Generative<\/strong> AI produces new content \u2014 text, mostly, in a business system.<\/p>\n That distinguishes it from the other two kinds you meet.<\/p>\n Extraction<\/strong> reads a document and pulls out what is there. The invoice total either is 4,500 or it is not.<\/p>\n Prediction<\/strong> estimates an outcome from history. The forecast is 400 units, and in three months you find out how close that was.<\/p>\n Generation<\/strong> writes something that did not exist. There is no right answer to check against, only a judgement about whether it is good and true.<\/p>\n That difference is the whole subject of this article.<\/strong> Extraction and prediction can be verified. Generation has to be read.<\/p>\n Six places, roughly in order of how safe they are.<\/p>\n Internal summaries.<\/strong> A long email thread, a customer’s history, a project’s status. Nobody outside sees it.<\/p>\n Internal notes.<\/strong> Turning a rough call note into something a colleague can act on.<\/p>\n Product descriptions.<\/strong> Written from attributes, at scale.<\/p>\n Customer email drafts.<\/strong> Replies and follow-ups, edited before sending.<\/p>\n Knowledge base articles.<\/strong> First drafts of support content.<\/p>\n Report commentary.<\/strong> A written explanation alongside the numbers.<\/p>\n FIGURE 1: SIX USES, SAFEST FIRST<\/p>\n Internal summaries<\/p>\n Call notes and internal write-ups<\/p>\n Product descriptions<\/p>\n Customer emails and articles<\/p>\n Here is what makes generative AI different from every other kind of software you have used.<\/p>\n Bad software produces obviously bad output.<\/strong> A broken report shows an error, or numbers that are visibly wrong. You notice.<\/p>\n Generative AI produces well-written wrong output.<\/strong> A product description with a specification it invented reads exactly as well as one that is correct. There is no formatting cue, no error message, nothing to catch your eye.<\/p>\n Confidence is not correlated with accuracy.<\/strong> This is the single most important thing to understand about it.<\/p>\n The practical consequence: everything generated needs reading by someone who knows whether it is true.<\/strong> Not proofreading for grammar \u2014 checking for facts.<\/p>\n Being specific is more useful than a general warning.<\/p>\n Invented specifications.<\/strong> Dimensions, materials, compatibility, certifications. It will produce plausible ones.<\/p>\n Wrong prices or terms.<\/strong> It has no reliable knowledge of your commercial terms unless given them.<\/p>\n Overstated claims.<\/strong> “Industry-leading”, “guaranteed”, “fastest”. These create expectations and occasionally legal exposure.<\/p>\n Tone mismatch.<\/strong> Fine in isolation, wrong for your company or that customer.<\/p>\n Confident answers to questions it cannot know.<\/strong> Lead times, availability, whether you can do something.<\/p>\n FIGURE 2: WHERE GENERATION IS SAFE AND WHERE IT IS NOT<\/p>\n Lower risk<\/p>\n Higher risk<\/p>\n Three situations where it is clearly worth it.<\/p>\n Volume that is otherwise impossible.<\/strong> Four thousand product descriptions is weeks of writing. Generated and reviewed, it is days.<\/p>\n The blank page.<\/strong> For many people the hardest part is starting. A rough first draft to react to is faster than writing from nothing.<\/p>\n Repetitive variation.<\/strong> The same message adapted for twelve customer segments.<\/p>\n The pattern:<\/strong> it is most valuable where the volume is high, the content is routine, and a knowledgeable person is reviewing.<\/p>\n Content requiring specific expertise you have not given it.<\/strong> Technical documentation, regulatory text, anything where being subtly wrong matters.<\/p>\n Anything genuinely original.<\/strong> It produces competent, conventional output. If your differentiation is your voice, generated text will flatten it.<\/p>\n Content that must be legally precise.<\/strong> Terms, contracts, compliance statements.<\/p>\n Small volumes.<\/strong> If you need three descriptions, write them. The review takes as long as the writing.<\/p>\n Five practices.<\/p>\n Give it examples.<\/strong> Showing it three descriptions you like produces far better output than describing what you want. This is the single highest-return technique.<\/p>\n Give it the facts.<\/strong> Feed it real attributes from your product records rather than letting it fill gaps. Most invented specifications come from unfilled gaps.<\/p>\n Review for truth, not style.<\/strong> The grammar will be fine. The question is whether the claims are correct.<\/p>\n Have the right person review.<\/strong> Somebody who knows the products, not just somebody who can read.<\/p>\n Keep a short list of things it must never claim.<\/strong> Certifications, guarantees, compatibility. Make it an explicit rule.<\/p>\n FIGURE 3: THREE HABITS THAT IMPROVE OUTPUT MOST<\/p>\n Show examples<\/p>\n Supply real facts<\/p>\n Review for truth<\/p>\n Worth naming, because it develops slowly.<\/p>\n Review quality decays.<\/strong> In the first week everything is read carefully. By the third month, output has been consistently good and reading becomes skimming.<\/p>\n Then one description with an invented certification gets published across four thousand products.<\/p>\n Two safeguards:<\/strong><\/p>\n Spot-check a sample deliberately<\/strong>, even after you trust it. A fixed percentage, reviewed properly.<\/p>\n Never remove review from anything published at scale.<\/strong> The volume that makes generation valuable is the same volume that makes one error expensive.<\/p>\n Odoo includes generative AI in the screens where the content is needed \u2014 product forms, email composition, knowledge articles, record summaries.<\/p>\n The practical advantage is context. Because it is one database, a draft can use the actual product attributes, the actual customer history, the actual order details. Less is invented, because less is missing.<\/p>\n That reduces the error rate. It does not remove the need to read.<\/p>\n Generative AI is very good at volume and very good at first drafts<\/strong>.<\/p>\n Its distinguishing feature is that wrong output looks exactly as good as right output. There is no error message and no visual cue.<\/p>\n Give it examples and real facts. Review for truth rather than style. Never remove review from anything published at scale.<\/strong><\/p>\n Used that way it saves weeks of routine writing. Used without review, it publishes confident errors faster than any tool you have owned.<\/p>\n Facing a catalogue of descriptions or a backlog of content?<\/p>\n Get in touch. We will set it up with your real product data and a review process that survives past the first month.<\/p>\n<\/div>","protected":false},"excerpt":{"rendered":" A specific word Generative AI produces new content \u2014 text, mostly, in a business system. That distinguishes it from the other two kinds you meet. Extraction reads a document and pulls out what is there. The invoice total either is 4,500 or it is not. Prediction estimates an outcome from history. The forecast is 400 […]<\/p>\n","protected":false},"author":3,"featured_media":9214,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[26],"tags":[],"class_list":["post-9213","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai"],"yoast_head":"\nWhere it is used in an ERP<\/h2>\n
\n
\n
\n
\n
The problem with fluency<\/h2>\n
Where errors actually appear<\/h2>\n
\n
\n
Where it genuinely earns its place<\/h2>\n
Where it is a poor fit<\/h2>\n
Making it work<\/h2>\n
\n
\n
\n
The quiet risk<\/h2>\n
What this means for Odoo<\/h2>\n
The short version<\/h2>\n