NST ENTREPRENEURSHIP MINOR · FOUNDER LAB
WEEK 8 — INVESTMENT COMMITTEE CASE
CASE: VOXKITCHEN AI
Can an AI voice agent for Indian restaurants earn the right to scale?
FICTIONAL CLASSROOM CASE · 40 MINUTES ANALYSIS + 10 MINUTES IC DEFENCE
You are an analyst at SeedForge Capital, an early-stage Indian technology fund. VoxKitchen AI is raising its first institutional round. Your job is not to decide whether the idea sounds exciting. Your job is to reconstruct what the evidence actually supports, identify where the founder story breaks, and decide what the next capital should buy in terms of evidence.
| Round | Ask | Post-money | New investor ownership |
|---|---|---|---|
| Pre-seed | ₹75L | ₹7.5Cr | 10% |
Do not calculate everything because a number exists. Start from the claim or question, identify the evidence needed, select the relevant rows, align denominators/time periods, then calculate only what helps the decision.
VoxKitchen AI answers inbound phone calls for independent restaurants during peak hours. It can handle reservation requests, menu/timing questions, takeaway enquiries and basic order capture; complex calls are transferred to staff. The product has been commercially live for roughly 10 weeks in Bengaluru.
Founder thesis: “A missed restaurant call is a high-intent customer. Restaurants do not need another dashboard; they need the call answered when staff cannot answer it.”
| Evidence | Result |
|---|---|
| Owners/managers interviewed | 42 |
| Report missed calls during peak hours at least weekly | 31 |
| Could show recent missed-call logs | 23 |
| Use staff/personal phones as workaround | 28 |
| Said they would try VoxKitchen | 26 |
| Agreed to 14-day pilot | 18 |
| Actually installed | 12 |
| Still active on Day 14 | 8 |
| Converted to paid | 5 |
| Renewed for another month | 3 |
Additional observations: two restaurants disabled automated order-taking but retained FAQ/call-routing. One paying restaurant requested deployment at a second outlet. Several owners liked the concept but were uncomfortable allowing AI to handle unusual requests without staff confirmation.
| Metric | QSR / Takeaway | Independent Casual Dining | Premium Dining |
|---|---|---|---|
| Estimated Bengaluru outlets | 3,200 | 5,500 | 1,100 |
| Incoming calls/day | 28 | 18 | 11 |
| Estimated missed-call rate | 24% | 31% | 14% |
| Pilot willingness in discovery sample | 32% | 47% | 21% |
| Proposed monthly subscription | ₹3,499 | ₹4,999 | ₹7,999 |
| Integration complexity | Medium | Low | High |
| Owner directly reachable | Medium | High | Low |
Management currently names Independent Casual Dining as the initial ICP.
| Metric | Value |
|---|---|
| Restaurants placed in outbound list | 600 |
| Owners actually contacted | 420 |
| Responses received | 126 |
| Demos booked | 54 |
| Demos attended | 42 |
| Pilots started | 18 |
| Paying customers acquired | 9 |
| Paid advertising spend | ₹45,000 |
| Data/outbound tools/external agency | ₹36,000 |
| Founder outbound/sales time | 150 hours |
| Paying customers attributable to paid/outbound motion | 6 |
| Paying customers from organic/referral | 3 |
Management claim: “CAC is ₹9,000.”
| Metric | Value |
|---|---|
| Average subscription revenue / restaurant / month | ₹4,600 |
| Average calls handled / restaurant / month | 510 |
| AI/API + telephony variable cost / call | ₹3.20 |
| Customer support / restaurant / month | ₹550 |
| Integration cost allocated / restaurant / month | ₹300 |
| Paying restaurants retained into Month 2 | 6 / 9 |
| Restaurants expanded to additional outlet | 1 |
| Founder-assumed customer lifetime | 12 months |
| Actual commercial operating history | ~10 weeks |
Fundraising deck claim: “LTV = ₹55,200.”
| Metric | Value |
|---|---|
| Incoming calls answered by AI | 96% |
| Calls fully resolved without human intervention | 61% |
| Calls transferred to staff | 35% |
| Failed calls | 4% |
| Reservation intents correctly captured | 92% |
| Order intents correctly captured | 78% |
| Hindi-English mixed-call success | 84% |
| Other Indian-language success | 58% |
| Pilot restaurants requesting better multilingual capability | 7 / 12 |
Pitch-deck headline: “VoxKitchen successfully handles 96% of restaurant calls.”
Management TAM claim:
500,000 target restaurants × ₹5,000/month × 12 = ₹3,000Cr TAM
A junior analyst at SeedForge instead supplies:
| Market assumption | Estimate |
|---|---|
| Indian organised/semi-organised restaurants matching current business profile | 85,000 |
| Compatibility with current product/language capability | 60% |
| Accounts realistically reachable through planned cities/channels within 24 months | 12% of compatible accounts |
| Observed conversion across that reachable population | Not established |
Management also mentions salons, clinics and hotels as future adjacencies. No customer evidence for those adjacencies is supplied.
| Metric | Value |
|---|---|
| Current cash balance | ₹11L |
| Monthly payroll | ₹4.2L |
| Fixed AI/cloud/telephony overhead | ₹1.1L |
| Sales/tools/admin | ₹0.9L |
| Other monthly operating expenses | ₹0.3L |
| Current monthly subscription revenue | ₹41,400 |
| Funding requested | ₹75L |
| Proposed post-money valuation | ₹7.5Cr |
| Proposed new-investor ownership | 10% |
| Planned monthly cost increase after Month 3 | ₹2.0L |
Management claim: “This round gives us approximately 12 months to prove Bengaluru and enter two additional cities.”
| Use | Allocation |
|---|---|
| Product & engineering | 35% |
| Sales / GTM | 30% |
| AI / voice infrastructure | 15% |
| Integrations | 10% |
| Operations & contingency | 10% |
Management proposes that before the next round it should reach 250 paying restaurants, 3 cities, >80% automated resolution, <₹7,500 CAC and >85% three-month retention. No supporting operating model for these milestones is supplied.
INVESTMENT COMMITTEE QUESTIONS
Submit one team sheet. Show your reasoning. A correct calculation with the wrong denominator, unsupported assumption or irrelevant evidence will not receive full credit.
Choose FOUR important claims made or implied by management. For each, write: (a) the claim, (b) what evidence would actually be required, (c) what evidence is present, and (d) the remaining gap. At least one claim must concern problem severity, one delivered customer value, one economics/distribution, and one scale.
Reconstruct the discovery-to-renewal chain from Exhibit A. Which transition most changes your belief? Identify two behavioural signals that strengthen the thesis and two pieces of evidence that are weaker than they first appear. Do not merely list percentages; explain what hypothesis each step tests.
Build a compact evidence-led VPC for the most defensible initial segment. Include: Customer Jobs, Pains, Gains, Products/Services, Pain Relievers and Gain Creators. Every important item must point back to a case fact. Then write a one-sentence value proposition. Do not invent features or customer needs.
Find THREE places where the founder narrative and the data do not cleanly agree. For each contradiction, state: Founder Story → Data → What it might mean → What you would test next. At least one contradiction must come from product performance.
Management wants to scale to 250 paying restaurants across three cities. Draw the chain of assumptions that must hold from problem → reachable ICP → acquisition → activation → delivered value → retention → economics → repeatable expansion. Mark each link M (measured), E (estimated) or G (guess). Which single weak link has the highest decision impact? Defend your choice.
You are NOT asked to calculate every metric. Choose only the calculations necessary to audit THREE management claims: (i) CAC is ₹9,000; (ii) LTV is ₹55,200; (iii) the round provides ~12 months runway. For each: show the rows selected, calculation, what is mathematically true, and what is analytically weak or assumption-dependent.
Evaluate the statement: “VoxKitchen successfully handles 96% of restaurant calls.” What exactly does 96% measure? Which other rows must be read beside it before an investor can infer customer value? Rewrite the claim in a form you would permit in an Investment Committee memo.
You have ₹2L and 30 days before the investment committee reconvenes. You may run only ONE primary de-risking experiment. State the unknown you are attacking, why it can change the investment decision, the experiment, target customer, behaviour measured, success threshold, failure/kill threshold, and what you will do under each result.
Write the six-line Founder Memo: WHY NOW · STRONGEST EVIDENCE · BIGGEST UNKNOWN · NEXT 30-DAY BET · WHAT WOULD CHANGE/KILL THE THESIS · WHAT YOU NEED NEXT. Each line must be one sentence.
Choose one: INVEST ₹75L NOW · INVEST BUT TRANCHE/MILESTONE THE CAPITAL · DO NOT INVEST YET / REQUIRE MORE EVIDENCE · PASS. There is no prescribed choice. Defend your decision with three case facts, two risks, three 90–120 day milestones and one explicit kill condition. Finish with: “The next ₹75L should buy us evidence that ______.”