CRM
Records, not relationships.
AI-powered relationships, sales execution, employee development, and performance intelligence, connected through one shared ecosystem.
Jessi's real-time voice, vision, AI-written replies, AI call grading and AI mystery shopping run in I'm On Social today (controlled beta). Every capability on this page is labelled with its true status.
Most customer-facing businesses run four systems that never learn from each other. The CRM stores records, the phone holds conversations, training happens elsewhere, and mystery shopping arrives as a report weeks later.
Records, not relationships.
Conversations stay on devices.
Occasional, generic practice.
A report, weeks later.
What a customer said on a call never reaches the next text.
No system knows what matters to this customer right now.
Leaders rarely hear real conversations, so training guesses.
Every tool brings its own seats, admin and integrations.
franchised light-vehicle dealers in the US in 2025: the first market, all relationship-driven.
NADA Data 2025for full-duplex AI voice by API, the engine behind voice practice and AI shoppers.
OpenAI GPT-Live-1 list price, Oct 2026price spread between frontier and small AI models, so routing by task decides margins.
OpenAI, Anthropic, Google pricing, Oct 2026VI Ventures is building a connected AI ecosystem that helps businesses capture relationships, execute follow-up, improve employee performance, and measure results. Each product keeps its own brand and workflows, and all four draw on the same shared intelligence.
Every contact, card, photo, call and conversation becomes structured context.
Illustrative example. The engines are LIVE inside I'm On Social; the cross-product loop is PROPOSED.
"Every relationship has context. Every interaction creates intelligence. Every opportunity deserves appropriate follow-up."
Current versus future
No capabilities in this group for I'm On Social.
"Relationship intelligence and sales execution in one place, with practice and coaching built in."
Current versus future
No capabilities in this group for TracTok.
"Practice the Conversation. Perfect the Performance."
Current versus future
No capabilities in this group for TokTrac.
"Where Performance Meets Potential."
Current versus future
No capabilities in this group for MVP Line.
The personal AI assistant that works on behalf of each salesperson or business professional. Jessi remembers relationships, understands conversations, searches notes, prepares follow-up and recommends what to do next, always in the rep's own voice and permissions.
Illustrative. Cited answers and approve-before-send are LIVE today; cross-customer semantic memory is PROPOSED.
Drafts and replies in the rep's own voice, built from a voice onboarding interview, with banned words honored.
Full-duplex voice on GPT-Live-1: interrupt her, ask about any customer, have her draft or remind.
Reads business cards, documents, VIN plates and odometers; photo intelligence on texted pictures is in limited rollout.
Voice memos, saved facts and cross-customer search today; time-aware fact engine built; semantic memory proposed.
Anything that reaches a customer or cannot be undone waits for the rep's yes.
"Your 3", hot opportunities and nudges today; autonomous campaign preparation proposed.
Explore the shared platform. Select a capability to see what exists today and which products use it, or pick a product to highlight what it draws from the shared layer.
Cross-cutting: tenant isolation · consent ledger · approval policies · audit log · retries and idempotency · observability · cost metering.
This reuse already exists inside the I'm On Social codebase: one grader serves call scorecards, practice calls and mystery shops.
Every agent action follows one path: request, plan, scoped tools, policy check, approval, execute, audit.
Read, summarize and answer with citations.
Drafts and proposals a person reviews.
Send, book or enroll after an explicit yes. The default for anything customer-facing.
Pre-approved routines with limits, audit and undo.
Today: Jessi's customer-facing actions wait for the rep's yes (LIVE). A central policy engine and an approval inbox across products are PROPOSED.
Research current to October 2026. VI Ventures stays vendor-neutral: every AI call passes through one gateway that routes by measured quality, meters cost and falls back across providers.
Million-token context and strong agentic coding are now standard at the top tier; mid-tier models cost about a fifth as much.
Examples: OpenAI GPT-6 Astra and GPT-6.1 Sol; Anthropic Claude Fable 5.1, Opus 5.5, Sonnet 5.5; Google Gemini 4 Argon
For VI Ventures: Campaign planning, coaching analysis, rubric design, executive insight
High-volume classification and extraction now cost cents per thousand calls.
Examples: GPT-6 Luna; Claude Haiku 5.5; Gemini 3.8 Flash and 3.5 Flash-Lite
For VI Ventures: Intent detection, tagging, routing, summaries, nightly intelligence
Full-duplex voice with natural interruptions is available by API at flat per-minute prices.
Examples: OpenAI GPT-Live-1 ($0.05/min) and gpt-realtime-2.1-mini; Google Gemini Live; ElevenLabs, Deepgram, Cartesia
For VI Ventures: Jessi Live, TokTrac practice, MVP Line shoppers
Managed agent runtimes add sandboxes, schedules, memory stores and approval policies.
Examples: Claude Managed Agents; OpenAI Agents SDK and Responses API; Google Gemini Enterprise Agent Platform and ADK
For VI Ventures: Campaign preparation, research, scheduled insight jobs with approvals
Images, documents, audio and video are read natively by frontier models.
Examples: GPT-5.4 vision (in use), Claude and Gemini multimodal models
For VI Ventures: Business cards, documents, vehicle photos, texted photos
Vector search now runs inside operational databases with automatic embeddings.
Examples: MongoDB Atlas Vector Search with Voyage AI automated embedding (GA Aug 2026); agent memory stores
For VI Ventures: Relationship memory with citations, privacy-scoped by tenant
Agents can operate websites, but direct APIs remain safer and cheaper where they exist.
Examples: OpenAI computer use; Claude computer use; Gemini computer use (preview)
For VI Ventures: Last-mile CRM or DMS steps where no API exists, always with approval
MCP is the open standard for connecting agents to tools, now governed by the Linux Foundation's Agentic AI Foundation.
Examples: MCP spec 2026-07-28 (stateless core, Tasks, OAuth alignment)
For VI Ventures: Expose the platform to authorized agents; connect CRMs and calendars
Scheduled agents and event triggers let AI surface opportunities before anyone asks.
Examples: Scheduled agent deployments; event-driven pipelines
For VI Ventures: Cooling-relationship alerts, coaching priorities, store risk signals
Why routing by evidence matters. In September 2026 the I'm On Social team moved customer texts to a newer flagship model. It drafted blander messages three to five times slower, and it was rolled back the same day. Model choice is decided by evaluations, not release dates.
Model names, availability and prices checked against vendor sources on October 16, 2026.
Subscriptions, usage-based voice practice, evaluation programs, enterprise licensing and white-label partners, with cross-sell built into the loop and one shared platform carrying the expensive capabilities. No customers or revenue are claimed here; only I'm On Social pricing is published.
Per-seat subscription
Published$100 per month individual; $75 per user per month for teams, 5-seat minimum
Per-seat subscription with a performance tier
IllustrativeNo pricing published
Per-learner subscription plus voice-practice minutes
IllustrativeSite says Request Pricing
Per-location evaluation program or per-shop pricing
IllustrativeNo pricing published
Enterprise license, white-label and partner revenue share
IllustrativePartner and white-label modules exist in I'm On Social; terms illustrative
AI cost of a 10-minute voice practice session at list prices
AI cost of one Jessi customer draft, depending on model tier
per user per month: the published I'm On Social team price
MVP Line opens the door without a software rollout, TokTrac fixes what it finds, and I'm On Social or TracTok runs the day. One platform team maintains voice, evaluation, memory, compliance and billing for every revenue line.
| Typical category coverage | Capture and engage | Personal AI assistant | Practice | Evaluate | Closed loop |
|---|---|---|---|---|---|
| Dealership CRMs | Yes | No | No | No | No |
| Conversation intelligence | No | No | No | Yes | No |
| AI role-play training | No | No | Yes | No | No |
| Mystery-shopping firms | No | No | No | Yes | No |
| VI Ventures ecosystem | Yes | Yes | Yes | Yes | Yes |
Examples of categories: dealership CRMs, conversation intelligence, AI role-play training and traditional mystery shoppers. Individual vendors vary.
Evaluate with MVP Line, train with TokTrac, execute with I'm On Social or TracTok, then measure again.
Each new AI capability ships once and lifts every product.
Jessi writes and speaks as the individual rep, inside their permissions.
Real-time voice, vision and AI grading already run in I'm On Social.
Opt-out and quiet hours are enforced today; consent capture and per-dealer sender compliance are built and being connected.
Automotive, equipment, home services, real estate and more share one engine.
What exists today, the next 90 days, and the 12-month build. Everything after "Foundation in place" is a proposal.
VI Ventures is moving beyond individual software applications toward an interconnected AI-powered business ecosystem. The opportunity is not simply to automate more tasks. It is to help businesses build stronger relationships, execute consistently, develop better employees, and continuously improve performance.
Jessi remembers every relationship with cited answers, prepares campaigns for one-tap approval and flags cooling relationships early.
A branded execution and development workspace on shared services, where each rep's day mixes customer work and targeted practice.
A global practice platform in English and Dutch. Adaptive training targets each employee's weakest skill, by voice in the browser.
A client console running AI shops across phone, text, email and leads, with benchmarks and remediation through TokTrac.
One intelligence. Four products. One loop.