SB StartupBasket
All ideas
68 /100 VALIDATE Medium complexity

Aprubo — estimate-approval closer for LATAM auto shops

AI advisor that calls and messages your customers on WhatsApp to get the repair approved before the car leaves the bay.

TradeTech SaaS LATAM WhatsApp-firstVoice-firstAI-agentMultilingualSMBCompliance-driven
— views
Evaluation Scores
68/100

VALIDATE

Overall Score

15
Problem
11
Demand
11
Build
11
Distrib.
10
Revenue
6
Time
4
Defense

Aprubo — estimate-approval closer for LATAM auto shops

1. One-liner

AI advisor that calls and messages your customers on WhatsApp to get the repair approved before the car leaves the bay.

2. Trend signal — why now?

The bottleneck in an auto shop isn’t the wrench — it’s the customer saying “sí, arréglalo.” The car sits on the lift, the bay earns nothing, and a service advisor burns hours chasing a stalled WhatsApp thread. That chase just became automatable.

Three things converged in the last 12 months:

  • WhatsApp Business Calling API went GA. Meta launched native business voice calls (July 2025); Telnyx shipped WhatsApp Business Calling to developers in April 2026, letting an AI voice agent place outbound calls over the same WhatsApp number the shop already uses. In LATAM, WhatsApp is where 80% of the conversation already happens — voice on that channel is new.
  • The pain is measured, not guessed. Shops lose ~4 labor hours/day chasing customer approvals; a five-bay shop cut average approval response from 90 minutes to under 20 with proactive follow-up. Declined/forgotten recommended work “disappears permanently” because shops track invoices, not stalled estimates.
  • The market is enormous and micro. INEGI counts 231,678 talleres in Mexico, overwhelmingly microenterprises. A ready-made GTM list already exists as a paid Excel directory of 98,017 shops.

Provenance:

3. The opportunity

Every incumbent workshop tool in Mexico — KIPUP, Mecanica MX, ServitechApp, Garage Master — treats the estimate as a document you send. You build the presupuesto, fire it into WhatsApp, and wait. In LATAM the customer opens it, doesn’t understand why the suspension job is $8,000, doesn’t want to admit they don’t understand, and goes quiet. The shop’s only recovery move is a human advisor calling back — the exact 4-hours-a-day sink.

Aprubo treats the estimate as a conversation you close. The moment a presupuesto is sent, an AI advisor works the thread: it messages, and if there’s no reply, it calls over WhatsApp, walks the customer line-by-line in plain Spanish/Portuguese (“el amortiguador está fundido, por eso vibra al frenar”), answers “do I really need this now or can it wait?”, offers to split the job into urgent-vs-later, and captures a timestamped, recorded authorization. US tools (Bolt On, Tekmetric) do mobile-approval links — but in English, at US price points, with no WhatsApp voice. Nobody in LATAM is closing the approval; they’re all just sending it.

4. Target market

  • Primary customer: Owner-operator of an independent multi-bay taller in Mexico (later Brazil/Colombia) — 3–15 bays, 2–8 staff, no dedicated service advisor. The owner is the advisor between turning wrenches.
  • Why they buy: “El cliente no contesta y el coche ocupa la rampa.” A stalled approval is a dead bay, and PROFECO requires a written, authorized estimate before work anyway — so the shop is already legally obligated to get and keep that yes. Doing it by hand costs half a person.
  • Rough TAM reasoning: 231,678 talleres in Mexico alone. Say 15% are organized enough to run software and have ≥3 bays → ~35,000 serviceable shops in Mexico; multiples of that across Brazil, Colombia, Argentina. Capturing 3,000 shops is a healthy business and a rounding error on the market.
  • Why now for them: WhatsApp voice AI in their own dialect just became buyable off the shelf, and their customers already live on WhatsApp. Two years ago this was a text-only chatbot they’d have ignored.

5. Product sketch (MVP)

  • Estimate hand-off: shop pastes/imports the presupuesto (or connects KIPUP/Mecanica MX via CSV/webhook); Aprubo takes over follow-up from send to yes.
  • WhatsApp AI advisor: answers the customer in Spanish/Portuguese, explains each line item in plain language, handles “why”, “how much”, “can it wait”, and “how do I pay.”
  • Escalate to voice: if no reply after a set window, the agent places a WhatsApp voice call and has the same conversation out loud — for the older, phone-first customer who never reads the message.
  • Split-approve: customer can approve urgent items now, defer the rest; deferred items go into a follow-up queue instead of vanishing.
  • Timestamped, recorded authorization: captures the approval as a PROFECO-defensible record (who approved what, when, transcript/recording attached to the work order).
  • Stalled-estimate recovery: a dashboard of every unapproved estimate with age, value, and one-tap “have Aprubo chase it again.”
  • Owner digest: daily WhatsApp summary — pesos approved today, pesos still stalled, oldest dead bay.

6. AI angle — what’s load-bearing

Remove the AI and this is just another estimate-link tool — which already exists and isn’t the product. The AI is the advisor: it holds a real, back-and-forth negotiation in regional Spanish/Portuguese, understands automotive line items well enough to justify them, adapts when the customer pushes back on price, and does it over both chat and live voice. That’s the exact labor Aprubo replaces — the human service advisor’s follow-up conversation, at 1/20th the cost and available at 9pm when the customer finally reads their phone. Text-macro reminders can’t negotiate; a scripted IVR can’t answer “why.” Only a conversational model can, and only recently on WhatsApp voice.

7. Localization angle

This is a LATAM-first play, not a global tool with translations bolted on:

  • Language/dialect: Mexican Spanish first, then Brazilian Portuguese and Colombian Spanish — automotive slang differs by country (“rines” vs “llantas” vs “aros”).
  • Channel: WhatsApp is the channel; the whole product lives inside it, message and voice. A US-style email/SMS approval link is dead on arrival here.
  • Payment rails: approval flows straight into SPEI/CoDi (Mexico) or Pix (Brazil) so “sí” and “pagado” happen in one thread.
  • Regulatory: PROFECO (Mexico) mandates written, authorized estimates before repair — the recorded approval doubles as the shop’s legal shield against unauthorized-work complaints, the single most common taller dispute. In Brazil the CDC analog applies.
  • Price: a ₹/peso-native tier (~$25–45/mo) works where a $199/mo US tool never lands.

8. Business model — path to $1M–$5M ARR

  • Pricing: $29/mo base per shop (up to N estimates chased/mo) + usage — WhatsApp voice minutes are billed per 6-second increment only when answered, passed through with margin. Premium tier at $79/mo for multi-location and split-pay.
  • ACV: ~$500–700/yr blended (base + voice usage + upsell).
  • Rough math to $1M ARR: ~1,700 shops × ~$590 ACV ≈ $1M. Out of ~35,000 serviceable Mexican shops, that’s under 5% penetration in one country.
  • Rough math to $5M ARR: ~8,000 shops across Mexico + Brazil + Colombia, plus expansion into service-advisor-seat upsell and parts-supplier referral fees. Requires proving retention past month 6 and a working self-serve onboarding.
  • Expansion path: more locations, more monthly volume (voice usage scales with the shop), then adjacent verticals with the same “expensive slot + phone-first customer + mandatory authorization” shape — motorcycle shops, appliance repair, veterinary.

9. Go-to-market wedge — first 100 customers

  • Buy the list, work it by hand. The 98,017-shop datomex directory is a purchasable Excel. Pull the 3,000 multi-bay shops in CDMX/Guadalajara/Monterrey, send each a WhatsApp voice note (in the product’s own voice) demoing Aprubo chasing a fake estimate on their shop name. Meta-demo: the tool sells itself by doing its own job.
  • Ride the incumbents’ gap. KIPUP/Mecanica MX/ServitechApp have the shops but not the closer. Offer a CSV/webhook “approval add-on” and pitch their customers directly in the taller-software Facebook/WhatsApp groups where owners already compare tools.
  • Refaccionaria (parts-store) channel. Parts distributors talk to every taller daily and win when repairs get approved faster. Co-sell through 5–10 regional distributors on a referral cut.
  • Profeco-dispute angle for cold outreach. Lead with “protégete de las quejas por trabajo no autorizado” — the recorded approval is a legal shield, not just convenience. Different hook, same product, cuts through “another software” fatigue.

10. Build complexity — justification

Medium. Off-the-shelf: LLM for the conversation, a WhatsApp BSP (Telnyx/others) for chat + Business Calling, TTS/STT for voice, standard web stack for the dashboard. Custom work is the automotive line-item reasoning quality (must justify repairs convincingly in regional dialect), the send→chat→voice→approval state machine, and light integrations into the top 3 Mexican taller tools. A pair can ship a credible v1 in ~10–14 weeks; the voice-quality tuning and dialect coverage are what push it past a pure API-glue weekend build.

11. Gating checklist

GatePass?Note
Legal in target market✅Aligns with PROFECO’s estimate-authorization requirement; call recording needs consent disclosure (standard).
Ethical — no harm / dark patterns✅The recorded approval protects the customer from unauthorized work too. Must avoid AI upsell pressure — guardrail against pushing unneeded repairs.
Market exists (evidence above)✅231K shops, measured 4hr/day approval-chase pain, existing paid software category.
1–5 person team can build this✅Pair, ~10–14 weeks to v1.
Launchable with <$50K / ₹40L✅APIs + a bought list; main cost is founder time and voice-minute float.

All five pass.

12. Feasibility score

AxisWeightScoreNotes
Problem intensity2015/20Real, daily, quantified (~4 labor hrs/day, dead bays). Painful but the shop has a human workaround, so not full hair-on-fire.
Demand evidence1511/15Strong indirect: measured approval-chase cost, a paid software category, PROFECO mandate. Weak spot: no direct verbatim taller-owner quote begging for this AI closer — inferred, not proven.
Build feasibility1511/15Off-the-shelf APIs; voice quality + dialect + state machine are the honest work.
Distribution clarity1511/15Named list (98K directory), named channels (owner groups, parts distributors). Conversion on cold WhatsApp outreach is unproven.
Revenue mechanics1510/15Pricing sane vs category; but voice-minute pass-through margin and retention are guesses, and micro-shops churn.
Time to first revenue106/10v1 in ~3 months, then pilot period; first paid shop realistically 3–4 months out.
Defensibility104/10Execution + dialect + integration lock-in only. Incumbents with the shops could bolt this on within a year. Speed and a wedge are the moat, not tech.
Total10068/100

13. Qualitative modifiers

Founder-fit tags

sales-heavy (cold list-work, distributor partnerships, owner-group hustle) · technical-heavy (voice-AI quality is the product; a weak agent kills it).

Key assumptions to validate (3–5)

  1. Assumption: Owners will let an AI speak to their customers, not just text. How to test: demo a live WhatsApp voice-call chase to 30 owners; measure how many are comfortable vs “no, mis clientes quieren hablar conmigo.”
  2. Assumption: The AI can justify line items convincingly enough that customers approve, not get spooked. How to test: run 50 real chases in a friendly pilot shop; track approval rate vs the owner’s manual baseline.
  3. Assumption: Shops will pay ~$29/mo + usage, and voice minutes don’t blow up margin. How to test: pilot pricing with 10 shops; measure minutes/estimate and gross margin per shop.
  4. Assumption: The PROFECO-shield framing beats the productivity framing in cold outreach. How to test: A/B the two hooks across 500 directory shops; compare reply rates.

Risk flags

  1. Platform dependency: Total reliance on Meta’s WhatsApp Business Calling API — pricing, policy, or rate-limit changes could reshape unit economics overnight.
  2. Defensibility: Thin moat. A funded incumbent (Mecanica MX, 400+ shops in 5 countries) that already owns the shop relationship could add an approval closer and out-distribute a newcomer.
  3. Trust/ethics: An AI that talks customers into repairs is one bad headline from a Profeco backlash. Must hard-guardrail against upselling and disclose it’s an AI.
  4. Market timing: Micro-shop software adoption in Mexico is real but slow; owner-operators are busy and skeptical of “otra app.”

14. Structured verdict

Score:                  68/100
Verdict:                VALIDATE
Confidence:             Medium
Best-fit builder:       Bilingual (ES) technical founder who can tune voice-AI + a scrappy LATAM sales partner
Time to revenue:        3–4 months (v1 ~10–14 weeks, then paid pilot)
Capital to launch:      $8–15K (APIs, voice-minute float, directory purchase)
Top 3 assumptions to validate first:
  1. Owners accept an AI voicing their customers — 30-owner demo test
  2. AI-driven approval rate ≥ owner's manual baseline — 50-chase pilot in one friendly shop
  3. Voice-minute usage keeps gross margin positive at $29/mo — 10-shop pricing pilot
Kill criteria:
  - Abandon if pilot approval rate is not clearly above the shop's manual baseline after 50 real chases
  - Abandon if <5% of 500 cold-outreach shops reply to the meta-demo
  - Abandon if voice-minute cost per approved estimate exceeds ~30% of the monthly fee at typical volume

15. Next step — 1-week validation sprint

  • Day 1–2: Buy/scrape 300 multi-bay shops from the directory. Hand-build one end-to-end chase (message → WhatsApp voice call → recorded approval) using an off-the-shelf voice agent on a single fake estimate. Get it working in Mexican Spanish.
  • Day 3–4: Recruit 1 friendly taller. Run Aprubo against 15–20 of their real stalled estimates. Track approval rate and time-to-yes vs the owner’s gut baseline.
  • Day 5: Send the meta-demo voice note to 100 cold shops. Go if: the pilot’s AI-chased approval rate beats the owner’s manual baseline and ≥5 of 100 cold shops reply asking for it. No-go if owners recoil at the AI voicing their customers, or approval rate is flat/worse.

Falsifiable outcome: a measured approval rate and a measured cold-reply rate — not “the owner liked it.”

Interested in a detailed proposal?

Get a deep-dive with market research, competitive analysis, and implementation roadmap.

Contact us

info@startupbasket.ai