Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.kowalah.com/llms.txt

Use this file to discover all available pages before exploring further.

Once installed, the Kowalah Agent is available in Slack, Microsoft Teams, and Google Chat. This guide covers how to interact with it and how to get the most out of it. The Agent is at its best when you bring it real work you’re doing right now. “Could AI help with this?” — asked while you’re actually in the middle of the task — is the question it’s built to answer.

​
Two ways to use the Agent

​
@mention in a conversation

Mention @Kowalah in any channel, thread, or group chat followed by your request: 
@Kowalah We're scoping the new customer onboarding process —
how could we use AI in this?
The Agent responds in the same thread. You can send follow-up messages without mentioning @Kowalah again — it stays active in the thread. This is great for bringing AI into a conversation your team is already having: a planning meeting, a working session, a problem someone just raised.

​
Direct message

You can also have a private 1:1 conversation with Kowalah. Find the Kowalah app in your platform’s app directory and open a direct message — no @mention needed. Direct messages are useful when you want to:
  • Work through an AI idea before sharing with the team
  • Have a longer, multi-step conversation about a workflow design
  • Explore a use case quietly first

​
Ask “could AI help with this?”

This is the headline use case. You’re doing some work — month-end close, a board pack, a hiring round, a customer review — and the thought lands: “could AI help us with this?” Take it to the thread. 
@Kowalah We do month-end variance analysis every month. How
could AI help us write the commentary faster?

@Kowalah I'm prepping for board reporting tomorrow. How would
you use AI to draft the narrative around the financials?

@Kowalah We've got 30 customer interview transcripts to
synthesise into themes. What's a good AI approach?

@Kowalah We get 400 support tickets a day. How would you
use AI to triage and route them?

@Kowalah Help me think about how to use AI to evaluate
RFP responses faster.
The Agent draws on patterns from across Kowalah’s customer base — what’s worked elsewhere on the same problem, where the friction usually is, which approaches generally fail — and gives you a concrete answer. Then it checks the accelerator library, and if needed scopes the work as an Opportunity.
Be specific about your work. “How could AI help with month-end?” is fine, but “how could AI help us speed up our month-end variance commentary across 12 entities?” gets you a sharper answer because the Agent can pattern-match to the real shape of the problem.

​
Design an AI workflow

When you’ve got a recurring task — or a one-off process you wish someone could just do — ask the Agent to design an AI workflow for it. It maps your steps, decides what can run in-thread vs. what needs proper build, and either does the work live or scopes it for Kowalah. 
@Kowalah Help me design an AI workflow for our weekly KPI
report. It takes me three hours every Friday.

@Kowalah Before each customer QBR I summarise the last 90
days of activity across emails, support tickets, and product
usage. Can we automate this with AI?

@Kowalah I write 15 job descriptions a quarter for new sales
roles. Help me build an AI workflow that drafts them properly.

@Kowalah We need to draft weekly competitive intel for the
sales team. Help me design the AI workflow.
For analysis, synthesis, drafting, and pulling things together from your workspace, the Agent often does the work right there in the thread. For workflows that need ongoing tracking or specialist build, it scopes the work and raises it as an Opportunity for the Kowalah team.

​
Find where AI fits in your role

When you don’t know where AI fits in your role, the Agent walks you through what’s on your plate and surfaces 2-3 high-value AI use cases. Every recommendation ties back to your own work. 
@Kowalah Where could AI make the biggest difference for me?
I run finance at a 1,500-person professional services firm.

@Kowalah I'm a regional sales director. What AI use cases
should I try first?

@Kowalah I'm Head of People at a manufacturer. Help me figure
out which AI workflows would be worth building first.
Expect 3-4 exchanges before the Agent gets to recommendations — that’s how it makes the suggestions specific to your role and work rather than a generic feature dump.

​
Find an accelerator

Before raising new work, check whether there’s an off-the-shelf accelerator (a prompt, GPT, training, or workflow template) that already solves the problem. The Agent searches Kowalah’s global library plus your organization’s private library. 
@Kowalah Is there an accelerator for writing job descriptions?

@Kowalah Find me a prompt for quarterly board reporting

@Kowalah Show me HR accelerators tagged for onboarding
Results often include a try-it-now prompt the Agent can run with you in the same chat.

​
Raise an Opportunity with the Kowalah team

Opportunities are the single inbound channel for anything you want Kowalah to take forward — an AI workflow to build, a question to scope, a piece of work that needs specialist hands. You don’t need to decide whether it should become an Expert Request, a Deliverable, or a new Project — Kowalah’s team triages from there. 
@Kowalah Capture an opportunity for automating our weekly
KPI report with AI — we just designed the workflow together

@Kowalah I'd like Kowalah's help building a Claude skill that
summarises customer interviews — please raise it as an opportunity

@Kowalah Help me scope an AI workflow for invoice processing,
then submit it as an opportunity
Before submitting, the Agent will typically:
  1. Check the accelerator library in case something already solves the problem
  2. Make sure your description has enough context — goal, what good looks like, anything Kowalah should know
  3. Ask which organization the opportunity is for (if you belong to more than one)
  4. Ask which discovery it relates to, if you have any pending invites
The Agent always asks before raising anything, and describes exactly what it will write.

​
Look up status across your portfolio

The Agent can also read what’s going on in your Kowalah workspace when you want a quick status check: 
@Kowalah List all our active projects and their health status

@Kowalah Show me the deliverables in our Finance Automation project

@Kowalah What expert requests are currently in progress?

@Kowalah Do I have any pending discoveries to respond to?

@Kowalah What's waiting on me right now?

​
What your role changes

Your organization role affects what the Agent shows you, not what you can do.
RoleWorkspace viewOpportunities visibleProjects, deliverables, ERs visibleCan capture opportunities
AdminFull org rollupEvery opportunity in the orgAllYes
Core teamFull org rollupEvery opportunity in the orgAllYes
MemberPersonal rollupOnly the ones you raisedOnly the ones you’re a stakeholder onYes
Tasks, risks, and decisions are filtered to client-visible items. Internal Kowalah fields (team assignments, contract values, payment details) are never exposed. Pending discoveries are personal — you always see your own invites, regardless of role.
If the Agent says it can’t find something you know exists (a specific Expert Request number, for example), it usually means you’re not on its stakeholder list — not that it’s been deleted. Ask an admin or core team member, or get added as a stakeholder in the platform.

​
Tips for better results

​
Bring real work

The Agent is at its best when the question is grounded in something you’re actually doing — “we do X, could AI help?” — not abstract questions about AI in general.

​
Be specific about the task

“How could AI help with month-end?” is fine. “How could AI help us speed up our month-end variance commentary across 12 entities, where the constraint is that we need an auditable trail?” gets a sharper answer.

​
Use threads

Start a thread when you have a multi-step conversation. The Agent remembers thread context, so you can iterate on a workflow design across several messages without repeating yourself.

​
Let it explore

For complex questions, the Agent may take multiple steps to gather information before responding. Give it a moment — it’s reading through your projects, deliverables, and accelerators to give you an accurate answer.

​
Ask it to design, not just describe

Don’t stop at “could AI help with this?” — once you’ve got a direction, ask the Agent to design the workflow. That’s where the real value compounds.

​
Response times

ScenarioTypical time
First message in a new thread~20 seconds
Follow-up in the same thread3-5 seconds
Simple lookups (list projects)5-10 seconds
Complex analysis (cross-project summary)15-30 seconds
The Agent shows a typing indicator while it works on your request.

​
Troubleshooting

The Agent can’t find a Kowalah account associated with your chat identity. Follow the link it provides to connect your accounts.

​
Agent doesn’t respond

Make sure you’re using @Kowalah (with the @ mention, not just the text). The Agent only activates when properly mentioned.

​
”I can’t find that”

The item may exist but be outside your visibility. Members only see opportunities they raised and entities they’re a stakeholder on. Ask an admin or core team member to confirm, or get added as a stakeholder in the platform.

Roles and permissions

See what each role can do

Agent overview

Back to Agent overview