Agents: Omni

Overview

Omni is a dealer-facing, secure GPT environment built by Fullpath. Unlike public LLMs, Omni is trained on the dealership’s own data, integrates with the dealership Customer Data Platform (CDP), and uses the Model Context Protocol (MCP) to leverage dealership agents (inventory, incentives, OEM offers, Shopper data, etc.). Omni uses what's called Deep Thinking, which means it has enhanced context for more accurate nuanced responses as well as:

• Ability to draw intelligent insights and recommendations
• Can handle multiple queries simultaneously and step-by-step

Omni is accessible via the Fullpath Dashboard and is positioned to preserve customer privacy, reduce per-user costs compared to individual accounts, and improve team performance while minimizing PII leakage risk.

Technical eligibility

Available for Fullpath Dealer customers who have CDP Pro.

Roles and permissions

Permissions: Omni will be available for Admin, Edit, or Read Only profiles. 

• Omni can talk about any products that are enabled on the store account and which the profile using Omni has access to.

• Omni cannot access group level metrics (at this stage)

   

Navigation

Omni will be included in the CDP Pro package and therefore will not sit as a separate product in the Dashboard. Omni will be enabled by default on all live CDP Pro accounts and will be accessible via the Dashboard side menu across all pages.

Omni interface highlights

• Familiar GPT-style chat window.
   
• Upper-left controls:
  • New chat / Quick start
  • Bookmark and save important chats for later
     
• Chat features:
  • Attach files to a conversation
  • Use supported toolsets (e.g., code interpreter, web/file search) if enabled
  • Can support up to 5 uploads per message
  • File types supported: txt, Csv, json, pdf, png,jpg, jpeg

  • File Size supported: up to 5MB each file
     

     

Privacy, security, and PII handling

Omni is privacy-preserving and trained on dealership data to reduce reliance on public LLMs. It is designed to prevent PII leaks and protect customer information when used correctly. 

Omni Data and MCPs

Omni has several built-in Model Context Protocols (MCPs) or context bundles that it can use to ground its responses in dealership-specific data. We're constantly adding more MCPs to the mix; some typical MCPs include (but are not limited to):

1. Dealership operating information (address, hours, contact details)
2. Dealership inventory
3. Current incentives and OEM offers
4. Dynamic Payments (if enabled on the store / group)
5. Shopper, customer profiles and behavioral data from the Fullpath CDP
6. Activation performance and metrics

Omni Use Cases

A sample of suggested prompts:
 

Tier 1: Executive / GM Questions (High-Impact Decisions)

Performance & Pacing

• "How are we pacing month-to-date versus the same point last month? Can you also help explain what is driving the increase or decrease by looking at lead sources, traffic sources, shopper behavior, and any other relevant contributing factors?"
• "What is our projected month-end sales/gross if the current pace continues?"
• "Where is the store leaking in the lead-to-sale funnel? Are we converting enough leads into appointments and shown appointments?"
• "Which lead sources or vendors are actually producing sales?"

Inventory & Demand

• "Do we have enough inventory for the vehicles shoppers are demanding? Which makes/models have demand but not enough stock?"
• "Which aged or promoted VINs need attention? Are VINs Acceleration units getting more VDP views after promotion?"
• "Which models are getting the most engagement?"

Store Health

• "How is my store doing today?" (morning briefing: sales, inventory, service, CRM workload)

Tier 2: Marketing Director Questions (Campaign ROI)

Traffic & Conversion

• "Which traffic sources are driving website conversions?"
• "How are my traffic sources performing relative to their benchmarks?" (close rate vs industry standard by source)
• "Perform a website audit" (traffic quality, engagement paths, drop-off analysis)

Channel Performance

• "How did paid media perform this month by channel and objective? What did we spend, and are we on pace with budget?"
• "Which channels are efficient or inefficient by CPC, CPL, CTR, and conversion rate?"
• "Which campaigns should we scale, pause, or reallocate budget from?"
• "Are we getting enough VDP views from paid campaigns? What is the estimated CPL after accounting for calls and lead events?"

Audience Activation

• "Are Audience Activation campaigns producing opens, clicks, leads, and sales? Which AA campaigns are causing unsubscribes?"
• "Is unsubscribe rate spiking for any campaign, store, or time period?"
• "Which shoppers were reactivated or influenced by marketing?"

Tier 3: Sales Manager / BDC Questions (Team Efficiency)

Funnel & Follow-Up

• "Which lead sources are producing appointments, shown appointments, and sold units?"
• "Where are leads getting stuck between lead received, appointment set, appointment shown, and sold?"
• "Are appointment show rates improving or declining?"
• "Which CRM leads need better follow-up? Are close rates different by source, vendor, make, or campaign?"

Prioritization

• "Which shoppers should sales prioritize right now?"
• "Who's in the sales lane today that I might be able to upsell to?" (equity + recent activity)
• "Show me five people to call today who were active yesterday on VDPs but haven't submitted a lead"

Tier 4: Compliance / CX Questions (Risk Mitigation)

• "Which campaigns are causing unsubscribes? Are unsubscribe rates unusually high?"
• "Are SMS/email opt-outs increasing? Are communication preferences being respected?"
• "Which campaigns risk list fatigue? Are shoppers receiving too many messages?"

Tier 5: Data Quality / CRM Ops Questions (System Health)

• "Are emails, phone numbers, addresses, VINs, and deal numbers usable? Where do CRM and DMS records disagree?"
• "Which clients or stores have data quality problems? Are shopper records merging correctly?"
• "Are opt-in and opt-out statuses accurate? Are lead sources mapped consistently?"