Job description

THE PROBLEM Flock's Fin/Intercom AI agent is the first response most customers get. That agent has to sound like Flock, resolve real issues, and know precisely when to stop and hand off to a human, especially on trust-critical accounts. Right now that design work is split across whoever has time for it. We need one person who owns the craft of conversation design as the agent portfolio grows past this first launch. WHAT YOU'LL OWN - End-to-end conversation design for Flock's AI support agents, from intent mapping through prompt development to production deployment. - A measurable lift in AI-resolved conversations without a rise in bad handoffs or missed trust-critical escalations. - The escalation and fallback logic that routes SEV incidents, VIP accounts, and Safe City accounts to a human every time. - A conversation analytics practice that turns agent performance data into a prioritized backlog of design fixes. - Documentation and design standards that let the agent portfolio scale past one bot without design quality slipping. - Partnership with Support, Product, and Trust stakeholders to keep agent behavior aligned with SLA and safety requirements. WHAT THIS ROLE IS NOT - Not a scripting job. This isn't static decision trees; it's designing and iterating on LLM-driven conversations with real performance data behind every change. - Not an unsupervised AI role. Trust-critical decisions, SEV incidents, and VIP/Safe City accounts stay human-reviewed. This role designs the guardrails; it doesn't replace the judgment call. - Not a general support operations role. This is conversation design and AI agent craft specifically, not queue management or ticket triage. WHAT YOU BRING - Experience with Intercom/Fin specifically, or a comparable modern AI agent platform. - Demonstrated ability to design and iterate conversational flows, including intent mapping, utterance training, and fallback strategies, in a live AI agent platform. - Hands-on prompt development experience with LLM-based conversational agents, not legacy decision-tree bots. - Fluency reading conversation logs and performance dashboards to diagnose failure points and prioritize fixes. - Writing that reads naturally in a customer-facing bot voice: clear, concise, on-brand. COMPENSATION In this role, you'll receive a starting salary between $85,000 - $95,000 as well as Flock Stock Options. Base salary is determined by job-related experience, education/training, as well as market indicators. Your recruiter will discuss this in depth with you during our first chat. WHY FLOCK 💚 Every community deserves to be safe. Flock builds the technology that makes that real: last year we supported over 1 million criminal investigations and helped locate more than 10,000 missing people. We're 1,700 people building the impossible with over $1B in funding, and the expectations are high on purpose. If you want a role where the stakes are real and the pace matches, this is it. 📍Some problems get solved faster in the same room, so we prioritize candidates in Atlanta and Boston. Hub-based roles mean real in-person time with your coworkers. Remote roles exist, and when a posting is open to remote work, it says so. 🌟 The impossible takes every kind of mind. Flock is an equal opportunity employer, and we know the best solutions come from diverse perspectives, experiences, and skills working together with mutual respect. Everyone is welcome to apply. If you need assistance or an accommodation due to a disability, email recruiting@flocksafety.com; your information stays confidential and is used only to arrange the right accommodation for your interviews. 💰 On compensation. We pay fairly for the work. Base salary is determined by job-related experience, education, training, and market indicators. The range in this posting covers base salary only and doesn't include equity, sales bonus plans where applicable, or benefits. The range may be adjusted over time, and this posting may span more than one career level. 📣 One last thing. All legitimate communication from Flock comes from an email ending in @flocksafety.com flocksafety.com or noreply@ashbyhq.com. We never make offers through messaging apps or third parties, and we never request payment or sensitive personal information during hiring. If you encounter suspicious outreach about a Flock role, report it to recruiting@flocksafety.com.

Requirements

Must-haves

End-to-end conversation design for Flock's AI support agents, from intent mapping through prompt development to production deployment.
A measurable lift in AI-resolved conversations without a rise in bad handoffs or missed trust-critical escalations.
The escalation and fallback logic that routes SEV incidents, VIP accounts, and Safe City accounts to a human every time.
A conversation analytics practice that turns agent performance data into a prioritized backlog of design fixes.
Documentation and design standards that let the agent portfolio scale past one bot without design quality slipping.
Partnership with Support, Product, and Trust stakeholders to keep agent behavior aligned with SLA and safety requirements.
Not a scripting job. This isn't static decision trees; it's designing and iterating on LLM-driven conversations with real performance data behind every change.
Not an unsupervised AI role. Trust-critical decisions, SEV incidents, and VIP/Safe City accounts stay human-reviewed. This role designs the guardrails; it doesn't replace the judgment call.
Not a general support operations role. This is conversation design and AI agent craft specifically, not queue management or ticket triage.
Experience with Intercom/Fin specifically, or a comparable modern AI agent platform.
Demonstrated ability to design and iterate conversational flows, including intent mapping, utterance training, and fallback strategies, in a live AI agent platform.
Hands-on prompt development experience with LLM-based conversational agents, not legacy decision-tree bots.
Fluency reading conversation logs and performance dashboards to diagnose failure points and prioritize fixes.
Writing that reads naturally in a customer-facing bot voice: clear, concise, on-brand.

Nice-to-haves

None flagged.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aillm

Platform: ashby
Effort: quick