# New Hire Checklist for ERP and AI: Onboard Users, Build and Roll Out AI Tools > Source: https://netray.co/tools/new-hire-erp-user-and-ai-tool-rollout-checklist > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: Free Tools ## Short answer Two checklists in one: onboarding a new employee into your ERP, and designing, building and rolling out an AI tool on ERP data. ## Category ai-automation ## Audience - ERP administrator - Department manager - AI project lead ## Tool **Kind:** checklist **Groups:** #### New hire: ERP access and role **Items:** **Text:** Written request from the hiring manager naming job role and site **Critical:** true **Text:** Role mapped to a standard ERP role, not copied from another person **Critical:** true **Text:** Segregation of duties check done, for example no create-vendor plus pay-vendor **Critical:** true **Text:** User account created with unique ID, no shared logins **Critical:** true Approval limits set for purchasing, expenses or journals Default site, warehouse, cost center and printer set Multi-factor or single sign-on enabled where supported #### New hire: training and first week **Items:** Training sandbox or test company login given before production access Five core transactions for the role practised with a buddy Where to find reports, saved searches and dashboards shown How to raise an ERP support request explained Data entry standards covered: naming, units of measure, item codes 30 day check that access still matches actual duties #### New hire: AI assistant use **Items:** **Text:** Acceptable use rules for AI tools signed or acknowledged **Critical:** true **Text:** Told which data may never go into public AI tools, such as drawings, pricing and customer data **Critical:** true Access to the approved internal AI assistant, if one exists, granted by role Shown how to check an AI answer against ERP records before acting #### Design the AI tool **Items:** **Text:** One specific task chosen with a measurable result, for example quote drafting time **Critical:** true **Text:** Data sources listed, with an owner for each **Critical:** true **Text:** Read-only access to ERP agreed first; any write-back requires human approval **Critical:** true Roles and permissions of the ERP carried through to the AI tool Where the model runs decided: cloud API, private cloud or on-prem Sensitive and export-controlled data classified before design #### Build and test the AI tool **Items:** **Text:** Test set of at least 50 real questions with known correct answers **Critical:** true Answers cite the ERP records or documents they used Wrong or unsupported answers are measured, not guessed Prompt, model and data version recorded for each test run Logging of questions and answers with user ID enabled Latency and cost per request measured on realistic volume #### Roll out **Items:** **Text:** Pilot with five to ten users in one department first **Critical:** true Users told what the tool cannot do and when to ignore it Feedback button or channel for wrong answers **Text:** Owner named for monitoring accuracy and cost monthly **Critical:** true Rollback plan if answers prove unreliable Review at 30 and 90 days against the original measure **Scoring Note:** Aim for every critical item ticked before production access or pilot launch. Open critical items in the access groups are audit findings in waiting; open critical items in the AI groups are the usual cause of tools that nobody trusts. ## Content **Intro:** This checklist covers two jobs that arrive together: getting a new hire productive and safe in your ERP, and designing, building and rolling out an AI tool on ERP data. Tick items, then export to PDF or Excel. **Sections:** #### How to use the checklist **Body:** Use the first three groups on the day a new employee is hired and the last three when your team starts an AI project. Tick each item as it is done and treat items marked critical as gates. Export the result to PDF to keep as the onboarding record, or to Excel to assign owners and dates. If you onboard many people, copy the groups into your HR workflow so ERP access is part of the standard hire process rather than an afterthought. #### ERP onboarding: access is the risk **Body:** The common ERP onboarding failure is not training but access. New hires get a copy of a colleague's access, which brings every permission that colleague accumulated. Map the job to a defined role, check segregation of duties, and set approval limits explicitly. In SAP this means roles and authorization objects, in Infor SyteLine user groups and form permissions, in NetSuite roles, and in Oracle Fusion job roles and data security contexts. Review access again after 30 days, since real duties usually differ from the request. **Bullets:** - Never share a login between two people - Remove access promptly when someone changes role, not only when they leave - Keep the approval for each access grant on file for audits #### Designing AI tools that people use **Body:** Efficient AI tools start with one narrow task and a measure, not a general chatbot. Decide what data the tool reads, who owns it, and whether the tool may write anything back. Read-only first is the safe default; if the tool proposes changes, a person approves them. Carry the ERP's role permissions into the tool so a clerk cannot ask for payroll or cost data they cannot see in the ERP itself. Decide early whether the model runs in a cloud API or inside your network, because that choice affects confidentiality, cost and what data you can include. #### Testing and rolling out **Body:** Build a test set of real questions with known answers before you show the tool to users, and measure wrong and unsupported answers explicitly. Require citations to ERP records so users can verify. Pilot with a small group, give them an easy way to report wrong answers, and name one owner who reviews accuracy and cost monthly. Compare results at 30 and 90 days with the measure you chose at design time. If the tool is not clearly better than the existing report or search, simplify it or stop. **Cta:** Netray builds private AI tools on ERP data with read-only access, role checks and logged answers. If you want a second pair of eyes on your design or rollout plan, we can review it. **Faq:** #### What should an ERP new hire checklist include? **Answer:** A written access request, a defined role instead of a copied profile, a segregation of duties check, a unique login, approval limits, training in a sandbox, and a review of access at 30 days. Add data handling rules for AI tools. #### How do I give a new user the right ERP access? **Answer:** Map the job to a standard role, grant only that role, check for conflicting duties, and set approval limits. Avoid copying another user, since that carries over unneeded permissions that auditors will question. #### What is a checklist for building an AI tool? **Answer:** Choose one task and a measure, list data sources and owners, start read-only, keep ERP permissions, test on real questions with known answers, log usage, pilot with a small group, and review results at 30 and 90 days. #### Can an AI tool write back to the ERP? **Answer:** It can, but only with care. Treat any write as a proposal that a person approves, log it, and restrict it to low risk transactions first. Read-only is the safe starting point for most teams. ## Search terms new hire checklist ai erp, erp new user onboarding checklist, build ai tools checklist, design ai tools checklist, efficient ai tools checklist, ai tool rollout checklist, erp access request checklist, ai tool onboarding checklist --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json