AI & Automation
Using data and AI to deliver support that’s faster, smarter, and more human — not instead of people, but so people can show up where it matters most.
When we migrated to Intercom, we didn’t just move to a new platform — we unlocked a new level of visibility. For the first time, we could see exactly what was driving support volume, where conversations were stalling, and which moments were costing customers the most effort. That data became the foundation for everything we built next.
The approach is deliberate: analyze conversation patterns to find the signal, design automation or workflow changes to address it, measure the result, and feed those findings back into the next cycle. Data tells us where AI helps. Data tells us where humans need to step in. And data tells us when we've actually made things better.
The goal was never to automate support. It was to use automation as a lever — so that when a customer has a complex, emotionally charged problem, a human agent is available, informed, and ready. High tech makes high touch possible at scale.
Each quarter built on the last. Data from one cycle became the brief for the next.
H2 2025 — Foundation
Data input: conversation volume by topic, CX score baselines, SLA gap analysis.
After completing the migration, achieved 45% self-serve rate with AI agent. Further improved docs and customized workflows. Established SLA framework with P0–P3 priority ratings. Built agent dashboards to surface volume drivers. Defined QA rubric to measure support quality consistently.
Q1 2026 — Refinement
Data input: Low CX score AI-to-human handoff failures, agent first response time.
Analyzed high-effort moments where AI was failing customers. Built proactive escalation logic routing complex conversations based on sentiment and complexity. Launched personalized AI automation using customer account data. Made first response time 55% faster.
Now: Q2 2026 — Proactive
Data input: missing payment info, early-stage user patterns, urgency analysis.
Mapping payment processor data fields to billing conversation types to power account-based automations. Building urgency detection to guarantee sub-1-hour response for critical issues. Developing proactive workflows for new users before problems arise.
Each initiative started the same: with data pointing at a problem worth solving.
Data signal
Billing inquiries mapped to specific Stripe fields: payout dates, charge status, subscription details.
Bridge account systems
Connecting Stripe and Intercom to automate routine account questions — payout dates, purchase verification, trial confirmations — so agents spend time on conversations that actually need them.
Data signal
Time-sensitive issues queuing with everything else despite requiring sub-1-hour response.
Meet critical moments
Building AI urgency detection that automatically identifies and prioritizes high-impact conversations based on sentiment and complexity — so a creator’s launch-day crisis never lingers too long in a queue.
Data signal
New users reaching out with blockers before their first sale — support waiting reactively.
Facilitate early success
Mapping and analyzing early-stage pain points, and then building page-specific proactive workflows — so we can guide new users to the answer even before the question arises, improving both conversion and retention.
The feedback loop
The most important thing we built isn’t any single automation — it’s the loop. Each quarter, we establish baselines, run experiments, measure outcomes, and feed those findings into the next cycle’s priorities.
The data doesn’t just tell us what’s broken. It tells us what to build next — and whether what we built actually worked.
That compounding effect — each quarter smarter than the last — is what “high tech, high touch” looks like in practice.
