Use Case

AI-powered support that actually knows your product

Deploy Conduity as your first-line support agent — on web chat, email, Slack, or WhatsApp. It answers questions from your docs, drafts responses, and escalates what it can’t handle. Your team focuses on the hard problems.

The Challenge

The Problem

Support doesn’t scale linearly. When your user base doubles, your ticket volume doubles, but your team size doesn’t. The result is predictable:

  • Response times creep up from hours to days
  • Simple questions eat your team’s time — “How do I reset my password?” answered for the 400th time
  • Knowledge is trapped in senior reps’ heads — new hires take months to become effective
  • Inconsistent answers across channels because every rep has a slightly different understanding
  • After-hours coverage means either paying for a night shift or letting tickets pile up until morning

Traditional chatbots don’t solve this. They follow decision trees. If the customer’s question doesn’t match a pre-written flow, they hit a dead end and hand off to a human anyway. Customers hate them because they feel robotic and unhelpful.

The Solution

How Conduity Is Different

Conduity agents don’t follow scripts. They read your documentation, understand your product, and have real conversations. When a customer asks a nuanced question, the agent reasons about the answer — pulling from your help docs, API references, and past conversations.

1

Grounded in your docs, not hallucinations

Upload your help center articles, API documentation, product guides, and FAQ. The agent uses RAG (retrieval-augmented generation) to find the most relevant sections and synthesize accurate answers. It cites its sources so customers can verify.

2

Multi-channel from day one

Deploy the same agent to web chat, email, Slack, and WhatsApp. A customer can start a conversation on your website and follow up via email. The agent has the full context because it’s the same agent with the same memory, not a separate bot per channel.

3

It learns from every interaction

When you correct the agent or add new information in a conversation, it remembers. If you tell it “we deprecated the v1 API last month,” it won’t recommend v1 endpoints to anyone again. The knowledge compounds over time.

4

Escalation, not abandonment

When the agent can’t answer confidently, it doesn’t loop the customer in a “sorry, I didn’t understand” cycle. It collects the context, summarizes the issue, and routes it to your team with all the details — so the human rep doesn’t start from scratch.

In Practice

A Day With Conduity Support

7:42 AM — Before your team is online

A customer in a different timezone hits your web chat:

Customer

I’m getting a 403 error when I try to access the /users endpoint with my API key.

Agent

A 403 error on the /users endpoint typically means your API key doesn’t have the required scope. The /users endpoint requires the users:read scope. You can check your key’s scopes in the Dashboard under Settings > API Keys.

If your key already has that scope, the issue might be IP allowlisting — if your workspace has IP restrictions enabled, requests from unlisted IPs will return 403. You can check this under Settings > Security.

Would you like me to walk you through updating your API key scopes?

The agent found the answer in the API documentation and the security guide. No human involved. The customer gets a solution at 7:42 AM.

9:15 AM — Email support, auto-drafted

A customer emails asking about pricing for their team of 15. The agent reads the email, checks the knowledge base for pricing information, and drafts a response:

Agent (Draft)

Hi Sarah,

Thanks for reaching out about team pricing.

For a team of 15, the Team plan would be the best fit:

  • Base price: $79/month (includes 5 seats)
  • Additional seats: 10 x $20/month = $200/month
  • Total: $279/month

The Team plan includes unlimited agents, integrations, automations, and knowledge documents. All 15 team members get full access.

If you’d prefer to save on an annual commitment, the same setup would be $232/month billed annually ($2,784/year — that’s a 17% savings).

Would you like me to set up a trial for your team?

The draft is saved in your conversation. Your rep reviews it, makes a small tweak, and sends it. What would have been a 10-minute research-and-write task took 30 seconds to review.

11:00 AM — Slack channel support for enterprise client

Your enterprise client has a dedicated Slack channel. A developer pings:

Client Dev

We’re seeing intermittent timeouts on the webhook endpoint. Is there a retry policy?

Agent

Yes — webhook deliveries are retried automatically with exponential backoff. The retry schedule is:

  • 1st retry: 30 seconds
  • 2nd retry: 2 minutes
  • 3rd retry: 15 minutes
  • 4th retry: 1 hour
  • 5th retry: 4 hours (final attempt)

If all retries fail, the delivery is marked as failed and appears in your webhook logs under Settings > Integrations > Webhooks > Delivery History.

For intermittent timeouts specifically: make sure your endpoint responds within 10 seconds. If processing takes longer, return a 200 immediately and process async.

Same agent, same knowledge base, different channel. The client gets an expert-level answer without your team context-switching from their current work.

2:30 PM — Escalation with full context

A customer reports a billing issue the agent can’t resolve:

Customer

I was charged twice for my February subscription and I need a refund.

Agent

I’m sorry about the double charge. Billing issues like this need to be handled directly by our support team to ensure your account is corrected properly. Let me get you connected.

I’ve flagged this for our billing team with the following details:

  • Issue: Duplicate charge for February subscription
  • Account: sarah@acmecorp.com
  • Amount: $79 x 2
  • Conversation context: Customer reached out via web chat at 2:30 PM, Feb 6

Someone from the team will respond within 2 hours. Is there anything else I can help with in the meantime?

The agent knows its limits. It doesn’t try to process refunds or access billing systems it shouldn’t. It escalates cleanly, with context, so the human rep can resolve it fast.

6:00 PM — Automated daily digest

Your end-of-day automation runs:

Agent

Support Digest — Thursday, Feb 6

Total conversations: 47 Resolved by agent (no human): 31 (66%) Escalated to team: 12 Pending response: 4

Top questions today: 1. API key scope errors (8 conversations) — consider adding a troubleshooting section to the API docs 2. Webhook retry policy (5 conversations) — frequently asked, well-documented 3. Team pricing for 10+ seats (4 conversations)

Sentiment: Mostly positive. 2 frustrated customers (both billing-related escalations).

Deployment Options

Web chat widget

Embed Conduity’s chat on your website or app. Customers get instant answers without leaving your product. The widget is customizable (colors, welcome message, position) on Team plan and above.

Email

Connect your support inbox. The agent reads incoming emails, drafts responses, and either sends them automatically or saves drafts for your team to review. Configure per-topic rules: auto-send for known questions, draft-only for billing or account issues.

Slack

Create a dedicated support channel for enterprise clients or internal support. The agent responds in-thread, keeping conversations organized. Works in shared Slack Connect channels too.

WhatsApp

For customer bases that prefer WhatsApp (common in LATAM, APAC, and Europe). Same agent, same knowledge, native WhatsApp experience.

Measuring Impact

After deploying Conduity for support, teams typically see:

MetricBeforeAfter
First response time2–6 hoursUnder 30 seconds
Resolution rate (no human)0%50–70%
Tickets per rep per day25–4015–20 (only complex ones)
After-hours coverageNone24/7
Onboarding time for new reps4–8 weeks1–2 weeks

The math is straightforward: if 60% of your tickets are answered by the agent, your team handles 60% fewer tickets. That’s either a cost saving or a quality improvement — your existing team spends more time on the hard problems instead of answering the same question for the hundredth time.

Knowledge Base Tips for Support

The quality of the agent’s answers depends directly on the quality of your documentation. Here’s what works:

Upload everything

Help center articles, API docs, product guides, FAQ, changelog, troubleshooting guides. More content means more answers.

Keep it current

When you ship a new feature or change a workflow, update the knowledge base. The agent’s answers are only as good as the docs it has access to.

Use the agent’s digest to find gaps

If the daily digest shows the same question coming up repeatedly with no good answer, that’s a documentation gap. Write the article, upload it, and the agent handles it from then on.

Separate internal and external docs

Use per-agent access control to make sure the support agent only sees customer-facing docs — not internal runbooks, pricing strategies, or employee handbooks. Different agents, different knowledge.

Support that scales without headcount

Deploy to web chat in under 10 minutes. No code required.

Start Free Trial