How To Work With Dobbin In Slack: Best Practices For Prompting
Dobbin’s Company AI vs Conventional AI Tools
A good prompt does not mean a long prompt. Generic AI tools require you to explain who you are and what matters to you. Dobbin already knows your company context, so you only need to provide exactly what you need from the answer.
How To Prompt Dobbin
1. Write a strong initial question/prompt
Stick to the core formula of context plus action to get a quick result without overthinking advanced rules for tone or formatting.
Tell Dobbin exactly what to do and state clearly what you need the final deliverable to look like.
Provide the required context by attaching a file, pasting your notes, or referencing specific starting information.
Basic prompt examples:
"Summarize [this attached file] into a short list of bullet points."
"Turn [these pasted notes] into a draft email to [audience]."
"Pull the next steps from [this thread] and list who is doing what."
"Clean up [this attached draft] to make it easier to read."
"Read [this document] and tell me the three main takeaways."
"Take [this transcript] and pull out the questions the customer asked."
2. See Dobbin’s first response as a first draft
Recognize that all AI — not just Dobbin — is trained to reach a conclusion as quickly as possible. That's how the models work.
A long answer is not the same as a complete one. Treat the first response as an opening move to react to, not a finished product to accept as-is.
3. Provide follow-up direction to Dobbin
Push back on the first draft to give direct feedback on what to change.
Volley back and forth to guide the system toward the exact answer you need.
Follow up examples:
"It sounds a bit too [current tone, e.g., formal]. Rewrite it to sound more [desired tone, e.g., conversational] for [target audience], and weave in the updates from this [attached document]."
"You captured the main points well, but we missed the [recent offline decision / late-breaking data]. Add a final section stating that [specific decision / metric], pulling from this [attached file]."
"Now that we see the overview, [Option A] is the clear winner. Regenerate the brief focusing only on [Option A], and highlight how it impacts [specific team / goal] based on this [attached report]."
"Now, adjust the [timeline / scope] to account for [specific change, e.g., a two-week delay], and add a specific track for [new requirement] using the guidelines in this [attached deck]."
“We need to address [specific concern, e.g., security] upfront. Swap out the [old section] with the new [rules / policies] in this [attached PDF], and weave in [specific quote / stat] to anchor the opening."
4. Identify and add helpful additional context
Ask Dobbin what additional context it needs — information it doesn't already know, meaning anything siloed within your organization — and add it to the conversation.
Here are a few prompt examples users can drop into the middle of a workflow to uncover missing context.
"Before you write this, what specific internal files or past examples do you need me to provide?"
"What data or context are you missing to make this recommendation actually actionable?"
"Are there any specific reports, transcripts, or past decisions you need to see before we finalize this?"
"To pressure-test this assumption, what internal information do you need me to give you?"
5. Collaborate with your team
Tag a team member directly in the thread so they're looped into what you've been working on, sharing your results to make it part of the collaborative work.
Managing upwards: "@[Name], Dobbin pressure-tested the [Q3 strategy]. The risk analysis is in this thread. Let me know if you agree with the recommendation to proceed with [Option A]."
Managing downwards: "@[Name], Dobbin pulled together a first draft of the [onboarding schedule] based on our notes. Take a look at the timeline below and fill in the missing [training modules]."
Working laterally: "@[Name], Dobbin summarized the action items from our [client discovery call]. Review the list below and flag if it missed any [feature requests] you caught."
Working cross-functionally: "@[Name], Dobbin turned the [sales objections] into a draft for the [new landing page]. Review the copy below to ensure it matches your [campaign guidelines] before we publish."
Expanded Prompt Library
Once you have the hang of the basics, you can start shaping the result of your work with Dobbin.
Intermediate Prompts: Directing the Deliverable
These prompts go beyond simple commands by adding specific constraints—like who the audience is, what tone to use, the exact length, or specific criteria to compare. By setting these boundaries upfront, you skip the generic first draft and start much closer to the final deliverable.
"Compare [option A] and [option B] on [specific criteria]. Pull from [source] and flag anything that's outdated."
"Summarize the key decisions from [attached meeting transcript/Slack thread] into a [length/format] brief I can forward to [department heads]. Format the key decisions as bullet points and put the next steps in a table."
"Draft an [objection-handling script] for [the sales team]. It should cover [responses to X, Y, Z objections] and land at around [3 sentences per response]. Start from [rough call notes/won deal history]."
"Write a first draft of [an RFP response] in the voice of [past winning proposal]. Keep it to [standard length/format] and pull product details from [specific reference document]."
Advanced Prompts: Engaging the Thinking Partner
This is where Dobbin shifts from a drafting tool to a strategic advisor. Level Three prompts ask Dobbin to analyze, challenge, or synthesize information against your company's specific goals and Foundation. Use these when you need to spot blind spots, evaluate tradeoffs, or pressure-test a high-stakes decision.
"I am considering [Decision X]. Based on our strategic objective to [Objective Y], what are the three biggest risks I am missing? Play devil's advocate."
"Review this [strategy draft] against our [Brand Foundation/Core Values]. Flag any areas where the logic is weak, generic, or misaligned with how we make decisions."
"Synthesize these [conflicting reports/threads] and recommend a path forward that prioritizes [Specific Goal]. State the tradeoffs clearly."
"Look at [proposed process/new initiative]. What assumptions am I making here that could cause this to fail based on [historical data/past project]?"
"Review these [new initiative options] against our [current resource constraints]. Recommend the safest path forward and flag any fatal misalignments with our [long-term goals]."
How Dobbin Learns From Slack Conversations
Dobbin learns from both the body of your conversations and from what you attach to a message (i.e. a report, a PDF of a related presentation, reference materials). If Dobbin sees it as valuable to the company as a whole, judged by its relationship to the company Foundation, it appears as a Learning in Dobbin HQ. There, it can be promoted to permanence in the Knowledge Base if your HQ admin reviews and approves it.