Calls as BI

Your phones ring all day. Some get answered, some go to voicemail, some get returned, some don't. Nobody on your team can tell you what % of inbound calls turn into revenue, which calls were missed during the lunch rush, or what your customers are actually asking about. Communication analysis — every call transcribed, classified, scored, and tied back to its outcome — turns the most overlooked data source in your business into the most actionable.

The black box you're currently operating

For most SMBs, the phone system is a black box. You know:

• Total incoming call count (maybe)
• Some sense of voicemails received
• Nothing about what was actually said

You don't know: which calls were reservation attempts vs vendor pitches vs complaints. Which calls were answered fast vs let-it-ring. Which calls converted to revenue. Which staff member is great on the phone and which is killing your close rate. Which ad campaign is driving the phone activity. What customers are repeatedly asking that suggests a content or process gap.

Every one of those is answerable data. It just lives in the audio nobody listens to.

What "call as business intelligence" means

The communications stack:

• Telnyx + Twilio for phone numbers, routing, recording
• Deepgram nova-3 for real-time bilingual EN/ES transcription
• Claude-based intent classifier — tags each call as reservation / complaint / question / vendor / sales lead
• Sentiment scoring — how the caller felt entering and leaving the call
• Inside Sales Agent — auto-callback within 90 seconds for any missed inbound during business hours
• Insights / BIE dashboard — weekly call-intel summary tied to revenue outcomes

Net effect: every call is a structured row of data. Searchable, filterable, joinable to your other operating metrics.

Restaurant example: the lunch-rush missed-call problem

A bar & restaurant we work with handles their phones in-house. Calls go to the host stand. Friday lunch peaks 12:00-1:30 PM. Going in, the owner thought maybe 5-10% of calls went unanswered during that window. The actual number was 41%.

Most of those missed calls were reservation attempts for that evening or the weekend. We tied call data to POS reservation outcomes for one month:

• ~67 missed calls per week during lunch rush
• Estimated 35-45% were reservation attempts (industry rule of thumb, validated by sample callback)
• Average party size: 3.2 people; average ticket: $54/person
• Estimated lost revenue: ~$5,400/week or ~$280K/year

The fix wasn't hiring more staff. The fix was deploying the bilingual AI IVR + the auto-callback Inside Sales Agent. Cost: included in the Standard tier at $1,500/mo. Recovered revenue inside the first 90 days: ~$50K. The owner's words after the first month: "I had no idea we were missing that many."

Multifamily example: leasing call quality

A multifamily property gets ~30 leasing inquiries per week. The leasing agent answers them. The owner has no idea how good the agent is at the phone — just that some calls become tours and tours become leases.

Call-intel surfaced the breakdown:

• 30 inquiries/week → 12 tours booked → 4 leases signed
• Of the 18 inquiries that didn't book a tour, 11 ended with sentiment-scored "disengaged" in the last 30 seconds
• Common pattern: agent quoted price up front before establishing fit, caller hung up

The fix was an objection-handling rewrite + a callback sequence for the disengaged calls. Tour booking rate moved from 40% to 67% over the next quarter. At average lease values for this property, that's $40K+/yr in incremental rent on the same inbound volume.

Law firm example: intake quality scoring

Law firm intake calls are high-stakes — one missed or fumbled intake can mean a $5K-$50K case walking to a competitor. Call-intel gives the partner:

• Real-time alert when a high-value intake call is happening (caller mentions specific case types)
• Quality score on every intake call against the firm's intake script
• Auto-escalation if sentiment turns negative — partner can quietly join the call
• Weekly summary tying intake quality scores to case-acceptance outcomes

Privacy + compliance

Two things matter here:

• Disclosure: a recorded-line greeting plays on every inbound call. Compliant with two-party-consent states (California, Florida, etc.).
• PII handling: transcripts are encrypted at rest in Supabase, accessible only to authenticated staff. Customer phone numbers are hashed for analytics; the actual number is only viewable on a per-call detail page.

What it costs

Communication intel is included in the Standard tier ($20K + $1,500/mo). Standalone deployment for operators who already have ASI 360 telephony: ~$3K setup + $400/mo per location for the transcription + classification + dashboard layer. See Software for the underlying engine and Insights / BIE for the dashboard.