Roman Roads – Launching Autonomous Taxis in Italy

“Please size the total daily taxi-like trip market in Rome and then estimate how many of those trips UrbanAI could realistically capture in the first two years.”

Watch for: Does the candidate structure before calculating? Do they segment residents vs. tourists? Do they sanity-check the output (e.g. rides per vehicle per day)?

OVERALL STRUCTURE

The candidate should use a bottom-up approach, segmenting demand by user type (residents vs. tourists) given their materially different trip frequencies and propensity to use taxi-like services. The total market is then multiplied by an AV-specific capture rate derived from the adoption curve (5% at Year 2).

STEP 1 — TOTAL RESIDENT TAXI-LIKE TRIPS

2.8M people × 2 trips/day × 2% taxi propensity = 112,000 trips/day. Assumption rationale: 2% taxi propensity reflects the mix of public transit dominance in Rome — most residents use the metro or bus for routine trips; taxi-like services are reserved for late-night, luggage, or time-sensitive journeys.

STEP 2 — TOTAL TOURIST TAXI-LIKE TRIPS

30M annual visitors ÷ 365 ≈ 80,000/day. Trips: 80,000 × 4 trips/day × 15% taxi propensity = 48,000 trips/day. Assumption rationale: tourists have higher taxi propensity (15%) — unfamiliar with local transit, carry luggage, and are willing to pay for convenience.

STEP 3 — TOTAL MARKET AND AV CAPTURE

STEP 4 — SANITY CHECK: FLEET UTILISATION

8,000 rides/day ÷ 500 vehicles = 16 rides/vehicle/day. At an average of ~25 minutes per ride, this implies 6.7 hours of active service per vehicle per day — plausible for a first-generation fleet. A well-run ride-hailing operator typically achieves 12–20 rides/vehicle/day. This validates the fleet size assumption.

STRATEGIC IMPLICATIONS

The 160,000 trip/day market is meaningful but highly concentrated: tourists, who represent only ~3% of daily people in Rome, account for 30% of all taxi-like trips. This has a critical go-to-market implication: UrbanAI should prioritise tourist corridors (Fiumicino Airport, Termini Station, Colosseum, Vatican) in its geofenced zone selection to maximise early adoption and unit economics. Resident adoption will be slower to build and should be treated as a Phase 2 objective.

An excellent candidate should:
•    Segment the market into residents and tourists — not treat it as a homogeneous population
•    State assumptions explicitly (2% resident taxi propensity, 15% tourist) with brief rationale
•    Convert the total market to a specific AV capture figure at the Year 2 pilot target (8,000 rides/day)
•    Sanity-check the rides/vehicle/day implied by the fleet size assumption
•    Draw a go-to-market implication from the tourist vs. resident split

Recommended score calculation: starting from 2, decrease by 0.5 for any of the following:
•    Candidate did not segment residents from tourists (treated Rome as one homogeneous demand pool)
•    Candidate did not sanity-check the fleet utilisation implied by the output
•    Candidate failed to connect the sizing result to an AV capture figure at the pilot target
•    Candidate did not state assumptions before calculating
 

PART A — REVENUE AND EBITDA

STEP 1 — ANNUAL REVENUE

8,000 rides/day × €15/ride × 365 days = €43.8M/year

STEP 2 — ANNUAL OPEX

€2.5M + €8.0M + €5.0M + €4.0M = €19.5M/year

STEP 3 — EBITDA

€43.8M – €19.5M = €24.3M EBITDA/year (EBITDA margin: ~55%)

STEP 4 — FLEET CAPEX AND PAYBACK

500 vehicles × €80,000 = €40M total capex. Payback: €40M ÷ €24.3M = ~1.6 years.

PART B — BREAK-EVEN ADOPTION RATE

STEP 5 — REVERSE-ENGINEER BREAK-EVEN RIDES/DAY

At break-even, Revenue = Fixed costs. Revenue per ride-day = €15 × 365 = €5,475/ride/year. Break-even rides/day: €19.5M ÷ €5,475 = ~3,562 rides/day.

STEP 6 — CONVERT TO ADOPTION RATE

3,562 ÷ 160,000 total trips = ~2.2% adoption rate. Given that the S-curve projects ~2.5% adoption by Year 1.5, UrbanAI reaches operating break-even well within its pilot window — a strong signal for investor confidence.

Strategic implication: The break-even threshold of 2.2% is low relative to the pilot target of 5%. This means UrbanAI has substantial margin of safety: even if adoption ramps more slowly than the S-curve projects, the business is operationally self-sustaining within the geofenced pilot zone before reaching its target adoption rate.

An excellent candidate should:
•    Correctly treat fleet cost as one-time capex, not an annual opex charge
•    Derive the correct EBITDA of €24.3M and payback of ~1.6 years without prompting
•    Attempt the break-even reverse calculation without being asked — or do it immediately when asked
•    Connect the 2.2% break-even to the S-curve timeline (achievable within ~18 months)
 

Recommended score calculation: starting from 2, decrease by 0.5 for any of the following:
•    Candidate treated fleet capex as an annual cost (inflating opex and suppressing EBITDA)
•    Candidate did not attempt the break-even reverse calculation
•    Candidate could not convert the break-even rides/day to an adoption percentage
•    Candidate did not draw a strategic implication from the break-even result
•    Candidate made arithmetic errors without self-correcting
 

Solution

An excellent candidate should:
•    Organise risks into genuinely MECE categories (not a flat list of 8+ disconnected points)
•    For each risk, name a specific mitigation — not a generic ‘monitor closely’ or ‘mitigate through planning’
•    Demonstrate awareness that the 18-month competitor window creates urgency — this is not a slow-burn market entry
•    Mention the trust/adoption barrier as a distinct risk from the technology risk

Recommended score calculation: starting from 2, decrease by 0.5 for any of the following:
•    Candidate produced an unstructured list without MECE categories
•    Candidate identified risks without proposing specific mitigations
•    Candidate ignored the regulatory risk (the most material near-term threat to the pilot)
•    Candidate did not connect competitive timing to urgency of launch pace

What distinguishes excellent from adequate:
•    Adequate: names 3–4 risks with generic mitigations (e.g. ‘invest in safety testing’, ‘run marketing campaigns’)
•    Good: MECE structure, specific mitigations, references the geofence constraint and competitor timeline
•    Excellent: MECE structure, specific mitigations linked to case data, proactively prioritises risks by probability × impact, identifies the B2B anchor demand as a risk mitigation lever

RECOMMENDATION

“My recommendation is yes — UrbanAI should proceed with the Rome pilot, and the numbers support a decisive launch rather than a tentative one. The total taxi-like market in Rome is approximately 160,000 trips per day. At just 5% AV penetration — our Year 2 target — we reach 8,000 rides/day, generating €43.8M in annual revenue against €19.5M in opex. That is €24.3M in EBITDA and a 1.6-year payback on the €40M fleet investment. Critically, our operating break-even is at just 2.2% adoption — well below the pilot target, giving us a substantial safety margin. The economics are strong, the investor timeline is tight, and we have a 6-month window on the Stellantis fleet deal. We should move now.”

RISKS & NEXT STEPS

“The one risk that could derail this is not the technology — it is the regulatory environment. The taxi lobby is already active and a national-level political intervention could override Rome’s municipal authorisation. Our first action this week should be locking in a formal municipal operating agreement with binding terms before we deploy any capital. Second, we must geofence Phase 1 exclusively to the widest, most AV-optimised corridors — airport highway, EUR district, main boulevards — to avoid any incident in complex historic-centre traffic that could trigger a shutdown and kill investor confidence. Third, close pre-launch commercial agreements with Fiumicino Airport and at least three major hotel chains to guarantee a baseline of ~3,000–4,000 rides/day from B2B anchors before we open to the general public. That gives us break-even coverage on Day 1.”

An excellent candidate should:
•    Lead with a clear yes/no recommendation — not context or caveats
•    Reference specific numbers from the case: €24.3M EBITDA, 1.6-year payback, 2.2% break-even adoption, €40M capex
•    Name the regulatory risk (not the technology) as the primary threat to the pilot
•    Provide 2–3 concrete next steps with specificity — municipal agreement, geofence selection, B2B anchor deals
•    Complete the recommendation in under 90 seconds

Recommended score calculation: starting from 2, decrease by 0.5 for any of the following:
•    Candidate gave a conditional or hedged recommendation (‘it depends on regulatory clarity’)
•    Candidate could not recall key numbers from the case (EBITDA, payback, break-even)
•    Candidate named technology risk instead of regulatory risk as the primary threat
•    Candidate provided no concrete next steps, or only generic ones (‘monitor the regulatory environment’)
•    Candidate spent more than 2 minutes on the conclusion