Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.onsomble.ai/llms.txt

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

Knowledge content is the first of the two workflow types in Onsomble. It lets you publish curated, authoritative information about your business — product details, pricing rules, policies, FAQs, anything a customer might ask — so that AI assistants can answer from your source of truth instead of guessing from whatever they scraped months ago. It’s the lightest-lift workflow to set up and often the first one a business activates after connecting their website.

​
Why curated knowledge matters

AI assistants build their picture of a business from public web content — websites, press, directories, reviews, anything crawlable. This picture is often:
  • Out of date — pricing from two years ago, product lines that have been discontinued, staff who no longer work there
  • Incomplete — the homepage has been scraped, but specific service pages haven’t, so the answer is shallow
  • Approximate — the assistant is stitching together fragments from different sources and filling gaps with reasonable-sounding guesses
Publishing knowledge content gives assistants a direct, authoritative source to draw from. When a customer asks a question your knowledge covers, the answer comes from what you said — not what a model inferred.

​
What to publish as knowledge

Start with the information customers most commonly ask about. Good candidates:
  • Products and services — what you offer, who each offering is for, what’s included, what isn’t
  • Pricing — rates, packages, discounts, eligibility rules
  • Process — how onboarding works, what the customer journey looks like, turnaround times
  • Policies — returns, cancellations, guarantees, compliance
  • FAQs — the questions your team actually hears over and over
  • Who to contact for what — sales, support, partnerships, billing
You don’t need to publish everything at once. A focused knowledge base of a dozen well-written topics usually outperforms a sprawling one of a hundred half-finished entries.

​
Adding knowledge

Knowledge content is added through the Onsomble dashboard. For each topic:
1

Create a knowledge entry

Give the entry a short, descriptive title that matches how customers would ask about the topic. “How pricing works” is better than “Pricing overview doc v2.”
2

Write the content

Keep it direct. Lead with the answer to the most likely customer question, then add supporting detail. AI assistants are particularly good at extracting information from content that’s structured the way humans would naturally ask for it.
3

Tag and categorise

Categories help both you (when managing a growing knowledge base) and assistants (when matching a customer question to the right entry). Use a consistent set of categories — products, pricing, process, policy, contact — rather than inventing new ones for every entry.
4

Save as draft or publish

Drafts aren’t visible to AI assistants. Publish when you’re ready for the entry to be live.

​
Structuring knowledge for AI retrieval

How you write knowledge content matters more than how much you write. A few principles that consistently produce better results:

​
Write in plain language

AI assistants relay the content to customers in something close to the original phrasing. Industry jargon, marketing copy, and overly formal prose all get passed through. Write the way you’d want the assistant to describe the business.

​
Lead with the answer

Put the direct answer in the first sentence or two. The context, exceptions, and caveats follow. A customer asking “what’s your return policy?” should hear “You can return items within 30 days of purchase” — not three paragraphs of legal framing before the actual number.

​
Use explicit structure

Clear headings, bullet points, and numbered lists all help. They give AI assistants obvious anchors to extract from, especially when the customer’s question maps to one part of a longer entry.

​
One topic per entry

Don’t combine “Pricing” and “Process” into a single sprawling entry. A customer asking about one shouldn’t have to sift through the other. Separate entries also make updating easier when one topic changes.

​
Keep it current

Inaccurate knowledge is worse than no knowledge. Set a rhythm for reviewing entries — every quarter is a reasonable default for most businesses. Pricing changes, staff changes, and service changes should trigger immediate updates.

​
Turning on knowledge retrieval

Once you have entries published, knowledge retrieval needs to be enabled for the Site so AI assistants can actually query it. This is a single toggle in the Site’s workflow settings. Once on, AI assistants that visit the connected website can discover and use your knowledge entries.
Enable retrieval only once you have enough published entries for it to be useful. A knowledge base with three entries tends to produce worse answers than no knowledge base at all, because the assistant will try to use what’s there even when it’s a poor match.

​
Keeping knowledge fresh

Businesses change. Pricing moves. Products launch. Policies get updated. Knowledge content needs regular maintenance to stay useful. A few small habits help:
  • Tie updates to source-of-truth changes. When pricing changes on the website, update the pricing entry at the same time. When a product line is retired, retire the corresponding entry.
  • Watch usage metrics. If an entry is being queried frequently, that’s a signal to keep it especially current — customers are asking about it.
  • Review periodically. Even entries that aren’t being queried often should be reviewed quarterly. Stale content is easy to forget about.

​
What’s next

Action workflows

Go beyond answering questions — let customers complete structured processes through their AI assistant.

Managing workflows

See which knowledge entries and workflows are live, what’s being used, and what to retire.