Get more done with AI.
You already have a powerful AI assistant in Slack. This guide takes you from basic questions to building automated tools — one level at a time. No coding experience needed.
Start Talking
Get comfortable with the basics
Just open a DM with the Hermes bot and start typing. It understands natural language — talk to it like a colleague. You can ask follow-up questions, change direction, or paste in long documents for it to process.
Try these
"Summarize these meeting notes in 3 bullet points: [paste your notes]"
"Explain [concept] like I'm new to this"
"What's the difference between [A] and [B]? Give me a simple comparison."
"Help me draft a follow-up email after my meeting with [person] about [topic]"
"Translate this into plain English: [paste jargon-heavy text]"
Be specific, not vague
Too vague
"Help"
Much better
"Help me understand why our MRR dropped in March"
Try it now
Ask it to explain something you Googled today.
Ask Better
5 prompt patterns for structured outputs
The trick to getting great results is telling the AI exactly what you want and how you want it. Here are 5 templates you can reuse for almost anything.
A The Analyzer
Break anything down into risks, opportunities, and action items.
"Analyze this [document/report/data]. Tell me 3 things: the biggest risks, the best opportunities, and concrete action items I should take."
B The Formatter
Turn messy information into clean, organized output.
"Take this information and organize it into a [table / bullet list / timeline / spreadsheet format]: [paste your data]"
C The Comparator
Get a clear side-by-side comparison of anything.
"Compare [Option A] and [Option B]. Use a table with these columns: [criteria 1, criteria 2, criteria 3]. Add a recommendation at the end."
D The Drafter
Generate professional documents with the right tone.
"Draft a [email / report / memo / proposal] about [topic]. Tone: [professional / friendly / urgent]. Audience: [who will read this]. Keep it under [X words / 1 page]."
E The Reviewer
Get a second pair of eyes on anything you've written.
"Review this [document / email / proposal / contract]. Point out: gaps in logic, unclear sections, factual errors, and specific suggestions to improve it."
Try it now
Pick one of the 5 patterns, fill in the blanks with something real from your work, and send it to the bot.
Work with Data
Upload files and get answers from your data
You can attach files directly to your DM — CSVs, Excel spreadsheets, PDFs, Word documents. The AI will read them and do whatever you ask.
How: In your Slack DM with the bot, click the + button (or paperclip icon) to attach a file, then type your question along with it.
Examples by role
"Here's the cloud spend report. Break down costs by service, compare to last month, and flag anything that increased more than 20%."
"Pipeline CSV attached. Rank all accounts by deal size multiplied by close probability. Show the top 20 in a table."
"Attached is our Q1 P&L. Calculate the margin trend month over month and flag any anomalies that look unusual."
"Attached are support tickets from last week. Categorize them by issue type, find the most frequent complaint, and suggest root causes."
"Employee survey results attached. Group responses by department, highlight the 3 lowest satisfaction areas, and suggest improvements."
Pro tip
Always tell it the output format you want: "Show as a table", "Give me bullet points", "Create a chart description", "Export-ready CSV format".
Try it now
Upload a spreadsheet you worked with this week and ask it to summarize the key findings.
Automate
Build tools and schedule recurring jobs
This is where things get powerful. You can ask Codex to write scripts that process your data, and even set them up to run on a schedule. You describe what you want in plain English — it handles the code.
One-off tools
Ask it to build a specific tool for a task you need done once.
"Build a calculator that takes a loan amount, interest rate, and loan term, then outputs a full monthly amortization schedule as a table."
"Write a script that takes two Excel files — this month's expenses and last month's — and highlights every line item that changed by more than 10%."
"Create a tool that reads a folder of PDF invoices and extracts: vendor name, invoice date, total amount, and due date into a single summary table."
Scheduled jobs
Set up recurring tasks that run automatically and deliver results to you.
"Set up a job that runs every Monday at 9am. It should pull our new signups from last week and send a summary to me here with total count, top sources, and any flagged accounts."
"Create a scheduled job for the 1st of every month that compares our actual spend vs. budget for each department, and sends the variance report to the #finance channel."
"Set up a daily job that checks our support ticket queue, categorizes any new unresolved tickets by urgency, and sends me the high-priority ones every morning at 8am."
The automation pattern
Trigger
(daily, weekly, monthly)
Action
(pull, calculate, process)
Output
(DM, channel, file)
Start simple
Get a one-off version working first. Once it works the way you want, then ask to make it recurring. Trying to automate something that doesn't work manually yet is frustrating.
Try it now
Think of one task you do every week. Ask Codex to do it for you — start with a one-off version.
Power Moves
10 high-impact prompts for your role
Quick-reference cards for each team. Pick your role, copy the prompts that fit your day, and level up your workflow.
Daily use
"Break down our cloud spend by service for today. Which service consumed the most, and is it normal for a [day of week]?"
"Scan our billing data for anomalies. Flag any charges that are more than 2 standard deviations from the 30-day average."
"Show me the resource utilization for our top 5 most expensive services. Are we paying for capacity we're not using?"
"Compare this invoice to last month's. Line by line, flag any new charges, removed charges, or items that changed by more than 15%."
"Calculate our unit cost per [transaction/user/API call] based on this month's infrastructure spend and volume data."
Weekly wins
"Analyze our spending trends and recommend 3 specific cost optimization opportunities with estimated savings for each."
"Check our reserved instance / committed use coverage. How much are we spending on on-demand that could be covered by reservations?"
"Generate a spend-by-team report. Which team's costs are growing fastest? Is it justified by their usage metrics?"
Level up — automate these
"Set up a daily job that posts a cost summary to the #finops channel at 9am: total spend, top 3 services, and any anomaly alerts."
"Create a weekly job that generates right-sizing recommendations for our compute resources and sends them to me every Monday."