In partnership with
Two weeks ago, I talked about using AI to significantly increase my output.
Today I’m sharing how I do it.
As a Product Manager at beehiiv I use claude code & cowork on a daily basis, and have been for the last few months.
For me, the goal is not to try and replace tasks entirely, but to try and reduce the time each task takes by >90% → unlocking more time for high-value tasks.
As it stands, my weekly time saving is easily approaching 12 hours per week and I’m able to do tasks on a weekly basis that I previously could only attempt once per quarter due to the time commitment.
Below are five of the most impactful skills I use on a weekly basis (with instructions, templates, and required connectors) so you can try them too.
(These can be copied directly into you claude account and then customized to suit your setup)
#1. /debug
Summary: A slack command to get claude to troubleshoot, debug and fix customer issues raised in slack thread.
How it works:
reads the thread to identify the affected user, error messages, and core problem
cross-references Sentry for stack traces and error counts, GitHub for relevant code and recent commits, and Metabase for a live SQL query on the affected user's data
classifies the issue as a user error or product bug and determines the fix path (instructions, script, or PR)
if a PR is required, creates a fully populated Linear ticket with root cause, steps to reproduce, and suggested fix, and links it back in the thread
if the issue is a user-error, it will draft a customer-facing email to the affected user (saved to Gmail, not sent)
posts a scannable debug report back into the original Slack thread, with CTA to action the fix (eg create PR)
Required connectors: Slack, Sentry, GitHub, Metabase, Linear, Gmail
#2. /prd
Summary: End-to-end PRD writer for beehiiv features. Produces three deliverables (research, PRD, implementation plan) using internal product context from Github, slack and linear.
How it works:
Creates a new folder under PRDs/ with a kebab-case feature name and confirms with the user
Phase 1 (Research): Queries Unblocked MCP for current architecture, historical decisions, related Linear issues, Slack discussions, and recent PRs. Also researches competitors. Writes findings to
research.mdPhase 2 (PRD): Drafts the full PRD following PRD_INSTRUCTIONS.md, covering overview, user stories, phased timeline, functional requirements, UI inspo, tiering/fees, analytics, non-functional requirements, risks, and open questions. Writes to prd.md
Phase 3 (Implementation): Translates the PRD into an engineering plan with architecture decisions, API contracts, migration plans, per-phase tasks, dependencies, and testing strategy. Writes to implementation.md
Required connectors: Unblocked (Slack, Linear, GitHub, internal docs)
#3. /changelog
Summary: Automatically generates a weekly customer-facing changelog report for users.
How it works:
finds all the completed tickets from the previous week and consolidates them into a list of customer-facing changes
groups the work by improvements & bug fixes
removes any mention of work that customers cant see/experience (eg backend tickets)
removes any mention of features that relate to an upcoming product launch/announcement (At beehiiv, this focuses on what is included in our fortnightly ‘Product Update’)
Required connectors: Linear, Slack (via Unblocked), Github
#4. /support-trends
Summary: Analyzes how support topics change over time, detecting volume spikes and identifying which issues are increasing or decreasing.
How it works:
Parses the latest support CSV and groups data by the chosen time period (daily by default)
Calculates period-over-period percentage changes in volume
Detects spikes where volume exceeds 2 standard deviations above the mean
Classifies each topic as increasing, decreasing, or stable
Displays volume trend, spike alerts, top 10 increasing/decreasing topics, and category-level trend
directions
Required connectors: Zendesk / metabase / CSV
#5. /livechat
Summary: Analyzes thousands of AI live chat transcripts to surface the top how-to questions and areas of user confusion, filtering out outage noise to reveal genuine product gaps.
How it works:
Locates the most recent AI livechat transcript CSV or queries metabase directly for the latest info. (Our AI livechat is provided by a tool called Ada)
Builds or updates an outages timeline by researching Slack (#devs-situationroom), Linear, and status pages via Unblocked MCP, then uses it to exclude outage-related tickets with word-boundary regex
Runs a Python script that categorizes remaining tickets as "how-to", "confusion/frustration", or both using keyword patterns from inquiry_summary text (not Ada's broad topic labels)
Generates ranked top-20 lists for how-to questions and confusion areas, plus deep dives on the top 10 with 10+ verbatim customer quotes each
Writes a full report to livechat_insights/ with executive summary, methodology, outages reference, rankings, and prioritized recommendations
Runs a verification checklist (ticket counts, false positives, duplicate categories, quote minimums) and cleans up the temp script
Required connectors: Slack, Linear, Status pages
Side note: If, like me, you're struggling to keep up with the latest AI news and updates, then The Deep View is for you.
It’s one of the easiest ways of staying up-to-date and one of the first newsletters I recommend to my friends if they’re looking to upskill:
Become An AI Expert In Just 5 Minutes
If you’re a decision maker at your company, you need to be on the bleeding edge of, well, everything. But before you go signing up for seminars, conferences, lunch ‘n learns, and all that jazz, just know there’s a far better (and simpler) way: Subscribing to The Deep View.
This daily newsletter condenses everything you need to know about the latest and greatest AI developments into a 5-minute read. Squeeze it into your morning coffee break and before you know it, you’ll be an expert too.
Subscribe right here. It’s totally free, wildly informative, and trusted by 600,000+ readers at Google, Meta, Microsoft, and beyond.
Before you go, I have two favors to ask:
If you know a PM, founder or creator that would find this useful, please send them this link ( https://the-release-notes.beehiiv.com/how-we-use-claude-at-beehiiv ) or forward this email on.
If you have any feedback, I'd love to hear it!
Thanks,
Jake
