Menteia · RosettaDMS · October 2026 · Confidential
# 24-month roadmap **RosettaDMS** — Connects a car brand with all of its dealers whatever their DMS: through an API or an AI agent that operates it like a service advisor, a central brain that translates everything, an MCP server and voice with Allo.
## Next 8 weeks (before asking for money) What YC recommends doing first: talk to users, charge early and get 10 customers who actually use the product. - Talk to 10 aftersales directors at brands and groups. - Connect a real DMS through computer use and measure reliability. - Sign the first 4-dealer pilot. - Legal review of DMS licenses. Milestones tied to the use of funds. Each quarter has a deliverable the investor can verify. ## Months 1–6: First pilot - Pilot with 1 brand or group (4 dealers, 4 DMS) - 3 DMS mapped with a playbook - Convert the pilot into a contract ## Months 7–12: One brand's network - 40 dealers connected - Voice with Allo in production - JIT parts ordering in use ## Months 13–18: Second brand - Second brand - 90 dealers connected - Quality alerts for the brand ## Months 19–24: Scale - ~150 dealers connected - First customer outside Mexico - Breakeven ## Features that answer competitors' customer complaints Taken from 09_Competencia_y_Quejas: what customers complain about with competitors that we do not fully cover yet. ### Months 1–6 - **Computer-use connector for DMS without API (desktop, AS-400, Citrix)** (partial, high): Demo only with fictitious dealers; has not operated a real DMS or measured reliability. - **Vision-based adaptation and versioned playbooks for screen changes** (partial, high): Designed to answer RPA brittleness, but untested on a real DMS version change. - **Rosetta Schema: common model and per-VIN history** (partial, high): Defined and shown in demo; at least 3 real DMS still need mapping (pilot goal). - **Bidirectional write-back (appointments, leads, parts holds) with validation** (partial, high): Differentiator vs read-only integrations; demo only today. - **Reusable API connectors (REST/SOAP, SFTP, webhooks)** (partial, high): Adapters described in the stack; none connected to a production DMS. - **Sync health monitoring with immediate alerts** (missing, high): Mentioned as risk mitigation but not built; competitors take days to flag stale data. - **Per-DMS legal opinion on automated users and data ownership** (missing, high): Legal review pending; large DMS vendors have blocked and sued third-party accessors. - **Security: credential vault, recorded sessions, revocable access, pseudonymization** (partial, high): Designed (LFPDPPP); no audit or certification (e.g. SOC 2) that an OEM will ask for after the CDK breach. ### Months 7–12 - **MCP server for agents (6 tools)** (partial, medium): Tools defined on the site; not operating on real dealer data. - **Voice and SMS via Allo booking against the real shop calendar** (partial, medium): Integration designed (API, webhooks, MCP), not built; production target months 7–12. - **Export to the OEM data lake and standard formats (STAR)** (missing, medium): The pitch promises data-lake delivery; a concrete connector and output in standards the OEM already uses are missing. - **Resilience to DMS outages (cache, write queue, degraded mode)** (missing, medium): Not described; after the CDK ransomware OEMs value data continuity when a DMS goes down. - **Public per-dealer pricing and contractual SLA** (missing, medium): Pricing only proposed (~USD 300/dealer/month); lack of public pricing is a common complaint about integrators. ### Months 13–18 - **JIT parts ordering and quality alerts in production** (missing, low): Predictive modules on roadmap; need accumulated real data from several dealers. ### Months 19–24 - **Data export for DMS migration on dealer request** (missing, low): Would leverage the common schema to ease DMS switching pain (Tekion vs CDK case). ## Metrics we will report every quarter - Dealers connected and DMS mapped - Onboarding time per dealer - Successful syncs (%) - Appointments booked through RosettaDMS (voice, WhatsApp, MCP) - Calls answered by AI - MRR per brand and churn