AI PM team Q1 2025 goals (#10292)

contents/teams/ai-product-manager/objectives.mdx

@@ -1,13 +1,45 @@
-### Q4 2024 objectives
+### Q1 2025 objectives
 
-This quarter is about answering product questions _really_ well.
+This quarter is about:
 
-Components we see as necessary:
+- excellence in solving level 1 analysis problems (broad, but straightforward)
+- reliable handling of level 2 analysis problems (deeper, ambiguous)
+- AI assistance as a core part of PostHog
 
-1. Agent's awareness of the product in question
-1. Proactive approach, e.g. with personalized question suggestions
-1. Ability to create insights of all types, including tweaking their options
-1. SQL skills, aware of ClickHouse and HogQL features
-1. Persistence of conversations
-1. LLM observability
-1. Metrics on usage of Max
+Our projects:
+
+#### Core intelligence
+
+- Long-term memory (**high impact**, **low effort**)
+- Retrieval of actions and cohorts for context (**high impact**, high effort)
+- Retrieval of insights and dashboards for context (medium impact, high effort)
+- Integrate with customer's docs for product understanding (medium impact, high effort)
+- Understanding of data warehouse, tuned for common sources like Stripe (**high impact**, high effort)
+- More examples in prompts, e.g. taxonomy planning or summarization (low impact, **low effort**)
+
+#### New skills
+
+- Automated pull request with event capture/SDK features (**high impact**, high effort)
+- HogQL insights (**high impact**, high effort)
+- PostHog docs search (**high impact**, medium effort)
+- Analysis of session recordings (**high impact**, high effort)
+- Analysis of experiment results (medium impact, **low effort**)
+- Analysis of surveys (medium impact, **low effort**)
+- Using feature flags in analysis (low impact, **low effort**)
+
+#### User experience
+
+- Max in the sidebar (**high impact**, **low effort**)
+- "Edit with Max" button when hovering over existing insight (**high impact**, medium effort)
+- Max on the insight page (**high impact**, medium effort)
+- Max in "New dashboard" (**high impact**, high effort)
+- Max in the dashboard "Add insight" flow (**high impact**, high effort)
+- History of conversations (medium impact, **low effort**)
+- Hard-coded great question suggestions (medium impact, **low effort**)
+
+#### Evaluation
+
+- Evaluation tests speeded up 3x (**high impact**, **low effort**)
+- Junk in eval test data (**high impact**, medium effort)
+- Comparison of OpenAI vs. Anthropic vs. alternative LLMs quality (low impact, **low effort**)
+- LLM observability in production (medium impact, **low effort**)