Migrating From Vapi to CallSphere: 14-Step Cutover Checklist

TL;DR

Migrating from Vapi to CallSphere is a 14-step process that typically runs 6–12 weeks end to end. The bulk of the time is dual-run — both stacks live, traffic split, comparing answer rate / AHT / sentiment — not the technical cutover. This is the production-grade checklist used by teams that have done this migration before.

Who This Playbook Is For

Engineering leads, ops directors, and IT managers responsible for an existing Vapi-based voice AI deployment who have decided to move to CallSphere. You already passed the build-vs-buy decision; now you need a credible cutover plan to get there without dropping calls.

Migration Timeline at a Glance

gantt
    title 14-Step Vapi to CallSphere Migration
    dateFormat YYYY-MM-DD
    section Discovery
    1. Inventory Vapi assets       :a1, 2026-04-15, 5d
    2. Export prompts and KBs      :a2, after a1, 3d
    3. Map tools to CallSphere     :a3, after a2, 4d
    section Build
    4. CallSphere tenant provision :b1, after a3, 2d
    5. Vertical template selection :b2, after b1, 1d
    6. KB ingest and re-tag        :b3, after b2, 5d
    7. Tool re-implementation      :b4, after b3, 7d
    8. Dashboard + RBAC config     :b5, after b4, 3d
    section Validate
    9. Synthetic call testing      :c1, after b5, 5d
    10. Dual-run setup             :c2, after c1, 3d
    section Run
    11. Dual-run period            :d1, after c2, 21d
    12. Number porting             :d2, after d1, 7d
    13. Cutover                    :d3, after d2, 1d
    section Stabilize
    14. Post-cutover monitoring    :e1, after d3, 14d

The 14 Steps in Detail

Step 1 — Inventory Vapi Assets

Before you migrate anything, document what you actually have. Most teams discover their Vapi deployment is more sprawling than they realized.

Capture:

• Every Vapi assistant (production, staging, experimental)
• Every system prompt with version history
• Every tool definition with webhook URLs
• Every phone number and routing rule
• Every voice configuration
• Every observability/analytics dashboard
• Every BAA, DPA, and MSA in force across the supporting vendor stack

Output: a single Migration Inventory doc.

Step 2 — Export Prompts, Knowledge Bases, and Transcripts

Pull the data out while you still have access:

• All system prompts as plain text or JSON
• Knowledge base content as Markdown / JSON / CSV
• The last 30–90 days of transcripts for regression testing
• Tool call logs to reconstruct expected behaviors

Store in version control. You'll need these for testing parity later.

Step 3 — Map Tools to CallSphere Equivalents

This is the highest-leverage step in the migration. CallSphere's vertical templates already include 4–14 tools per vertical. Your Vapi tools likely map to a subset of those, plus a few custom ones.

Build a table:

Vapi Tool	CallSphere Equivalent	Action
book_appointment	booking_tool (Healthcare/Salon)	use template
verify_insurance	insurance_lookup_tool	use template
custom_loyalty_check	(none)	implement as custom webhook
send_sms_reminder	reminder_tool	use template

Custom tools become CallSphere webhook tools — same pattern, cleaner contract.

Step 4 — Provision CallSphere Tenant

Sign up. Pick the tier that matches your current volume (Starter, Growth, Scale, or Enterprise). Provision a tenant. Configure RBAC roles for your team — admin, manager, sales_rep/agent, requester as fits.

Step 5 — Select the Vertical Template

Six options: Healthcare, Real Estate, Sales, Salon, After-Hours, IT Helpdesk.

If your business cleanly maps to one, pick it. If you straddle two (e.g., dental practice with after-hours coverage), Healthcare + After-Hours patterns combine naturally.

If you straddle none — i.e., you're in a vertical CallSphere doesn't have a template for yet — talk to CallSphere sales. The closest template is usually a strong starting point.

Step 6 — Knowledge Base Ingest and Re-Tag

Ingest the KB content you exported in step 2. Re-tag and structure it for CallSphere's data model:

• Hours, locations, contact info as structured fields
• Services as a services table with codes (CDT for dental, CPT for medical, etc.)
• Policies as Markdown
• FAQs as Q/A pairs

CallSphere's ingestion handles both bulk upload and ongoing CMS sync.

Step 7 — Tool Re-Implementation

Implement custom tools as CallSphere webhooks. The contract is straightforward — JSON schema in, JSON out, with a clear set of allowed actions.

For each Vapi tool not covered by the vertical template:

1. Identify the upstream API the tool calls
2. Define the input schema
3. Define the output schema
4. Implement the webhook handler
5. Test in isolation
6. Wire to CallSphere

This step is the one most often underestimated in time. Budget 1–3 days per non-trivial custom tool.

Step 8 — Dashboard and RBAC Configuration

Configure the operational UI:

• Who has admin access (full config + billing)
• Who has manager access (config but not billing)
• Who has sales_rep / agent access (operational use)
• Who has requester access (read-only specific scopes)

Branding: logo, color scheme, sub-domain if you're using white-label.

Step 9 — Synthetic Call Testing

Run scripted test calls covering the top 30–50 conversation patterns from your transcript export. Compare CallSphere's responses against your Vapi historical responses.

Look for:

• Functional parity (booking succeeds, verification returns correctly)
• Voice quality acceptable
• Latency within target (sub-2 second time-to-first-token in most flows)
• Escalation triggers fire correctly
• Transcript and analytics capture is complete

Iterate prompts and tool configurations until parity hits 95%+ on your top patterns.

Step 10 — Dual-Run Setup

This is the safety net of the entire migration. You configure both Vapi and CallSphere to receive a portion of live traffic in parallel.

Two patterns:

A — Number-level split. Some phone numbers route to Vapi, others to CallSphere. Lower technical complexity, but harder to compare outcomes on the same conversation type.

B — Caller-cohort split. A subset of callers (e.g., specific area codes, or random %) route to CallSphere; the rest stay on Vapi. Cleaner comparison.

Most teams use pattern B with a 10% → 25% → 50% → 100% ramp.

Step 11 — Dual-Run Period

Run both stacks live for 2–4 weeks. Measure on identical KPIs:

KPI	Vapi	CallSphere	Target
Answer rate	baseline	track	parity or better
Average handle time	baseline	track	parity or better
Booking conversion	baseline	track	parity or better
Sentiment score	baseline	track	parity or better
Escalation accuracy	baseline	track	parity or better
First-call resolution	baseline	track	parity or better
p95 latency	baseline	track	parity or better
Call cost	baseline	track	better expected

Daily standups review the metrics, surface issues, fix them, continue.

Step 12 — Number Porting

Once dual-run shows CallSphere matching or beating Vapi on the KPIs, port the numbers.

Number porting takes 5–10 business days per batch. Submit Letters of Authorization (LOA) to your current carrier (typically Twilio under Vapi). CallSphere's integrated Twilio relationship simplifies this — same underlying carrier, just different orchestration.

Schedule porting on a low-traffic window (Sunday morning is common) to minimize risk.

Step 13 — Cutover

The actual cutover is anti-climactic if dual-run was thorough. Update routing so 100% of calls land on CallSphere. Keep Vapi running in standby for 48–72 hours as a fallback.

Communicate to the team:

• "We've cut over to CallSphere as of [date/time]"
• "Here's the dashboard URL"
• "Here's the support contact"
• "Here's the rollback procedure if needed"

Step 14 — Post-Cutover Monitoring

For 2 weeks post-cutover:

• Daily KPI review against the dual-run baseline
• Spot-check 10–20 calls per day for quality
• Address any regressions immediately
• Document new edge cases discovered in production

After 2 weeks of stable operation, decommission the Vapi stack. Cancel vendor contracts. Reclaim the budget. Update the security posture documentation to reflect the simplified vendor footprint.

Common Migration Pitfalls

Pitfall 1 — Skipping the inventory. Teams who skip Step 1 inevitably miss a tool, a phone number, or a prompt and discover it post-cutover when something breaks.

Pitfall 2 — Underestimating custom tool re-implementation. Step 7 is the time sink. Budget realistically.

Pitfall 3 — Cutting over before dual-run shows parity. Dual-run exists for a reason. If CallSphere is lagging Vapi on 2 KPIs out of 8, fix those 2 before cutover. Don't ship and pray.

Pitfall 4 — Forgetting observability cutover. Your team's dashboards, alerts, and on-call rotations need to point to CallSphere too. Don't leave Datadog dashboards monitoring an empty Vapi stack.

Pitfall 5 — Decommissioning Vapi too fast. Keep Vapi running in standby for 48–72 hours. The cost is trivial; the safety net is invaluable.

Migration Timeline by Organization Size

Org Size	Timeline	Notes
Solopreneur	2–4 weeks	Single tenant, single number, simple KB
Small business (5–25 employees)	4–6 weeks	Single vertical, 1–3 custom tools
Mid-market (50–250 employees)	6–10 weeks	Multi-location or multi-line, 3–8 custom tools
Enterprise (250+)	10–14 weeks	Multi-tenant, complex integrations, formal change mgmt
Agency reseller (10–50 clients)	14–20 weeks	Phased per-client migration

Cost of Migration

Honest accounting:

Cost Line	Range
Internal eng time	60–200 hours
Number porting fees	$20–$50 per number
Dual-run overlap (paying both)	2–4 weeks of both bills
Knowledge base re-tagging	10–40 hours
QA / synthetic call testing	20–60 hours

Most migrations pay back in 8–14 weeks based on monthly run-rate savings.

When NOT to Migrate

Be honest. Don't migrate if:

• Your Vapi build is genuinely working well, your team is happy, and the savings don't justify the change-management cost
• You're mid-acquisition and stability matters more than optimization
• You have <90 days of operational history on Vapi (give it more time before deciding)

For everyone else who has decided to migrate, this 14-step playbook is the production path.

FAQ

How long does the typical migration take?

6–12 weeks end to end. Dual-run consumes the most calendar time (2–4 weeks).

Can I migrate without engineering help?

For simple deployments (one number, one vertical, no custom tools), yes — CallSphere's onboarding team can guide. For anything with custom Vapi tools, you'll need 1 engineer involved.

What about my Twilio numbers?

Numbers port cleanly. CallSphere's Twilio integration means the underlying carrier relationship can stay intact through the migration.

How do we test parity rigorously?

Use the synthetic call testing pattern in Step 9, plus the dual-run KPI tracking in Step 11. Aim for 95%+ functional parity on your top 30–50 conversation patterns before cutover.

Can I keep my custom tools?

Yes. Custom tools port to CallSphere as webhook tools with the same input/output contract. The shared voice + chat tool surface usually reduces total tool count by 30–50%.

What if cutover goes sideways?

Keep Vapi in standby for 48–72 hours. Roll back at the routing layer if a critical regression appears. Most cutovers don't need rollback if dual-run was thorough.

Will CallSphere help me migrate?

Yes. Enterprise-tier customers get a dedicated CSM and migration engineer. Orchestrate and Custom Build tier customers get structured onboarding. Book a migration scoping call to plan yours.

---

Migration-ready resources: Pricing | Industries | Book a migration scoping call | Contact us

#VoiceAIMigration #VapiToCallSphere #VoiceAIPlaybook #CallSphere #CutoverChecklist