Your ERP’s data, in your shop’s words, in minutes
We stopped making machine shops learn our software’s vocabulary. The software learns theirs.
One standard data model underneath, for every shop. During onboarding, AI reads whatever your current system exports, maps it to that model, and learns your language along the way. Then the app shows your data in your words from day one. No admin renames a single field.
Every shop has a language. One calls a job a “job.” The next calls it a “WO.” A third calls it a “Doc No.” A part on hold for inspection is “QC hold” here, “Hold-Insp” there, “behind” somewhere else. That language took years to settle. It lives on the whiteboard, on the travelers, and in how the floor talks at 6 a.m. Then the shop buys software, and the software makes everyone relearn the operation in its terms. We think that is backwards. Here is what CitrusWeb built instead, and an honest account of what is new about it and what is not.
The problem with “just rename the fields”
Big enterprise software has let admins rename fields for years. Salesforce has it. NetSuite has it. It works, and it is a real feature. But it is a manual project. Someone sits down and renames the fields by hand, one at a time, and makes those same judgment calls again for the next deployment. At the small-shop end of manufacturing, where JobBOSS, E2, and Global Shop live, that kind of deep relabeling mostly does not exist. A 20-person shop does not get to make the software speak its language. So the shop adapts to the software. New hires learn two vocabularies: the one on the floor and the one on the screen. Reports read in a dialect nobody says out loud.
What we built
Two things, and the combination is the point. First, the data model underneath never changes. There is one standard model for every shop. One set of rules, one set of features, one set of AI tools. That is what lets us build things that work the same across every customer and keep the numbers straight. We do not fork the database per shop. That path looks flexible and ends in a swamp. Second, the words on top of that model are yours. During onboarding you upload whatever your current system exports. A CSV, a spreadsheet, an accounting dump. Our AI reads it, maps your columns onto our model, and along the way it learns your language. It learns that “Doc No” is your word for a job, that “Hold-Insp” means quality hold, that a part described as “4140 shaft” means you turn steel. Then the whole app speaks back in your words. Your column headers. Your status names. Your nouns. The owner never renames a field. The mapping is a byproduct of reading the shop’s own data. We tested it on deliberately messy exports. Columns named “Sold To,” “Ext Value,” and “Prod Stage.” Status values like “WIP,” “Released,” and “Hold-Insp.” It mapped every column correctly, translated the status words to our internal states, pulled the materials out of the part descriptions, and reported how confident it was and why. Ninety-two to ninety-three percent confident, with a plain-English explanation an owner can check in ten seconds. This is the onboarding step in CitrusWeb Works, our AI-native manufacturing platform.
Why this is possible now
Column mapping by keyword matching is old and brittle. “Reference” could be anything. “Amt” might be dollars or cents. A status vocabulary matches nothing by string. This works now because a model can read the column names and the actual values together and reason about them the way an experienced person would. That was not practical two years ago. It is practical today, and it is the difference between a mapping screen a person has to fight and one that is usually right before they look.
The honest version
We are not going to tell you nobody does relabeling. Salesforce and the big ERPs clearly do. What we have not seen anyone do, at the small-manufacturer price point, is make it automatic and zero-config, sourced from the shop’s own data, with the app speaking the shop’s vocabulary from day one and no admin touching a settings screen. That is the line. Your ERP’s data, in your shop’s words, in minutes. The floor already has a language. The software should learn it, not the other way around.