# AI-Augmented Customer Experience at HackerOne
- Session: 9058b934-6970-4aaa-8895-ba4b3a7d3466
- Channel: Discord #🐣│cohort-voice
- Started: 2026-06-23T17:00:10.802Z
- Ended: 2026-06-23T17:33:30.941Z
- Participants: duckanbro, 0xHunter, Aphilos • Pharo, chickenparm, ECWireless, louchi, samkuhlmann
- Tags: ai-customer-support, hackerone, claude-code, human-in-the-loop, support-automation, slack-app, ai-adoption
## TL;DR
Jake from HackerOne discussed how his customer experience team is using Claude Code and internal AI tooling to manage a sharp increase in support ticket volume, build a Slack-based support workflow, improve response suggestions with historical data, and keep humans in the loop for trust, security, and judgment-heavy issues.
## Summary
The conversation centered on how HackerOne is adapting customer experience operations to an AI-driven increase in activity. Jake, who leads customer experience at HackerOne, explained that AI agents are changing both vulnerability discovery and support operations: hackers and companies are using agents to find and submit more issues, and support ticket volume has increased dramatically because it is easier for users to generate requests.
Jake described building an internal Slack app, Dispatch, using Claude Code despite not being a developer by trade. The app pulls in tickets from Freshdesk, suggests responses based on historical tickets, GitLab and Linear issues, documentation, and resolved cases, and lets support agents approve, edit, escalate, or approve-and-resolve tickets directly from Slack. The team is intentionally keeping humans in the loop while trust and confidence scores improve, with the long-term goal of fully automating high-volume, low-risk ticket types once confidence reaches a high threshold.
The discussion covered operational improvements, including reduced context switching, automated ticket property filling, easier reporting through tagging AI-handled tickets, and faster iteration on tooling. Jake contrasted adding a new workflow button in minutes through Claude Code with the week or more it previously might have taken through a formal development request. He also emphasized HackerOne's security posture: AI tooling moved from tightly restricted to broadly encouraged once the company was comfortable with approved environments, data classification policies, and internal controls.
Jake also shared current pain points. Access to data remains the biggest blocker: if there is no API key, if he is not allowed to use it, or if an API does not exist, the agent cannot help as much. Rate limits were another learning point, leading to caching and scheduled background fetches. Confidence scoring is improving as more historical data is made available, but still needs refinement, especially when a response is mostly correct but contains one unacceptable section.
The meeting closed with a broader reflection on AI's effect on support roles. Jake was neither fully optimistic nor pessimistic, but argued that human-in-the-loop workflows will remain important for the next few years, especially for payment, mediation, technical troubleshooting, and other cases requiring judgment. He described upskilling support agents through a three-month training program so they can handle more technical work and use AI tools effectively rather than feel replaced by them.
## Action Items
- Track Dispatch impact with reporting tags: Use the Dispatch tag on tickets handled or suggested by the Slack app so the team can compare AI-assisted ticket volume and outcomes against prior months. (owner: chickenparm)
- Continue tuning confidence scoring: Refine how approvals, rejections, edits, and partial corrections affect confidence scores so the system can better distinguish mostly correct responses from sendable responses. (owner: chickenparm)
- Expand safe data access for support automation: Work with internal teams to obtain approved API access to additional systems where support-relevant data lives, while respecting security and data classification policies. (owner: chickenparm)
- Use meeting transcript for AI adoption research content: Ingest the recording and transcript into the broader research effort on how different industries and roles are adapting to AI. (owner: duckanbro)
## Notable Quotes
- chickenparm: "Cloud Code has kind of changed my life and the way I work."
- Jake said Claude Code has dramatically changed how he operates as a non-developer business and customer experience leader.
- chickenparm: "We went from getting X amount of tickets a month to like three X overnight."
- AI has made it much easier for users to submit support requests, causing a sharp increase in ticket volume.
- chickenparm: "A year ago, I would have had to submit a dev request... today I have it done in five minutes."
- Jake highlighted how AI coding tools let him ship internal workflow changes in minutes instead of waiting on scarce engineering capacity.
- chickenparm: "Data, data, data, data is the biggest way to be heard."
- Within HackerOne, Jake said measurable evidence is essential for getting leadership buy-in.
- chickenparm: "I think human loop is going to be very important."
- Jake expects human review to remain central in support automation, especially for sensitive or judgment-heavy cases.