Challenges

A knowledge base where you can ask technical "how/where to implement X?" questions. Example prompt the KB should answer: "How and where to implement a package classification flow for images coming from the dimensioners?"

Transform GitHub PR/tag-based technical changelogs into stakeholder-friendly product changelogs. Context: Current changelogs (e.g., wilkins-dimensioner) are auto-generated from PRs/tags and are mainly technical task lists. Problem: Not understandable for non-technical stakeholders; focuses on implementation instead of user value. Expected behavior: A workflow that produces changelogs that are (1) non-technical, (2) visual (images/gifs), (3) grouped by feature/value, not by engineering tickets.

Score support cases, detect missing info, and prompt for what's needed to make analysis reliable. • Define a quality rubric (minimum fields + clarity + reproducibility) • Use AI to score existing cases and suggest missing fields/questions • Add a "pre-check" when creating support tickets (auto-prompt missing info) • Create a quality dashboard + continuous improvement loop Team: Support/Ops + Product + AI + Engineering

Create a clear "source of truth" of repos, owners, and risks to improve onboarding and accountability. 1. Inventory repos (name, description, language, last activity, dependencies) into a single accessible doc/table 2. AI classification + likely owner: classify repos (core/support/tooling/legacy) and propose probable owner from activity; flag single maintainer 3. AI risk detection: flags like no README, no tests, unclear owner, low activity, high bus factor; propose 2-3 minimal actions per repo 4. Onboarding outputs: 1-page "quick map" for new joiners + proposed ownership checklist for critical repos

Move from opinions to objective prioritization using support data + AI clustering. • Connect to the support tickets/records database (inventory + access) • AI clustering by topic, severity, impact, recurrence • Weekly output: top problems + evidence + focus recommendation • Define a standard "planning input" format Team: Support/Ops + Product + Data/AI + Engineering

Generate draft SOPs from notes/chats, then validate and publish centrally. • Define a SOP template (steps, inputs/outputs, owner, DoD) • Take 3 real processes and feed the system with notes/chats • AI extracts steps, roles, dependencies and generates a SOP draft • Human validation + publish in a central repo Team: Ops/Procesos + process expert + AI/Engineering

Convert client notes/emails into a clean checklist and automatically detect what's missing. • Define a SOP/checklist template • AI extracts requirements from text and classifies them (technical, commercial, logistics, etc.) • Mark missing items and generate follow-up questions • Save final checklist ready for the proposal DoD examples: approved template; AI extraction quality documented; auto follow-up question generation; checklist used in 3 pilot proposals

Generate a complete proposal draft (scope + overview + terms) in minutes using AI. • Create a standard proposal template • Connect to product docs (overview) + terms library • Use AI to draft scope and fill sections with client data • Export Doc/PDF and provide a missing-items checklist Team: Sales, Product, Backend/Engineering, AI (NLP)