Write the business casefrom evidence you already gathered
Before anyone builds anything, somebody has to justify it in writing: the problem, the options, what it costs, what it returns, what could go wrong. The evidence exists, scattered across a Confluence page, a vendor email, a screenshot from finance and four tickets. Khint gathers it into one work session and lets a saved AI action write the case from it, in your house structure, every time.
The hard part is not the writing,
it is the assembling
A business case fails for boring reasons. The cost line is three weeks old because it lived in an email. The option someone rejected in a workshop reappears in the document, because the workshop notes were never opened again. The benefit is asserted without the ticket volume that would have proved it. None of that is a writing problem: it is an assembly problem, and it is why the document gets written on the last evening, from memory. Khint attacks the assembly. Every pull, capture and note lands in one work session that keeps a compacted working state, and the drafting action reads that state instead of your recollection of it. The evidence and the document stop being two different piles.
One business case, step by step
From scattered evidence to a document a committee can decide on.
Open a session named after the decision
Start a Memorysession called “Payments provider, business case” and leave it active. From here on, everything you pull, capture or note is logged into it, and Khint keeps a compacted working state of the whole file: objective, open threads, artifacts, decisions. That working state is what your draft will be written from, so the gathering phase and the writing phase stop being two separate jobs.
Pull the evidence out of the tools it is stuck in
Connect your stack once from the MCP page and the palette gains read rows. Search and Pull issue bring the tickets where the pain keeps resurfacing. Pull page brings the discovery notes out of Confluence. Pull thread brings the vendor's answer out of Gmail or Slack. Each one lands as an entry in the session instead of in a scratch file you will lose.
Capture the numbers that only exist on a screen
Finance sent the cost model as a spreadsheet screenshot, and the licence tiers live in a vendor portal that will not let you copy. Pick Extract text in the palette, drag a region, and the text is on your clipboard and in the session. How Capture works →
Write the business case Action once
In the Action editor, describe it in one line: problem, options considered, recommended option, benefits, costs, risks, and the decision you are asking for. Press Generate for a first prompt, then Improve to tighten it. Set its input source to Active session so the run reads your gathered working state rather than a selection, and give it a keyboard shortcut.
Run it, then argue with the draft
One shortcut and the case comes back in your house structure, written from the evidence in the session rather than from a blank page. It will be wrong somewhere: a benefit overstated, an option you already ruled out. That is the useful part. You are now editing a draft with named sections instead of staring at a heading called Options considered at 22:40.
Turn the trade-off into a slide, then publish
On macOS, switch on the PowerPoint companion from the MCP page, select the options paragraph in your deck and pick Transform to diagram: Khint picks a framework for the content, and you can force another one from the selector, then save it as native editable shapes. Diagram generation → When the case is agreed, publish it with the Confluence Create page row, which asks for the space and the parent page and shows you a preview before anything is written.
Every section traces back to something
What a case assembled in one session looks like, and where each part came from.
Nothing in that column on the right was retyped. It was pulled or captured into the session while you were doing the research, which is also why the second draft is cheap: the session is still there, so a rewrite after the steering committee pushes back does not mean gathering it all again. Once the case is approved, the same session feeds the project charter and the first steering committee deck.
What Khint will not do for you
It will not produce a number you did not give it. An Action reads the text it was handed plus, when you opt in, the compacted context of your session: there is no browsing, no market data, no benchmark pulled from somewhere. If the licence cost never entered your session, the model has nothing to base a figure on, which is exactly why the honest move is to tell your saved prompt to mark missing figures as TBD and list its assumptions. It also will not make the decision. It produces a readable, comparable case; the committee still decides, and you still sign your name under the recommendation. If what you need is a scoring pass across suppliers rather than a case for one investment, that is a different job: vendor evaluation with AI.
Common questions
Does the AI invent the numbers in the business case?
No, and you should write your prompt so it cannot try. An Action only ever sees the text you gave it plus, if you opted in, the compacted context of your active Memory session. If a cost or a benefit figure is not in what you gathered, it is not in the model's input either. Add a line to your saved prompt telling it to mark any missing figure as TBD and to list the assumptions it had to make, and the draft comes back with the gaps flagged instead of quietly filled. Khint writes the structure and the wording. The arithmetic stays yours.
My evidence is scattered across five tools. Do I have to paste it all in?
No, you pull it into one place first. Connect Atlassian, Linear, Notion, Slack or Gmail from the MCP page and the palette gains read rows: Pull issue, Search, Active sprint, Pull page, Pull thread, Pull channel. Each pull lands as an entry in your active Memory session, which keeps a compacted working state of everything you collected. Numbers that only exist as a screenshot of somebody's spreadsheet go in through Capture, which extracts the text and logs it in the same session.
Can I reuse the same business case structure next quarter?
That is the point of saving it as an Action. The structure lives in the prompt, not in your memory of how you did it last time, so every case your team submits arrives in the same order with the same sections. Bundle it into a Pack with the other decision documents and, if you own a team, Share the pack: teammates see it as a read-only Team Pack and clone their own copy in one click. Chain the draft step and the publish step into a Workflow and the whole thing becomes one shortcut.
Where do the cost figures and vendor names actually go?
An action sends the text it was given to Khint's proxy for the model call. What does not travel: your Memory sessions and their entries live in a local database on your machine and are never synced, and the input and output bodies in your local history stay on the device too. If a document names people, switch on the per-action Redact PII option and emails, phone numbers, SSNs and IBANs are replaced with placeholders before the text is forwarded.
Stop writing it from memory
Free with 10 AI actions and 5 captures per day. No credit card. Gather the evidence once, and let your own saved action write the case.