Skip to content
YC Root AccessYC Root Access

Parahelp: The End-to-End AI Support Agent

Parahelp is building an end-to-end AI support agent. One that doesn’t just reply, but actually resolves tickets by taking real actions inside a company’s tools. The team recently raised an $18M Series A to scale this vision as their agent now closes thousands of tickets every day. In this conversation with YC's Jared Friedman, founders Anker Ryhl and Mads Liechti share how they went from cold-emailing 700 prospects during YC to powering support at companies like Perplexity, ElevenLabs, and Replit. They talk about the mid-batch pivot that changed everything and how their new agentic architecture lets them onboard customers in hours instead of weeks. Learn more about Parahelp at https://parahelp.com. 00:00 – Intro: What Parahelp Is Building 00:46 – The First Version (and Why It Didn’t Work) 03:12 – YC, Cold Emailing 700 Prospects, and Finding the Problem 06:05 – The Mid-Batch Pivot That Changed Everything 08:40 – Landing Perplexity as Customer #2 11:28 – What “End-to-End Resolution” Actually Means 14:20 – How the Agent Takes Real Actions (Refunds, Fixes, Workflows) 18:05 – Building the Agentic Architecture 22:18 – Scaling to Thousands of Tickets Closed Every Day 29:04 – Onboarding Customers in Hours, Not Weeks 34:40 – Raising the $18M Series A 40:52 – Advice for Founders

Anker RyhlguestMads LiechtiguestJared Friedmanhost
Dec 1, 202547mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Parahelp builds end-to-end AI support agents that take real actions

  1. Parahelp pivoted from an NFT investment app to B2B support automation after discovering that software companies needed not just AI replies but end-to-end ticket resolution with actions.
  2. The product differentiates by safely calling tools (e.g., Stripe refunds) with guardrails, approvals, validations, and continuous evaluation to prevent regressions as company knowledge changes.
  3. A new “assistant” architecture automates onboarding and ongoing configuration by generating evals, editing policy/knowledge files, testing in sandbox, and publishing changes—reducing reliance on forward-deployed engineers.
  4. A research agent mines thousands of historical tickets to find resolution patterns, propose policy/tool updates, and drive a continuous improvement flywheel that measurably raises resolution rates.
  5. The company focuses on fast-moving software and AI-native teams, positioning value around leverage and revenue enablement (retention/upsell) more than pure support cost reduction, and raised an $18M Series A led by Jack Altman (Old Capital).

IDEAS WORTH REMEMBERING

5 ideas

End-to-end resolution requires action, not just good replies.

Parahelp’s core claim is that an AI support agent must execute the full workflow—refunds, account fixes, debugging steps—by using integrated tools, not merely draft responses for humans.

Tool calling at scale is an evals-and-guardrails problem.

They argue “tools” aren’t just API endpoints; reliability comes from deterministic validations, human approval flows for sensitive actions, and always-on monitoring so changes in prompts/knowledge don’t silently degrade behavior.

Fast-growing software companies need living knowledge, not static RAG.

Default support AIs and internal builds often start strong then decay because context in Slack/Linear/Notion changes daily; Parahelp focuses heavily on keeping policies and “memory files” current and de-duplicated.

Automating onboarding is the next scaling breakthrough for vertical agents.

Their upgraded assistant effectively replaces forward-deployed engineers by generating evaluation sets, editing policies like “cloud code,” running sandbox tests, and prompting humans only for missing inputs.

Multi-agent ‘modes’ help manage context and reliability.

The assistant switches between Ask/Configure/Test/Research modes with distinct prompts and tools, compacting context between modes to keep conversations bounded while still completing long, structured work.

WORDS WORTH SAVING

5 quotes

Parahelp is an AI support agent for customer service that resolves support tickets end to end.

— Mads Liechti

Tools are way more than just calling the API endpoint. It’s having the proper guardrails… and an evaluation system… to make sure we’re not regressing.

— Anker Ryhl

Since launching this, we haven’t written a single custom evaluation set because our agent does it for our customers.

— Anker Ryhl

We often give the analogy… Parahelp is kinda like having 1,000 super capable employees that are also getting better every day.

— Anker Ryhl

As soon as we start thinking more about what we’re afraid of losing… then we’ve already lost.

— Anker Ryhl

Origin story and pivots (Denmark to SF/YC)Mid-batch YC pivot to “end-to-end resolution”Cold outbound to land first customers (Captions, Perplexity)Tool execution and safety (Stripe, approvals, validations)Knowledge management across Linear/Notion/Slack/RetoolAgentic architecture: agent vs assistant, multi-mode workflowContinuous evaluation, monitoring, and research-driven improvementsScaling to thousands of tickets closed dailySlack-native workflows and enterprise supportSeries A rationale and founder advice

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome