Customer Support
What goes wrong after launch, and how we handle it.
Support scenarios — what clients actually hit
1. Connectivity & sync
- Email account disconnected / drafts not appearing in their Mac Mail
- Calendar not syncing (meetings booked, then vanish)
- Social account auth expired (Sonny can't post)
- Website/blog publishing access revoked
2. Agent behavior
- Tone is off — too formal, too casual, doesn't sound like them
- Eva drafted a reply that should've been skipped (or skipped one that mattered)
- Penny's blog angle doesn't match their brand
- Stan sourced leads outside the ICP
3. Phone (Rachel)
- Calls routing to voicemail when they should transfer
- Rachel handling a call they wanted, or transferring one they didn't
- Client wants the dedicated number changed or call script tweaked
4. Workflow & outcomes
- Leads not landing in the right CRM/Odoo pipeline
- Stan's emails hitting spam
- Booking link broken or pointing to wrong calendar
- Agent promised a discount/refund beyond what the client allows
5. Administrative
- Adding/removing an agent, changing business hours, swapping languages
- Billing questions, plan changes, adding seats
- Client-side admin changes (new owner, new approver)
6. Emergencies
- Agent stopped responding entirely
- Something wrong went out (post, email, call) and needs pulling back fast
How we provide support
Tier 0 — Self-serve (always available)
Help center + short video walkthroughs for the top 10 issues (reconnect email, change voice, add agent, adjust hours). Most clients can fix a tone tweak themselves in 2 minutes.
Tier 1 — Eva triages (same day)
Support inbox runs through Eva: labels, answers common questions from the playbook, and escalates what needs a human. Client never waits to hear "we got it."
Tier 2 — Human (JV/team)
Anything Eva can't resolve: complex config, billing, angry-customer situations, workflow redesign. Owner is always in the loop on anything sensitive.
Tier 3 — Platform (Marblism)
Actual platform bugs (agent down, tool failure) get escalated to Marblism with full context so we don't ping-pong.
SLAs
| Severity | Example | Response |
|---|---|---|
| 🔴 Critical | Agent down, wrong thing went out | Same business day, within 2 hrs |
| 🟡 Normal | Tone tweak, sync hiccup, question | 24 hrs |
| 🟢 Low | Feature request, optimization | Weekly batch / roadmap |
Ongoing optimization
- First 2 weeks: daily 15-min client review — we tune agents as real traffic hits
- Monthly: review call — what's working, what needs adjusting, new content/voice updates
- Quarterly: health check — are the agents still aligned with where the business is going?