Fin AI Agent and Copilot can consume all your support content to immediately offer your customers and teammates accurate and trustworthy answers.

Use this article to optimize your help center content for Fin AI Agent and Copilot. You'll learn how to identify content gaps, update existing articles, and apply the 14-factor content readiness framework — so Fin can retrieve accurate answers and customers can find what they need. This article is for workspace admins managing a knowledge base.

The more simple, straightforward, and comprehensive your articles are, the easier it will be for AI (and humans!) to consume them. If it’s confusing for a human to read, it’ll be confusing for an AI agent or copilot as well. Focus on:

Here are some tips to help you prioritize which content to optimize for Fin first:

The following best practices use public articles as an example, but apply to all content enabled for AI — including internal articles, snippets, and synced or external sources. They're organized around a 14-factor content readiness framework: the set of criteria that determines whether a piece of content is ready to be used by Fin and Copilot to answer real customer questions.

If it’s confusing or ambiguous for a human to read, it’s also going to be ambiguous for AI! (And that’s when there’s a chance it’ll give the wrong answer, or make an incorrect inference, or just not answer something that otherwise should be answerable).

The table below shows how restating the subject removes ambiguity for both customers and Fin:

You don’t need to do a question-answer pair, but write as if you’re doing a radio interview and you don’t want to be quoted out of context.

The table below shows how restating the question in the answer keeps it self-contained when retrieved out of context:

Use headers (H1, H2, H3) to divide your content into focused, single-topic sections. Each section should cover one thing — if a section mixes multiple subjects, break it up. This directly controls what Fin retrieves and presents to a customer: a section covering three different things will generate unfocused answers. Also include some of the header text in the paragraph below it, in case Fin or Copilot fails to capture all headings in the HTML.

The table below shows how using separate headers with context under each one keeps sections self-contained for Fin retrieval:

Multi-step processes must use numbered lists. Lists of options or items must use bullet points. If a sentence says 'the following:' it must be followed by an actual list, not a paragraph. AI performs better when long, detailed paragraphs are formatted using clear structure instead.

The table below shows how converting a dense paragraph into a structured list helps Fin extract specific points:

All numbers, thresholds, limits, durations, and quantities must be stated precisely. 'A few minutes,' 'some requests,' 'shortly' — none of these pass. Use exact values where known. If you want Fin to handle questions that may involve calculation, include worked examples that show clearly how the sum is calculated.

The table below shows how stating exact timeframes and breaking them down eliminates the need for a customer or Fin to calculate durations:

Every piece of content should state who it's for and what access or permissions are required. Don't assume the reader knows their own role or plan. 'Available on all plans' or 'requires Workspace Owner permissions' gives readers the context to know whether this applies to them before they follow five steps to discover it doesn't. The same principle applies to contact details — if certain numbers or addresses are specific to a particular situation, clearly delineate them.

The table below shows how labeling contact details by audience makes each entry self-descriptive and unambiguous:

An AI agent drives the most value when it resolves high volume, repetitive questions. So if you have small pieces of information that don’t require a full article or have no other obvious home, you should include them as a list of FAQs (frequently asked questions) in one public / internal article or create snippets. Fin will still be able to find them and serve them to customers.

The table below shows a well-structured FAQ entry that's complete enough for Fin to retrieve and use in isolation:

When setting up FAQ articles it is strongly recommended to use headers (H1, H2, H3) when structuring your FAQs.

Fin can send relevant images/GIFs from your support content in it's AI answers (up to 3).

However, if you've included multimedia in your support content, it's still important to also include step-by-step text instructions above or below any image or video which guides the user through the process.

Bonus: Including step-by-step text instructions makes your content more accessible to users with low vision, while also catering for multi-mode learners:

The table below shows how pairing an image with step-by-step text ensures Fin can answer process questions even if it doesn't surface the image:

The opening paragraph of every article must state what the reader will accomplish, not just what the article is about. Fin uses the opening to understand what an article is for — a weak opening degrades how well it surfaces for relevant queries.

The table below shows the difference between a topic-description opening and a jobs-to-be-done opening:

Step-by-step instructions must be complete end-to-end. That means including what happens after the final step: confirmation dialogs, what the customer should expect to see, and what to do if they don't see it. An incomplete instruction leaves the customer stranded — and gives Fin nothing to draw on when a customer asks 'did it work?'

The table below shows how completing an instruction — including the confirmation step — removes ambiguity about whether an action succeeded:

If a feature has known limitations, gaps, or edge cases, document them specifically — not vaguely. Which configurations are affected, what the limitation is, and what the customer should do instead. This is the factor that most directly prevents Fin from giving a misleading answer. If a limitation isn't documented, Fin doesn't know to flag it.

The table below shows how documenting a specific limitation with a workaround prevents Fin from giving a misleadingly incomplete answer:

Abbreviations and product-specific terms must be defined on first use. Don't assume the reader knows what CSV, GDPR, or 2FA mean. The first time an acronym appears, spell it out in full — then use the short form for the rest of the article.

The table below shows how defining an acronym on first use makes content accessible to all readers:

Key product and feature names should appear throughout an article — not just in the opening paragraph. When Fin retrieves a section mid-article, it needs enough context to know what's being discussed. A section that only says 'click the button in the top right' with no mention of the product or feature it refers to is effectively invisible to retrieval.

The table below shows how repeating the feature name in each section gives Fin the context it needs to retrieve and use that section accurately:

Tables and standalone blocks must include an introductory sentence that explains what they cover and where they apply. Fin can retrieve a table without the heading above it, so the table itself needs to make sense in isolation. A table that jumps straight into rows with no context fails this — Fin has no way to know what the table is about.

The table below shows how adding a single introductory sentence makes a standalone table self-explanatory for both customers and Fin:

Before publishing any piece of content, check it against these 14 factors. Content that passes all of them will be easy for customers to follow and easy for Fin to retrieve, parse, and use to give accurate, complete answers.