Read time: ~3 minutes · Series: Product Manager advanced · Previous: Rollout playbook
30-second version (internal or external)
Problem: Analysis conclusions stuck in AI chat—Monday definition follow-up gets summaries, no SQL, no signed version.
Lantide Data: Local SQL analysis IDE with built-in Agent. AI explores and drafts Plan; you annotate on documents; when ready you press Execute; Agent runs SQL, produces Report—definitions stay in SQL tabs; delivery reaches meetings.
vs chat BI: Not faster charts—auditable, rerunnable, Execute by human.
Boundary: Not cloud ETL, not fully automated reports; fits definition-aligned analysis collaboration.
2-minute version (four parts)
1. Category (what we are)
Desktop SQL analysis IDE + governed Agent. Data on local DuckDB; CSV, Excel, Parquet out of the box.
2. Difference (why not ChatGPT / Julius)
- Plan holds definition contract, not only chat consensus
- SQL tabs are pull evidence, not hidden Python
- Execute only by human separates exploration from formal pull
- Report has numbers and limitations, HTML export
3. When to use
Promo retro, funnel disputes, adhoc projects, local files not in warehouse—when "this analysis" will be questioned.
4. When not
One-off chart never reviewed; or mature BI + ETL and adhoc isn't pain.
Common pushback and short answers
| Pushback | Short answer |
|---|---|
| "We have ChatGPT" | Chat lacks Plan / SQL / Execute gate; Lantide adds sign-off delivery |
| "Won't it be slower?" | Plan review takes time; fewer meeting reversals and reruns |
| "Who's responsible if Agent is wrong?" | Human reviewed Plan before Execute; Report has limitations; SQL rerunnable |
| "Need a warehouse?" | No; local files work; optional external DB |
| "Replace BI?" | Adhoc and retro; dashboards still BI |
Wiki section titles
- Why Lantide (problem)
- What you get (Plan / SQL / Report / Execute)
- When to use / when not
- Pilot checklist (link five-step rollout)