Updates

Verification Mode is live in PlotStudio AI. Before any analysis runs, you can now review the AI's plan, ask questions about it, propose your own methodology, and edit it directly. The model proposes a starting point, you shape it. A few things you can do: → Ask the agent why it picked a specific test or method → Propose your own approach ("use a fixed-effects panel regression instead of pooled OLS") → Request changes in plain English ("drop the cardiovascular task and add an outlier check first") → See a diff summary of every edit before it lands → Undo any change with one click → Execute the final plan when you're ready This was the most-requested feature from data analysts who wanted to drive their own analysis instead of letting the model run unchecked. Especially in econometrics and other domains where the choice of model, control variables, or test specification is the entire ballgame, you can't just hand that to an agent and hope. The point of agentic analytics isn't to replace judgment. It's to do the mechanical work fast enough that the analyst spends more time on the parts that actually matter: defining the question, choosing the methodology, and deciding what to do with the answers. Verification Mode lets you work alongside the agent instead of after it.

Export your entire analysis in one click. A Jupyter notebook with every step. Your source datasets, derived tables, and all visualizations as PNGs. Open it anywhere. Hit Run All. Same results. Hand it to a reviewer, a collaborator, a professor — they see exactly what was done. No black box. Just a ZIP file that stands on its own. Reproducibility, built in.

We just shipped built-in PII detection — and it works before your data even uploads. When you drop your data, our AI scans the column names — just the names, never your actual data — and flags anything that looks like personally identifiable information. Names, emails, SSNs, phone numbers, addresses, IP addresses. Before a single row hits disk, you choose what to do: 👉 Mask it (j*@g*.com) 👉 Hash it (one-way scramble for joins) 👉 Pseudonymize it (consistent labels for grouping) 👉 Drop it entirely 👉 Keep it One click. Done before upload. Your data gets anonymized, then analyzed. Full analytical value, zero privacy risk. Most teams either skip anonymization because it's too much friction, or do it manually in a separate step that nobody does consistently. We made it zero-friction. The AI flags it. You confirm. Upload. That's it. Privacy shouldn't be a separate workflow. It should be built into the moment you first touch the data.

We shipped keyboard shortcuts because a user asked for them — the next day. Someone using PlotStudio told us: "I love the different analysis modes, but switching between them should be faster." PlotStudio has five analysis modes: 👉 Auto — AI chooses the right depth for your question 👉 Quick Analysis — fast answers like "plot revenue by region" 👉 Planned Analysis — deep, multi-step investigations with stakeholder-ready output 👉 Response Only — skip the code, just get the answer 👉 Workspace Insights — connect patterns across all your analyses Switching between them should feel effortless. Now it's one keystroke.