Advanced Questions & Troubleshooting

Complex questions from enterprise customers, technical teams, and power users — with detailed answers.

Multi-Agent Architecture

“Should I use one Oppy or multiple?”

One Oppy when: You’re talking to one audience type (e.g., all consumers) and the behavior differences are minimal. One prompt can handle conditional logic like “if they’re a buyer, do X; if they’re a seller, do Y.”

Multiple Oppies when: You’re talking to fundamentally different audiences (e.g., loan officers AND consumers), or you want different phone numbers/emails for different campaigns, or you want to A/B test different prompting approaches.

As models get smarter, one Oppy can handle more complexity. Start with one and split only when needed.

“Can Oppies talk to each other?”

Yes. The traffic controller pattern: one Oppy receives leads and routes them to specialized Oppies. The router can email or message another Oppy to hand off a conversation. Example: lead comes in → router determines it’s a mortgage inquiry → forwards to the mortgage specialist Oppy.

“Can I A/B test different Oppies?”

Yes. Create two Oppies with different prompts/approaches, split your leads between them, and compare metrics. The conversations dashboard at app.oppy.pro lets you filter by Oppy to compare performance.

CRM Integration (Advanced)

“How does the Salesforce integration work without a native connector?”

Several approaches, from simplest to most technical:

Email forwarding — Forward lead notification emails to the Oppy’s email. The AI creates contacts and starts conversations.
Webhooks/N8N — Use N8N or Zapier to send webhooks when lead status changes in Salesforce. The webhook triggers Oppy actions.
API — Post data directly to Oppy’s API endpoints.
HTTP Requests from Oppy — The AI can make HTTP requests to your Salesforce API during conversations (requires admin whitelisting).

For data back to Salesforce:

Needs attention emails → parse with automation tools
Subscribe to Oppy webhook events
Poll Oppy’s API for updates
The Oppy can write back to Salesforce via HTTP requests

“Can the Oppy read Salesforce data during a conversation?”

Yes, if you configure HTTP request access. In the prompt, you can instruct it to call specific Salesforce API endpoints to look up lead data, deal stages, or any other information. This requires the Oppy to have admin-level function access.

“How do we handle state changes between Salesforce and Oppy?”

Recommended pattern: Your system (Salesforce) is the source of truth for lead status. When status changes, your system sends data to Oppy (via webhook or API). Oppy acts on it. When Oppy needs to update Salesforce, it sends data back. Don’t make Oppy responsible for reading Salesforce state — push state changes to it.

Voice Calls (Advanced)

“Can we feed it existing call transcripts to improve its performance?”

Yes, indirectly. Take your a set of your best call transcripts, strip PII, and use Claude/ChatGPT to extract: objection handling patterns, successful closing techniques, key phrases, and qualification best practices. Put those into your Oppy’s prompt as behavioral guidelines.

“Does it say ‘this is a recorded line’?”

Only if you tell it to in the prompt. Common phrasing: “Hi, this is [Name], a digital assistant with [Company] on a recorded line.” Consult your legal team for jurisdiction-specific requirements.

“Can it handle someone who doesn’t want phone calls?”

Yes. In the prompt: “If someone says they prefer text/email over phone, respect that preference and switch channels. Update the contact with their preferred method.”

“What’s the latency on voice responses?”

The target is under 400ms (the natural pause threshold in English conversation). The system uses a brain/mouth architecture — the mouth (voice AI) can keep the conversation flowing while the brain (GPT) processes complex requests. Background sounds (typing, thinking) bridge any gaps.

“Can we make outbound calls to thousands of contacts?”

Yes. Bulk call feature at app.oppy.pro → Contacts → Select → Call Contacts. Set delay between calls. Results export as CSV. One team was making a high daily outbound calling volume.

Data & Privacy

“Where does the conversation data live?”

In Oppy’s infrastructure. Each account is isolated. Conversations, contacts, notes, and knowledge bases are scoped to your account. We don’t use customer data to train other customers’ Oppies.

“Can we export all our data?”

Yes. Contacts can be exported from the Contacts page. Conversations, notes, and other data are accessible via API. For bulk exports, contact our team.

“What happens if we cancel?”

Your data remains accessible for a period after cancellation. Contact our team for data export before your account is fully decommissioned.

Reseller / White-Label

Real Reseller Discussion (Example Digital Agency)

Example Agency is a digital marketing company serving hundreds of real estate clients. Their vision:

Phase 1: Use Oppy internally for their own sales and support
Phase 2: Offer Oppies to their client websites as a managed service
Technical approach: API-based provisioning — loop over client list, create Oppy per client with website URL (auto-crawled), standardized ISA prompt template, custom colors
Client experience: Clients get phone/email notifications when leads qualify. They never need to log into Oppy. The Oppy’s embed code goes on their website automatically.
Pricing model: Example Agency manages everything, charges clients as part of their marketing package. Rev share with Oppy for clients who want more advanced features.

Atlas Software Group / Enterprise Market Discussion

an industry executive (Atlas Software Group, a $10B vertical software company) shared industry insights:

Smallest customers (lower-price-point accounts) are the stickiest — long average tenure, lowest churn
Enterprise clients require the most support — dedicated account managers, frequent hand-holding, high cost to serve
80,000-90,000 small brokerages that nobody is serving because they’re “not worth the sales effort” — but they can make decisions fast (no C-suite, no 6-month evaluation cycles)
Key quote: “The reality is there are loads of small brokerages that have a need and a desperate need because if they want to survive, they’ve got to evolve. But nobody has been able to simplify it to a point where it’s just a no-brainer.”
Industry support ratio: Typically a relatively high support ratio. At eXp, Scott got it to a much leaner support ratio — and that was before AI.

“Can we offer Oppy to our customers?”

Yes. Several approaches:

Managed service — You create Oppies in your account on behalf of customers. They interact via phone/email/chat only (no dashboard access). You manage everything.
Rev share — Customer gets their own Oppy account. You share revenue. We help with setup.
White label — For larger partnerships, we can discuss custom branding.

“How would we deploy Oppies across 500 client websites?”

Via API. A single API call creates an Oppy with: website URL (auto-crawled for knowledge base), prompt (standardized template or AI-generated), phone number, email, and admin contacts. You can loop over your client list and provision in bulk.

“Can clients manage their own Oppy without seeing our dashboard?”

Yes. Give them: the Oppy’s phone number (they can call and give verbal instructions), the Oppy’s email (they can email changes), or a link to the live chat page. The Oppy can update its own prompt based on admin instructions. They never need to log into app.oppy.pro.

Troubleshooting

“The Oppy is making promises it can’t keep”

Add explicit guardrails to the prompt:

NEVER promise to create, deliver, or produce any materials, 
documents, presentations, or physical items. If asked about 
deliverables, explain what's available through the platform 
and offer to connect them with a human team member.

“The Oppy keeps asking for information it should already have”

Tell it to check existing data first:

Before asking the contact for their address, check if it's 
already in their contact record. If you have it, confirm it 
instead of asking again: "I have [address] on file — is that still correct?"

“Notifications aren’t going to Slack”

Verify the Slack channel’s email address is correct
Check it’s added as a Manager on the Oppy at app.oppy.pro
Make sure needs_attention is actually being triggered (check conversation status)
Test by manually setting a conversation to needs_attention

“The knowledge base seems empty after crawling”

Login-required pages show as crawled but contain no useful content
JavaScript-heavy SPAs may not render server-side
Check the actual chunks in the knowledge base to verify content was captured
Try indexing specific pages directly instead of relying on the crawler

“Voice calls sound robotic”

Try a different voice preset — some sound more natural than others
Check if background sounds are enabled (they help with naturalness)
The newer Cartesia voices (Gavin, Leo) tend to sound more natural
Consider cloning a real voice for maximum authenticity

“The AI is too aggressive with follow-ups”

Reduce max attempts in follow-up settings
Add explicit instructions: “If they haven’t responded after 3 attempts, wait 2 weeks before trying again”
Enable “Skip Weekends” and “Respect Business Hours”
Switch to Auto Schedule so the AI picks optimal timing

“Conversations show ‘generating’ for too long”

Normal generation takes 5-30 seconds depending on complexity
If stuck for >5 minutes, the generation may have failed
Check if the conversation has an error status
Contact support if persistent

“My FUB contacts aren’t syncing”

Verify the Owner API key (not a regular user key)
Check that the integration shows as connected at app.oppy.pro
Initial sync can take time for large databases
Look for a fub_id in contact metadata to confirm sync

Performance & Metrics

Representative Metrics

Example Realty Group / your Oppy:

strong website chat engagement from website chat alone
Multiple website chat visitors
Some failed calls (testing)
One contact (the prospect) had an extended single-session exchange

Example Homes Group / your ISA Oppy:

3 positive/engaged responses out of initial outreach
Conversation summaries auto-posted to FUB within 15 minutes
ISA team picks up qualified leads from Slack within minutes

Citywide Realty (Facebook Ads Campaign):

~200 leads at ~$10 CPL
15-20 qualified leads raised to human team (7.5-10%)
Most leads were junk, but AI filtered them automatically

Example Lending Team:

200-300 outbound calls per day
3X improvement in qualified lead conversion
Previously needed 5 human AEs, now handled by AI

Oppy.pro (our website Oppy, 3-year-old Oppy):

Handles all inbound sales inquiries
Books demos on Alex’s calendar automatically
Filters individual agents (waitlist) from enterprise prospects (calendar)
our website Oppy has been running for 3 years — longest-running Oppy

“How do I measure if my Oppy is working?”

Key metrics to track:

Contact rate — What percentage of outreach gets a response?
Qualification rate — How many conversations result in a qualified lead?
Needs attention rate — How often does the AI escalate to humans?
DNC rate — How many people opt out? (High DNC = adjust approach)
Average conversation length — Engaged contacts have longer conversations
Time to needs attention — How quickly are leads being qualified?

Ask the Account Concierge for analytics: “Give me a performance summary for the last 30 days across all Oppies.”

“Can I compare performance between different Oppies?”

Quick Troubleshooting Flowchart

Oppy isn’t responding to contacts:

Check delivery state → Is it “play”? (not pause/stop)
Check conversation status → Is it stuck in “generating”?
Check if contact is tagged dnc → DNC blocks all outreach
Check SMS provisioning → Voice works immediately, SMS needs A2P approval
Check the Oppy’s admin functions → Are the required tools enabled?

Oppy responds but says wrong things:

Check the prompt → Is the instruction clear and unambiguous?
Check knowledge base → Search for the topic, see what chunks exist
Check advanced view → What context did the AI have when it responded?
Check for prompt conflicts → Two instructions contradicting each other?
Update via concierge → Describe the issue, let it fix the prompt

Leads aren’t being notified:

Check managers list → Are email addresses correct?
Check needs_attention triggers → Is the prompt telling it when to escalate?
Check Slack channel email → Is it added as a manager?
Check spam folders → Notification emails might be filtered
Test manually → Set a conversation to needs_attention and see if the email arrives

FUB contacts not syncing:

Verify Owner API key (not regular user key)
Check integration status in settings
For large databases, initial sync takes time
Look for fub_id in contact metadata to confirm sync
Check if the FUB account has API access enabled

Yes. Filter the Conversations page by Oppy at app.oppy.pro. Compare response rates, conversation lengths, and qualification rates. We also provide cohort analytics showing DNC rates at different touch points.

Back to Master Guide