SaaS
An AI chatbot for SaaS that reads your docs properly
Support volume in SaaS is dominated by a small number of questions asked endlessly: how do I reset this, why was I charged, does the plan include that, how do I invite a colleague. The answers are almost always in your documentation, and almost nobody reads documentation. Paperchat trains on your docs, changelog, help centre and internal notes and answers those questions in one message, in the product where the user already is. Connect Stripe and it can check a specific invoice or subscription during the conversation. Anything it cannot resolve goes to your support team with the full history.
- Trained on your documentation
- Billing lookups with Stripe
- Escalation with full context
Why did my invoice go up this month?
You added two seats on 14 March, which was prorated for the remainder of the cycle.
Invoice INV-2043
Can I remove one before next month?
Yes, seat changes apply from the next cycle. I can pass this to support to action it now.
Your documentation, finally used
Most SaaS teams have decent docs and a support queue full of questions those docs already answer. The gap is not content, it is retrieval: users will not search a knowledge base, but they will ask a question in a chat window.
Paperchat trains on your docs site, help centre, changelog, PDFs and pasted internal notes, so the answer that already exists gets delivered in the moment somebody needs it. When you ship a change, retrain that source and the answers update.
Docs and help centre
Crawled from your address or sitemap
Changelog and release notes
So answers reflect the current build
Internal runbooks
Pasted text your users never see
Connected systems
Live account and billing detail
Billing questions answered with the actual invoice
Billing generates a disproportionate share of tickets and almost all of them need one specific fact: what was charged, when, and why. Documentation cannot answer that, which is why these tickets always reach a human.
With Stripe connected, the agent can look up the invoice or subscription in the conversation and explain the charge, including proration and plan changes. What remains for your team is the genuinely contested case rather than the routine explanation.
- 1The user asks why they were charged
- 2The agent looks the invoice up in Stripe
- 3It explains the change in plain language
- 4Anything that needs a refund goes to your team
Put this on your site today
Train a chatbot on your own content, connect what it should be able to do, and have it answering in the time it takes to make coffee. The free plan needs no card.
Onboarding questions in the first session
Activation depends on the first fifteen minutes. A new user who cannot work out how to connect their data or invite a colleague usually does not come back to try again later, and they rarely open a support ticket about it either.
An agent embedded in the product answers those questions immediately, with the specific steps for your product rather than generic advice, which turns silent drop off into a completed setup.
How do I give my developer access without paying for a seat?
Invite them as a viewer. Viewers do not consume a billable seat on the Team plan.
Where is that setting?
Settings, then Team, then Invite, and choose the Viewer role in the dropdown.
Escalation your support team will actually accept
Support engineers dislike chatbots mostly because of what arrives after them: a frustrated user and no context. Paperchat hands over the whole conversation, including any lookups it performed, into the same thread.
Your team is notified in Slack, picks up in the shared inbox, and the user carries on typing where they already were. Nothing is repeated and nothing is lost.
The full transcript
Everything the user already explained
Lookup results
Invoices or records the agent fetched
Slack notification
In the channel your team watches
One inbox
Web, Telegram, Instagram and Messenger
Frequently asked questions
Support automation for a SaaS product
Support volume in SaaS follows a predictable shape. A small number of question types account for most of the tickets, and those questions are answered somewhere in the documentation that almost nobody reads. Adding more documentation does not fix it, because the problem is retrieval rather than content. Users will type a question into a chat box in the product; they will not go and search a knowledge base first.
That makes an AI chatbot a good fit, but only if it answers from your material rather than from general knowledge about software. A model improvising about how your permissions system works will produce confident, wrong answers and generate more tickets than it deflects. Paperchat trains on your docs, help centre, changelog, uploaded files and pasted internal notes, and says when something is not covered rather than filling the gap. Keeping the changelog in the knowledge base is the single most effective habit, because it stops the agent describing a version of the product that no longer exists.
Billing is where documentation always fails, and where a connected system changes the outcome. Questions about a specific charge require a specific record: the invoice, the plan change, the proration. With Stripe connected the agent can fetch that during the conversation and explain what happened in plain language. That converts a category of ticket that previously always reached a human into one that mostly does not, while genuine disputes still escalate.
Onboarding deserves separate attention because the failure there is silent. A new user who cannot work out how to connect their data or invite a colleague rarely opens a ticket; they simply stop. An agent embedded in the product during the first session answers those questions immediately with steps specific to your product, which is a cheaper way to protect activation than another onboarding email sequence.
Finally, the handover is what determines whether your support team tolerates any of this. The classic failure is a frustrated user arriving with no context after four unhelpful bot replies. Paperchat moves the whole conversation, including anything it looked up, into the same thread and notifies your team in Slack. Pricing by message credits rather than per seat matters here too: the entire support team can work in the shared inbox without the cost changing, so nobody is incentivised to keep humans away from the queue.
More ways to use Paperchat
Use cases
AI customer support chatbot
Take the repetitive questions off your inbox and escalate the rest to a person.
Read moreIndustries
AI chatbot for agencies
Deploy trained chatbots for clients, with your branding and one dashboard.
Read moreUse cases
Lead generation chatbot
Answer the question first, then capture the contact while interest is high.
Read morePricing
Start on the free plan with no card. One credit is used per answer, so you only pay for conversations that actually happen.
Free
$0per month
- 100 message credits (~20+ conversations)
- 1 AI chatbot
- 800 KB knowledge base
- 1 team member
- Shopify, WooCommerce & WordPress actions
Basic
Most popular$35per month
- 4,000 message credits (~1000+ conversations)
- 1 AI chatbot
- 25 MB knowledge base
- 3 team members
- Conversation history
Pro
$99per month
- 10,000 message credits (~2500+ conversations)
- 3 AI chatbots
- 40 MB knowledge base
- 4 team members
- Conversation history
Larger plans and a lifetime option are on the pricing page.