15-minute first call
Discovery owns the first call. Demo expands on later calls. Objection handling is reactive — distributed across all beats, not a separate segment. Budget ~10–15% of responsive bandwidth without disrupting structure.
The 15-minute target is the structured backbone. Calendly books a 30-minute slot, so there's a full 15 minutes of headroom for prospect follow-up questions, deeper objections, or multi-thread setup. If the buyer is engaged and discovery surfaces real depth, let the call breathe — don't truncate to hit the clock. The goal is to land the math and the demo within 15 minutes so the remaining time is theirs.
| Beat | % of call | Minutes |
|---|---|---|
| Open + rapport | 5% | ~1 |
| Discovery + live ROI calculator (cost-of-inaction emerges during diagnosis) | 50% | ~7 |
| Live product demo (Zendesk feed + synthetic tickets + console backup) | 30% | ~5 |
| Outcomes + Close + Multi-thread | 15% | ~2 |
Verbatim — deliver in order, then move into §3
~45 seconds total. Senior buyers know who you are when they accept the meeting — don't over-credential. As soon as you get a one-sentence response from Beat 2, push to Beat 3.
"Saw on LinkedIn you've been at [Company] about [X] years — how's [the eCom side / your queue / the team] handling [recent context] this year?"
"Saw the news about your [Q3 raise / product launch / new market] — congrats."
Diagnostic, not sales weapon (~7 min)
Have paraison.io/roi open on your shared screen from the start. Type her numbers in real time as she answers. Cost-of-inaction emerges naturally as her current monthly spend appears on screen — you don't have to pitch it.
SPIN questions, mapped to calculator fields
Eight questions in SPIN order, two per beat. Type her answers into the calculator as she gives them. By the end of discovery, she will have watched her current annual spend appear, and the savings/payback math will be visible — without you having to "reveal" anything.
Then pivot to demo: "Now let me show you how those deflections actually happen."
Zendesk + synthetic tickets + operator console (~5 min)
After discovery, show — don't tell. This is the segment most vendors can't deliver because their demo is a sandbox or a pre-recorded video. You'll fire actual tickets at the live tenant during the call and let her watch real classification, real routing, real timing end-to-end.
Setup before the call
- Three pre-loaded Zendesk API curl commands — one per bucket (Tier 0, Tier 1, VIP). Creates a real Zendesk ticket → fires the Zendesk webhook → ACP classifies → ACP writes back routing + reply. End-to-end pipeline visible in Zendesk in ~5–7 seconds. Payloads in §5.
- Fallback screenshots for each bucket — if Anthropic API is slow mid-demo, you don't watch dead air.
- Zendesk view bookmarked — sorted newest-first, refreshed in a tab.
- console.paraison.io — pre-logged-in to
paraison_internaltenant. - Tenant scan — eyeball the visible ticket history before sharing screen; nothing in the test data should expose anything that shouldn't be visible to a prospect.
The 90-second opener (talk track)
— VIP tickets — high-value accounts tagged in your Zendesk. Bypass automation entirely; route directly to a senior agent. Deterministic, no model decides.
— Tier 0 — inbound matches a KB article you already approved. ACP synthesizes a reply from your approved content; you set the confidence threshold per category. This is where the deflection economics come from.
— Tier 1 — doesn't match a KB. ACP classifies into your existing Zendesk groups and routes — same workflow your team already runs, but with routing decisions logged and confidence scored.
Anything the classifier isn't confident about queues for human review. Let me show you live."
Live demo (~3.5 min)
Fire a Tier 0 test ticket
"Let me show you Tier 0 end-to-end." Run the Zendesk API curl. Tab to Zendesk. Watch the sequence land in real time:
- Ticket appears (within 1 sec)
- Internal note appears with category + confidence + matched KB article (~2–4 sec)
- Public reply appears (AI-generated from the approved KB content)
- Status changes to Solved
Fire a Tier 1 test ticket
"Now one that doesn't match KB." Run the curl. Watch:
- Ticket appears
- Internal note appears with category + confidence + routing rationale
- Group assignment updates to the correct Zendesk team
- Status stays open, assigned and waiting
Fire a VIP test ticket
"Now a VIP-tagged ticket — different path entirely." Run the curl with a VIP requester. Watch:
- Ticket appears with VIP tag visible
- Internal note: "VIP override — bypassing classifier"
- Assignment lands directly on a senior agent
Console as the "why" backup
Buyer will often ask "how do I see WHY ACP decided that?" Pivot to console.paraison.io, pull the acp_decisions row matching the most recent trace_id, walk the captured fields: category, confidence, predicted agent, action taken, timestamp.
The three pre-loaded commands
Export your Zendesk creds once per terminal session. Auth format is email/token:API_TOKEN — the literal string /token follows the email (Zendesk-specific Basic Auth).
requester block, Zendesk defaults the requester to the authenticated user — every ticket would show as from Ron Paraison, which a prospect will spot in two seconds. Each payload below sets a synthetic end-user on paraison-ai.io (plus-aliases route back to your inbox if anything auto-replies). "tags":["demo"] filters these out of real KPI rollups. Edit subject/body to match a real KB topic in paraison_internal before the call.# Export once per terminal session export ZD_SUBDOMAIN="paraisonai" # your Zendesk subdomain export ZD_EMAIL="ron@paraison.io" # auth user export ZD_API_TOKEN="..." # Zendesk Admin → Apps & integrations → APIs # Tier 0 — matches a KB article, ACP replies from approved content curl -sS -u "$ZD_EMAIL/token:$ZD_API_TOKEN" \ -H "Content-Type: application/json" \ "https://$ZD_SUBDOMAIN.zendesk.com/api/v2/tickets.json" \ -d '{"ticket":{"subject":"How do I reset my password?","comment":{"body":"I forgot my password and need to log back in. How do I reset it?"},"requester":{"name":"Avery Demo","email":"ron+avery.demo@paraison-ai.io"},"tags":["demo"]}}' # Tier 1 — no KB match, ACP classifies + routes to the right Zendesk group curl -sS -u "$ZD_EMAIL/token:$ZD_API_TOKEN" \ -H "Content-Type: application/json" \ "https://$ZD_SUBDOMAIN.zendesk.com/api/v2/tickets.json" \ -d '{"ticket":{"subject":"Webhook payloads arriving out of order","comment":{"body":"Our processing pipeline expects sequential events but ~3% of webhooks arrive late and out of order over the past week. Can you confirm whether there is queue contention on your side?"},"requester":{"name":"Jordan Demo","email":"ron+jordan.demo@paraison-ai.io"},"tags":["demo"]}}' # VIP — tagged ticket bypasses classifier, lands on a senior agent curl -sS -u "$ZD_EMAIL/token:$ZD_API_TOKEN" \ -H "Content-Type: application/json" \ "https://$ZD_SUBDOMAIN.zendesk.com/api/v2/tickets.json" \ -d '{"ticket":{"subject":"SSO down for our org","comment":{"body":"Our SSO integration is down. CISO is asking for an incident report."},"requester":{"name":"Casey Demo","email":"ron+casey.demo@paraison-ai.io"},"tags":["enterprise","demo"]}}'
Nine outcomes — five lead, two risk reducers, two reserves
Memorize the five "lead" outcomes — you'll use these on every Maya call. The other four are tiered for specific moments (risk hesitation, governance review, CFO depth on demand).
Lead outcomes — drill cold, use on every Maya call
Routing accuracy improvement
Tier 1→Tier 2 escalations drop because tickets land in the right team the first time. Misroutes from her existing triage logic get absorbed by the classifier; operator corrections are logged and ACP learns from them.
Agent time freed for resolution
Less triage drudgery, less context-switching. Agents spend their minutes on resolution, not routing — the front of the workflow compresses, average handle time comes down.
Deflection lift on the easy stuff
Net $21.35 saved per Tier 0 ticket the RAG path absorbs. Even at a conservative 10% deflection — well below her likely real number — the math is positive in month one.
OPEX, not headcount
Per-resolution lives in OPEX. Not a headcount story, not a layoff story. Cost-reduction without HR risk — passes budget committees even in a hiring freeze. Lead with this when Marcus joins.
Speed to value — live in three business days
From completed intake to operational tenant in three business days. Her team gets the routing accuracy + agent time + deflection lift within the same week, not the same quarter.
Risk reducers — volunteer only if she hesitates
No customer-facing outage if ACP goes down
Sits above Zendesk, not in front of it. If we go dark, her queue keeps moving in Zendesk exactly as it did before ACP.
Month-to-month, cancel anytime
Thirty days notice on either side. If we raise the rate, she walks free — no termination fees, no minimum commitment.
Hold in reserve — only if discovery surfaces the specific need
Per-category auto-thresholds
For the "I don't trust automation yet" concern — gradual trust-building.
Full audit trail in acp_decisions
For CIO/governance review — cluster with security brief, not a lead outcome.
Three-beat short answers
Three beats, not two: punchline (answers the literal question), frame (mechanism that makes the punchline true), bridge (ties back to an outcome she felt during discovery). The bridge keeps the rebuttal from being purely defensive — it forwards the deal.