Aline's AI Chat is where you analyse, draft and redline contracts, run your team's saved ways of working, and ask questions across your whole repository. This guide covers what the chat can do today and the gotchas worth knowing before you rely on it.
🧭 Overview
With AI Chat in Aline, you can:
Analyse a contract and generate redlines
Run an AI Skill — your team's packaged way of handling a task — with a single /
Build a new skill by describing it, then publish it for everyone
Ask questions across your entire repository or a single report and get a counted answer with links to the source documents
Create reports, playbooks and workflow phases from the chat
Turn any chat answer into a document you can share
Saved Prompts Are Now AI Skills
If you saved prompts in the past, they are now skills. A skill is more than a stored prompt: it packages your team's way of handling a task — reviewing a vendor MSA, marking up an NDA — together with the templates and fallback language that go with it. It runs a sequence of steps rather than firing a single instruction.
Skills are grouped two ways:
Community skills — ready-made skills Aline shares with every customer.
My skills — the ones your organisation has built.
When you update a skill, everyone who runs it gets the new version. You do not have to tell your team to change anything.
Step 1: Run a Skill
Type a forward slash (/) in the chat — this lists every skill available to you.
Pick the skill you want.
Attach the document and press enter.
💡 The first time you use a skill, open it and read what it does. Skills written by someone else may make assumptions that do not match your matter.
💡 Adding your own instructions alongside a skill does not break it. You can type extra context after invoking the skill and it will be taken into account.
Step 2: Build a Skill
You can describe the skill you want and have the AI write it:
In the chat, type what you want — for example, build me a skill that analyses NDAs.
Tag any reference material with the @ symbol. To base a skill on a playbook, tag the playbook and ask it to list all of the rules of the playbook before building the skill.
Tell it how the skill should behave at run time — for example, have the skill ask for the context before each run.
Review the proposal, approve it, and click the skill card in the chat.
⚠️ A skill you build is not usable until you publish it. This is the most common reason a skill you just created does not appear when you type /. Publish it from the skill card, or go to Settings → Skills under Agent Personalization and click Publish.
💡 Before sending a long instruction, click improve prompt. It rewrites your request into something more robust, and — more usefully — it reads back what the AI understood you to be asking for, so you catch a misunderstanding before it costs you a run.
What a Skill Cannot Do
A skill cannot tag a playbook. If you want a skill to apply playbook rules, the rules have to be written into the skill itself — ask the chat to list the playbook's rules and build them in.
A skill has no document you have not given it. If the skill needs to compare against an original, build it to ask the user to upload that file at run time.
There is no voice input — skills are written and run as text.
Skill or Playbook?
Use a playbook for your standard, repeatable position-by-position review. It is structured and predictable, and it is the right tool for first-pass review of a known agreement type. See Building AI Playbooks.
Use a skill for bespoke work — an unusual agreement, a one-off analysis, or a task that is a sequence of steps rather than a list of positions. Skills have far more freedom.
📌 They are not mutually exclusive, and some teams use skills without playbooks at all. A skill is also the better tool for reviewing returned redlines, because a playbook re-examines the whole document rather than focusing on what changed.
Step 3: Ask Questions Across Your Contracts
The chat can answer questions across your whole repository or across a single report, and it returns a counted answer with links to the documents behind it.
For example: how many contracts do I have that expire after October 31, 2026? You get the number, and you can open the underlying documents to check it.
If some contracts have not been analysed yet, Aline tells you and fills the gap.
You can open the raw data behind any answer.
When you want the answer scoped to one report, name the report in your prompt.
Step 4: Create Reports, Playbooks and Workflows from the Chat
You can build things from the chat, not just ask about them:
Reports — for example, create a report for all my NDAs and extract provisions around trade secrets. Aline proposes a plan, chooses the properties and the input and output types, and runs once you approve. See Building AI Reports in Aline.
Playbooks — tag a template or a set of redlined agreements and ask for a playbook. See Building AI Playbooks.
Template phases and workflows — describe the process you want and the chat will build it.
⚠️ The chat can create a report but it cannot edit an existing one. To change a report, open it and use Filters → Add filter, then save your changes. Asking the chat to modify a report you already have will produce a new one instead.
💡 Any chat answer can be turned into a document, which is the quickest way to get an analysis out of the chat and into something you can share or store.
Step 5: Run a Prompt on a Document
You can bring a document into the chat in several ways:
Drag and drop the file straight into the chat.
Paste a URL — a link to a public document such as a set of terms of service — and Aline retrieves and analyses it.
Insert files as context chips inside the prompt, so the AI knows which document each instruction refers to.
⚠️ Aline AI does not read spreadsheets. Export or print the sheet to PDF first. See File Format Requirements & PDF-to-Word Conversion.
Step 6: Review the Analysis
Once you run a prompt, Aline analyses the document and returns a structured breakdown — clause summaries, risk assessments, and recommendations, often as a green / amber / red ranking with negotiation guidance.
💡 A long response can be stopped mid-run with the stop button, so you are not waiting on an answer you have already decided to rephrase.
Step 7: Generate Redlines
You can go beyond analysis and ask the AI to apply redlines directly. It reprocesses the document and inserts the suggested edits as tracked changes.
⚠️ Redlines from the chat behave differently from playbook suggestions. Chat redlines are already written into the document — approving one accepts a change that is already there. A playbook suggestion, by contrast, is a proposal you choose to apply.
⚠️ On a PDF you get flags, reasoning and comments but no applied redlines, because Aline cannot write tracked changes into a PDF. Convert to Word first — see File Format Requirements & PDF-to-Word Conversion.
💡 Your editing mode — Editing, Suggesting or Commenting — now stays as you left it when you switch tabs.
Chat History and Models
Your past conversations are under the clock icon in the chat. Prompts are not saved automatically — only the ones you deliberately save are kept.
Aline runs on several frontier models behind the scenes, and you can see which model was used for which action in the usage tab. Every AI action consumes tokens — see Understanding Token Usage in Aline.
💡 Best Practices
Publish a skill as soon as you build it, or nobody — including you — can run it.
Open an unfamiliar skill and read it before you run it on a live matter.
Use improve prompt on anything complicated, and read what it says the AI understood.
Playbook for the standard first pass, skill for the bespoke work and for returned redlines.
Name the report in your prompt when you want an answer scoped to it.
Send Word when you want redlines; PDF is fine when you only need the AI to read.
📚 Related Articles
❓ Need Help?
Click the Support button in Aline to reach our team.