AGP Picks
View all

YTBE says NetSuite AI works better after fixing field descriptions

Nov. 2, 2026
By AI, Created 17:45 UTC, Nov 02, 2026, AGP -

YTBE, Inc. says a SuiteWorld 2026 demo showed how small data-description changes can flip NetSuite’s built-in AI from wrong to right. The review argues that companies need clearer field notes, role-based testing and AI unit planning before they can rely on NetSuite’s AI in production.

Why it matters: - NetSuite users may get inaccurate AI answers if their records and fields are poorly labeled. - YTBE says companies that clean up their data definitions now will be better positioned to use ERP AI effectively next year. - The issue affects day-to-day work such as customer follow-up, finance reviews and operations planning.

What happened: - At Oracle's SuiteWorld 2026 conference, an Oracle presenter asked NetSuite's built-in assistant, Ask Oracle, which customers needed follow-up using a custom record type. - The first answer found no customers. - After short plain-English descriptions were added to the record type and its fields, the same question returned the correct customer. - YTBE, Inc., which attended the sessions, published a review based on the demonstration and on its own client work.

The details: - YTBE concludes that NetSuite's AI is only as useful as the descriptions companies give their own data. - NetSuite's AI Descriptions feature lets an administrator attach descriptions to a record type or field, with up to 280 characters per description. - Writing useful descriptions requires input from the people who understand the data and an administrator or NetSuite partner to enter them. - Testing is a separate step and should use real questions under each role, with answers checked against the underlying data. - YTBE says that evaluation is where many teams need help. - AI usage is measured in AI units. - Each General Access user adds 1,000 units a month to a shared annual pool. - Unused units do not roll over. - Additional units can be purchased through a NetSuite account representative. - YTBE says a 20-user account would have about 240,000 units a year. - In a developer session, one scheduled agent run used about 74 units. - YTBE's own testing found that analytics questions start at about 30 units and multi-step tasks use more. - A daily agent at that level would use about 27,000 units a year, or roughly 11% of the pool, before testing. - Assistants connected through Oracle's AI Connector use no NetSuite AI units, according to Oracle's documentation and YTBE's experience. - Answers vary by role. - A sales rep and a controller can ask the same question and correctly get different answers. - A scheduled agent only sees what the role chosen at setup allows. - Transactions arrive as drafts for review. - AI-filled fields are marked "Needs review." - Agents stay off until an Administrator turns them on. - YTBE says many AI dashboard examples shown at SuiteWorld, such as sales against plan or items nearing safety stock, can be handled by standard NetSuite reports instead of AI. - Those reports can return the same answer every time, be checked line by line and use no AI units. - YTBE's approach is to use AI to build those reports and scripts. - The full review is available at the company's review. - YTBE also listed its AI readiness services as writing AI Descriptions, building and running role-based tests, and setting agent and AI unit policies.

Between the lines: - YTBE's critique is not that AI is useless in NetSuite, but that AI performance depends on data governance and role design. - The review also argues that some recurring business tasks do not need a language model at all. - That puts pressure on companies to separate true AI use cases from routine reporting work.

What's next: - YTBE recommends that companies start by listing the questions executives, finance and operations ask most. - The next step is to have an administrator or partner describe the fields those questions depend on, working with the people who use them. - Companies should test each question under every role, verify the answers against the data and repeat testing after changes. - Teams should monitor AI unit use from the start, decide who can turn on agents and budget for testing. - YTBE says companies should use AI to build recurring reports, not rerun them. - Hansbury says effective use requires more than a feature switch and careful planning of questions, records, steps and AI unit use.

Disclaimer: This article was produced by AGP Wire with the assistance of artificial intelligence based on original source content and has been refined to improve clarity, structure, and readability. This content is provided on an “as is” basis. While care has been taken in its preparation, it may contain inaccuracies or omissions, and readers should consult the original source and independently verify key information where appropriate. This content is for informational purposes only and does not constitute legal, financial, investment, or other professional advice.

Sign up for:

The Government Daily Review

The daily local news briefing you can trust. Every day. Subscribe now.

By signing up, you agree to our Terms & Conditions.

Share this page:

Advanced Search Options

Search for:

Search scope:

Type:

Search in:

Date range:

The last

Sort by:

Sign up for:

The Government Daily Review

The daily local news briefing you can trust. Every day. Subscribe now.

By signing up, you agree to our Terms & Conditions.