nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Quick Take #7: AI meets analytics

AllisonHines

In this Quick Take, Dan Valente (@Daniel_Valente), Scott Allen (@scott_allen), and Kristen Slappey (@kristen-slappey) explore the ever-evolving intersection of artificial intelligence and analytics, tackling how JMP maintains its distinct edge.

Drawing on insights from JMP user David Meintrup, the discussion highlighted that JMP’s seamless, trusted workflow offers a distinct advantage over the cumbersome back-and-forth of traditional AI prompting. And using Anderson Mayfield’s perspective on data prep, the team acknowledged that while JMP excels at heavy lifting and getting messy data ready, its true value lies in empowering the exploratory analysis that follows. (Be sure to check out the next issue of JMP Foreword, coming in October, to read the full articles by Meintrup and Mayfield.)

Reflecting on these customer perspectives and product philosophy, the conversation focused on ensuring AI enables human capability rather than replacing statistical intuition, with a particular look at how innovations like JMP LearnBot support this mission.

If you have questions about the chat or want to dig deeper, leave a comment in this article or reach out to Valente, Allen, or Slappey directly.

If you have a suggestion, please share your ideas for future Quick Take topics via the Feedback Form. You can also find the request link in the Helpful Links section of the home page. You can find all Quick Takes on the Company Updates 2026 page.

Comments

saradoudt

Very exciting! Here is the link for sales teams to request a product management customer engagement (if customers want to learn more about the roadmap, whats coming around AI) that Scott and Dan mentioned: https://jmpdiscovery.atlassian.net/wiki/spaces/SALESENABLEMENT/pages/570655157/Product+Management+Customer+Engagements+PMCE

Daniel_Valente

Here is the blog post we mention in the presentation: https://community.jmp.com/t5/JMPer-Cable-Blog/An-infrastructure-that-allows-organizations-to-implement-AI-on/ba-p/964862

LaurenAldrich

I love these quick takes! It was great to learn more about how we're approaching the use of AI with our already amazing product.

Tim_Pierce_JMP

Blog Post is greatly appreciated. I am using today in follow up to a meeting earlier this week.

russ_wolfinger

Thanks for your commentary here Dan, Kristen, and Scott.   A few “yes, and” follow-ups:

LearnBot is great and has been our first successful foray into AI. Its simple, powerful chat interface is perfect for asking general questions about JMP.   It has two key limitations:

1. LearnBot requires a public internet connection.   We and customers should not prompt it with any kind of proprietary questions or data.

2. While you can ask LearnBot to generate JSL and copy-paste-run it in a JMP Script window, this practice is strongly discouraged. LearnBot generates JSL errors 35-40% of the time and provides no safety guards on the JSL.

Key notes on Assistant:

- After a year of intense R&D and rapid response to tons of feedback, the latest version of Assistant has the JSL error rate down to 5% using doc retrieval and auto-correction + now includes JSL safety scanning with both blocking and flagging.

- Besides JMP CustomGPT (the same public provider as LearnBot), Assistant offers a comprehensive set of 20 bring-your-own LLM providers to plug into any kind of AI-aware software ecosystem in a secure way.

- Assistant has a new Auto mode that automatically recognizes if you are asking a question and or want to run a task in JMP with generated JSL along with strict safety checks.

- Assistant’s chat has modern styling and is interactive with editable, moveable, and deletable cards. You can chat both in a main window or now directly in JMP data and report windows.

- Many customers are recognizing Assistant’s broad and powerful functionality and several are evaluating it for broad deployment.

By the way, both David Meintrup and Anderson Mayfield are fans of Assistant.

If you want to do things in JMP with your data in a safe way using natural language prompting, try Assistant. The latest version is here (not yet published to the Marketplace): https://jmpdiscovery.atlassian.net/wiki/spaces/JMPPRODMGMT/pages/2710175789/Assistant+Feedback+Page

Please keep the feedback on Assistant coming; it has been contentious at times but overall tremendous!

yusuke_ono

I think that I can reconfirm the current guiding principles for AI in JMP. But because of my poor English skill, I cannot catch well the following three small minor points. I am a tester, and I do not often communicate or talk with our JMP users, so the following questions are based mainly on my personal interests.

(1) In the slide titled "What we shouldn't do", it is said that "[We should not] Provide capabilities that generate AND execute JSL (in product or in extension)." But the current Assistant extension in JMP Marketplace can generate AND execute JSL. Is JMP Assistance just an exception from this basic principle? Or am I misunderstanding what "generate AND execute JSL" means?

(2) In the slide titled "AI Integration guiding principles", it is said that "Utilize LLMs intelligently to improve search, access, and retrieval of information from results published to JMP Live." I cannot guess well what this feature is. Is there an internal Confluence page in which this feature is explained?

(3) It seems to me that the main principle of our parent company, SAS Institute, is AI First, and that SAS' principle is inconsistent with JMP's one. If so, how can we reconcile this inconsistency? (In the past, we said something that "SAS is a truck and JMP is a sports car.")