Brian Chess, who leads AI product for Oracle NetSuite, says companies must now design loops with far more AI and fewer humans.
Ask Oracle taught a controller at MCNC to run payroll when the payroll administrator was out, so payroll ran on time.
Chess says NetSuite users should start with read-only AI access and gain write permissions gradually as trust grows, with humans approving transactions entering the system of record.
NetSuite's plan to turn its AI from advisor into worker
NetSuite is betting that AI becomes the main way businesses operate their ERP, moving from answering questions to carrying out work. Brian Chess argues its customers start ahead because their data and permission systems are already structured. The real work is building trust gradually so business loops can run with much more AI and fewer people.
Takeaways
Teaching users builds trust faster than answers. Ask Oracle returns both the answer to an account question and an explanation of how it found it. At MCNC, with the payroll administrator out, it walked an untrained controller through the process and payroll ran on time.
Agents get employee-style roles and permissions. Ask Oracle carries exactly the user's permissions and cannot do anything that user cannot. Chess wants agents assigned roles modelled on employees. He sees teams start with read-only access and grant write access as trust grows.
Automate the financial close incrementally. Many companies still run the close on spreadsheets. NetSuite models the spreadsheet a finance team already uses and automates it piece by piece, because trust in AI builds one step at a time. Supply chain and workforce management are other targets.
Hallucinations mostly come from missing information. Chess says the best defence is supplying the model the data it needs. A human should still approve any transaction entering the system of record, and NetSuite logs which agent touched each transaction and when.
Skills turn one expert's knowledge into team knowledge. Written, shared procedures let a go-to person like Joe show colleagues how to do a task themselves. Institutional knowledge then spreads across the team instead of staying with a single employee.
ROI is real but rarely transferable. Customers report wins such as payroll running on time, or savings found by giving AI their vendor contracts and billing records. Chess says these results are individualised and do not carry over to the company next door.
Smaller businesses stand to gain most. With fewer specialists, small firms benefit more when software absorbs expert work. Chess ties this to NetSuite's history as the first SaaS company, which told customers they would no longer need IT staff for servers or backups.
How AI use matures inside NetSuite
Users first ask for numbers that come straight from reports, often ones they already know, to test the assistant.
Questions then move into analysis of the business.
Next comes automation of routine tasks.
Finally agents keep the system working for users twenty-four hours a day, with Ask Oracle as the first agent most users meet.
Tap a name to go there · tap a line to see why they are linked