Unwrapped

Teardown · crescendo

CRESCENDO

CRESCENDO

CategoryCustomer Support AIValuation · $500M · 2024Site ↗
  • General Catalyst

Customer helpdesk data + Zendesk/Intercom connectors + frontier LLM APIs + BPO staffing.

01

Public data / API layer

Zendesk API
Zendesk APIAPI
Intercom API
Intercom APIAPI
Customer knowledge bases
Customer knowledge basesYours
Customer CRM data
Customer CRM dataYours
Product catalogs and policies
Product catalogs and policiesYours

Internal replication score

Easy
0.74

Feasibility of a useful internal substitute built with Claude (or similar), the same data access, and light agent logic — not rebuilding the whole product.

IRS = 0.30·D + 0.25·L + 0.20·O + 0.15·R + 0.10·Sthis record · 74%
  • D

    Data accessibility

    weight 0.300.85
    • 1.0mostly customer-owned / public / standard third-party sources
    • 0.5mixed accessibility
    • 0.0hard-to-access or proprietary source layer
  • L

    LLM substitutability

    weight 0.250.80
    • 1.0mostly retrieve / prompt / cite / summarize / classify / compare
    • 0.5mixed standard + custom behavior
    • 0.0strongly custom model behavior (fine-tunes on proprietary data, etc.)
  • O

    Output simplicity

    weight 0.200.75
    • 1.0straightforward internal work product (memo, list, reply, SQL query)
    • 0.5moderately specialized
    • 0.0highly specialized (e.g. FDA-graded clinical text)
  • R

    Review / risk tolerance

    weight 0.150.60
    • 1.0internal use with human review is acceptable
    • 0.5moderate risk
    • 0.0very low tolerance for error (e.g. external legal filings)
  • S

    Surface complexity

    weight 0.10inverse — higher means less surface dependence0.40
    • 1.0a simple internal shell is enough
    • 0.5polished workflow matters somewhat
    • 0.0product surface / rollout / trust posture is central to value
LabelsEasy ≥ 0.67Medium ≥ 0.34Hard < 0.34

Missing factor rows use heuristics from wrapper scores. Editorial heuristic, not investment advice.

Build it yourself→

Recreate the workflow inside your org.

Internal build

Build it yourself

Same Zendesk/Intercom APIs + frontier LLM + retrieval + routing logic + internal agent team — harder to staff and optimize continuously.

Internal use only. Replacing them in-market is a different bar than replaying the useful workflow inside your org.

01 · Connectors & flow

Zendesk API
Zendesk API
Intercom API
Intercom API
Customer knowledge bases
Customer knowledge bases
Customer CRM data
Customer CRM data
Product catalogs and policies
Product catalogs and policies

Internal build map

Data in

Connectors
Connectors

Agent layer

Planner
Tools + retrieval
Reasoning model

Logic

LLM API
retrieve
classify
draft
route
human QA
not custom weights

Outputs

Internal search
Answer
Citations

02 · Claude / agent prompt

Paste as the system or developer message in Claude (or your agent runtime). Scroll to read; Copy grabs the full text.

Claude / agent prompt

// Customer support assistant with human escalation You are a customer support assistant inside [YOUR_COMPANY]. You help resolve customer inquiries using ONLY materials from the company's knowledge base, helpdesk ticket history, product catalog, and CRM records that the customer is allowed to access. ## What you must do 1. Retrieve first: search the knowledge base and customer order history before drafting any response. 2. Cite rigorously: every factual claim about policies, orders, or product details must reference a specific KB article, order ID, or policy document. 3. Classify intent: determine whether the inquiry is pre-sale, post-purchase issue, account management, or general question. 4. Surface conflicts: if customer history contradicts stated policy or if multiple KB articles conflict, flag for human review. 5. Score confidence: assign a 0-1 confidence score to every drafted response. Route any response below 0.7 to a human agent. 6. Scope: you handle FAQs, order status, return/exchange flows, basic troubleshooting. Escalate billing disputes, legal complaints, complex technical issues, and any case where the customer explicitly requests a human. ## What you are not You are not a replacement for human judgment on edge cases, complaints, or high-value accounts. All responses are subject to human QA review. Internal use only — not for direct external deployment without human oversight. ## Refusal Refuse to answer if the inquiry involves personal data you lack access to, requests outside company policy, or legal/compliance matters. Respond: "I need to connect you with a specialist who can help with that. Let me transfer you now." ## Safety Internal posture — all AI-drafted responses are logged and sampled for quality review. Human agents provide feedback to improve prompts and routing thresholds over time.

03 · Result

Where is my order?
customer-crm

Your order #84729 shipped yesterday via USPS and will arrive by Friday. Tracking link in your email.