"}],"version-history":[{"count":5,"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/posts\/34190\/revisions"}],"predecessor-version":[{"id":34791,"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/posts\/34190\/revisions\/34791"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/media\/34805"}],"wp:attachment":[{"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/media?parent=34190"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/categories?post=34190"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/dr-business.com\/en\/wp-json\/wp\/v2\/tags?post=34190"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}What the export contained but could not explain<\/h2>\n
Approved examples travel without their reasons<\/h2>\n
The material that lived only in people<\/h2>\n
The Exit Drill<\/h2>\n
Exportable and portable are different tests<\/h2>\n
Normal switching cost, and the kind worth worrying about<\/h2>\n
Run the drill while it is still a rehearsal<\/h2>\n