dusk botan ai assistant for product managers with scattered context
give a product bot a named project and a few relevant channels. it can read linear or jira issues, find the discussion behind them in slack and bring back a short list of decisions that need you. keep the bot read-only at first. the useful handoff is evidence and a proposed next step, with prioritisation and delivery promises still owned by your team.
from morning blockers to a release note
| when | what a bot does |
|---|---|
| before standup | review unresolved issues in one project and pull out explicit blockers from their comments |
| after customer feedback | ask the bot to find related tickets, keeping the customer's request separate from the proposed solution |
| before planning | compare a small set of jira or linear issues and list missing acceptance criteria, without changing estimates |
| after shipping | draft a release update from the issues you confirm shipped, with uncertainty called out |
| end of day | read a slack channel digest for decisions and open questions in the channels you chose |
the routines
the standup evidence list
every weekday at 8:45, read open issues and comments in the linear project i named. show only explicit blockers, unanswered questions and work needing a decision from me. link each item to its issue and say who last raised it. don't infer that quiet work is blocked. don't change status, priority, assignees or estimates.use linear reads; swap in one named jira project if that is your team's source
the feedback trail
every tuesday at 2, read the past week's messages in the slack feedback channel i chose. group requests by the problem described. search the jira project i named for related issues. give me the message reference and issue reference together, distinguishing a possible match from a confirmed one. propose questions for discovery; create nothing.jira issue searches and slack history are enough; no ticket-writing access needed
the release update draft
every friday at noon, read the release issue list i supplied and its latest comments. draft a short update here: confirmed shipped work, unresolved rollout questions and what i need to verify. don't treat a closed ticket as proof a customer can use the feature. don't post the update or edit the roadmap.give the bot the exact release scope so it does not combine unrelated teams' work
choose one home for each fact
tell the bot which tracker owns delivery status and which slack channels carry discussion. a copied ticket in another system should not become a second vote that the same work shipped. name project boundaries in the routine, not just a broad keyword.
linear and jira both let a read-only bot search issues and read their comments. you do not need to connect both if your team uses only one. a smaller source set makes a missing or contradictory detail easier to spot.
ask for the disagreement, not a smoother story
a useful product brief preserves conflict. if a ticket says one thing and a later message says another, ask the bot to show both references and leave the conclusion open. include the date of the evidence so an old decision does not sound current.
leave priority calls to the people who understand the tradeoffs. a list of repeated requests can inform discovery, but repetition alone is not a reason to move an item ahead of reliability, accessibility or an existing commitment.
read first, publish deliberately
set tracker and slack permissions to read-only in Connected apps for this pack. the bot can report in its own chat without editing tickets or announcing dates. if you later allow tracker changes, add "ask me before you change anything" to the instructions. those changes are not covered by the sending prompt.
sending a slack message asks first unless you choose a daily sending amount. one exception: a bot on Full access with no sending rule of its own may send up to 25 a day without asking, so set Ask every time for it.7 company work also needs a work license; the app's free personal use does not cover running product work. see account and billing.
questions
what would a product manager actually schedule an ai to do?
collect evidence for standup, connect feedback to existing issues and prepare a release update for review. start with work whose result you can check against a handful of source links.
can it work with jira instead of linear?
yes. a read-only bot can search jira issues and read their comments. name the project and rewrite the routine around your jira scope; connecting linear is optional.
will it update tickets while summarising them?
with read-only access it cannot. keep that permission while trying the pack. granting write access later is a separate choice, so give explicit instructions about which changes need your review.
sources
- dusk bot docs: connected apps and permissions
- dusk bot docs: routines
- dusk bot docs: account and billing
- composio: linear actions, checked 12 november 2026
- composio: jira actions, checked 12 november 2026
- composio: slack actions, checked 12 november 2026
- dusk bot docs: permissions and sends
read next
a jira assistant that explains what changedread the issue history before the verdict.
a linear bot that spots missing context before planningbetter issue context before the planning call.
an end-of-day digest of the slack channels you chooseread the decisions, then follow the thread.
claude connectors: connect the job you actually needpick a task before you connect an app.
ai agents for founders and solopreneursa starter team, and the routines to copy.
a github bot that reads the work before you weigh inissues and pull requests, with context.
drafted with ai help from the dusk bot docs and the sources listed on this page, then checked against them. if something here is wrong or out of date, write to [email protected].