CRM: Technical Overview of Agentic CRM

Overview

The Agentic CRM is a dealer-facing sales management layer built on a real-time ingestion model. Rather than relying on batch imports, leads and appointments enter the system immediately and trigger instant updates to the customer record, the task queue, and the AI agent workflows that manage outreach.

The system is built around three core AI components that work in tandem:

• The Task Builder agent — analyzes lead signals and shopper activity to generate a prioritized task queue for each sales rep.
   
• The Lead Handling agent — manages outbound follow-up for new leads (default: first 13 days, configurable per dealership playbook), operating autonomously until a human rep intervenes or the lead responds.
   
• Omni — a conversational AI assistant available across the CRM and CDP that has read-only access to live dealership data, customer records, inventory, and performance metrics.

The AI operates as a feedback loop. When a rep changes a suggested vehicle of interest, marks a task as Won't Do, or intervenes in a Lead Handling conversation, these actions are treated as training signals that improve future task generation and outreach recommendations.

The architecture enforces a permissions-first access model. Visibility and editing rights are governed by role mappings configured in the admin portal. The CRM also acts as a proxy for several external integrations — for example, forwarding requests to Fullpath Desking for deal management.

Roles and permissions

The CRM supports three roles, each with a distinct default experience:

Role assignment is managed through the Fullpath admin portal. Administrators can use the role toggle in the portal to switch between management and rep views for troubleshooting and training purposes.
 

Under the hood, these roles are bundles of resource permissions assigned per user profile. Each CRM resource is gated at one of four levels — Read, Create, Update, or Delete — so a profile can be given view-only access to a page or full edit rights. The manager-versus-rep distinction comes down to one thing: a profile with Read access to the other-profiles resources sees every rep's records, while a profile without it is scoped to its own. Landing page and section visibility follow from the same permissions (for example, Read on Overview versus Home).
 

Page and feature access (Read controls whether the page or section is visible and reachable):

• OVERVIEW and HOME — the two landing pages.
• LEADS, SALES, and CRM_ACTIVITY — the Event Logs pages.
• REPORTS — the Reports section.
• AGENTS — the Agents section.
• APPRAISALS — the Appraisals page.
• DEALS_DESK — the Deals Desk page.
• LIVE_WEBSITE_ACTIVITY — the Live Website Activity view.

Cross-rep visibility (controls whether a user sees other reps' records, not just their own):

• OTHER_PROFILES_LEADS — other reps' leads, and the Assignee column/filter in the Lead Pipeline and Leads views.
• OTHER_PROFILES_CUSTOMERS — other reps' customers.
• OTHER_PROFILES_TASKS — other reps' tasks.
• OTHER_PROFILES_APPOINTMENTS — other reps' calendar appointments.
• OTHER_PROFILES_ACTIVITIES — other reps' activity records.

Integrations and dependencies the CRM relies on:

• DMS — customer record import, sold/lost status signals, and appraisals data.
• ADF/XML lead feeds — inbound lead ingestion.
• OEM lead providers — OEM lead and event ingestion.
• Fullpath CDP / Audience Activation — shopper identity, website-tracking and behavioral data, and lead creation from shopper records; this layer supplies the data Omni and the Task Builder rely on.
• Fullpath core platform — underlying data import and synchronization.
• Inventory feed — vehicle of interest matching and inventory-based task generation.
• SMS / Twilio — outbound SMS via the task panel and the Lead Handling agent.
• Fullpath Desking (Transact) — the Deals Desk flow.

Customer records

What is a CRM customer?

In the CRM, the primary entity is a customer record — not a shopper. This is an important distinction:

• A shopper is the platform-wide identity record in the Fullpath CDP. It is built from behavioral data — website visits, VDP views, form submissions, email engagement — across all Fullpath products.
   
• A customer is the CRM-specific operational record. It is the entity a sales rep works from: it holds the lead history, task queue, appointment records, communication log, appraisals, and deal data for that person.

Every CRM customer is backed by a customer_id, which is the primary identifier used across all CRM operations. When a customer also exists as a shopper in the CDP, the two records are linked — but they serve different purposes and different teams.
 

If a person exists only in the DMS and has no prior CDP footprint, they will still appear as a customer record in the CRM. The CRM does not require a CDP shopper profile to exist.

How customer records are created

Customer records are created in three ways:

1. DMS import — When the DMS integration is active, customer records are pulled in from the dealer's DMS. These records bring in contact details, vehicle ownership history, and prior transaction data. DMS customers are searchable and manageable in the CRM immediately on import.
    
2. Lead ingestion — When a lead arrives — via ADF/XML feed, OEM lead provider, a website form, or a manual showroom entry — the system creates or matches a customer record automatically. If a matching record already exists (by email or phone), the lead is attached to that customer. If no match is found, a new customer record is created.
    
3. Manual creation — Any CRM user with Edit access can create a customer manually from the Home page quick action or the customer search bar. The rep fills in the required fields and the record is created immediately.

Core system flows

Lead intake

When a lead arrives in the CRM:

• The system ingests the lead via the real-time ingestion model — no batch delay.
• The lead is matched to an existing customer record by email or phone. If no match exists, a new customer record is created.
• The lead is assigned to a sales rep based on the dealership's assignment rules (configured in the admin portal).
• The Lead Handling agent evaluates the lead and, if active, begins its outreach sequence automatically.
• The Task Builder generates an initial task for the rep and places it in their queue.
• The lead appears in the rep's Lead Pipeline under New Lead.

Task execution

The rep's primary workspace is the Tasks page. The workflow is:

• Rep opens a task row — the task panel expands with the full context, AI-generated rationale, and a suggested outreach script.
• Rep reviews the suggestion, optionally changes the action type (call → SMS, for example), edits the message, and sends.
• Clicking Send from the task panel automatically logs the communication to the customer's Activity and Conversations tabs, marks the task as Done, and updates the lead's last activity timestamp in the pipeline.

No manual status updates are required. The system handles the logging and status change on send.

Lead Handling agent

The Lead Handling agent manages autonomous outreach for new leads. Its behavior is governed by two independent settings in AI Conversation Preferences, accessed from the Conversations page:

• AI Reply — the agent automatically responds when a customer sends a message in (reactive).
• AI Follow up — the agent proactively initiates outreach to customers who haven't responded, following the cadence configured in the agent's settings, within the follow-up window (default: first 13 days, configurable per dealership). This is what drives the outbound follow-up sequence.

The two settings are linked: enabling AI Follow up also enables AI Reply, and AI Reply cannot be turned off while AI Follow up is on (turn off AI Follow up first). AI Reply can run on its own.

Other behavior:

• Sends follow-up messages by email and SMS based on lead source and dealership playbook settings.
• Does not initiate follow-up on 6 federal holidays (New Year's Day, Memorial Day, July 4th, Labor Day, Thanksgiving, Christmas) but will still reply to inbound messages received on those days.
• Hands off to the rep automatically when the customer books an appointment, or when the lead replies and the rep takes over.

When the agent is actively handling a lead, an AI Handling tag appears in the Lead Pipeline. Reps should not send manual outreach while this tag is visible — doing so can disrupt the agent's sequence. To take over a lead, open the conversation and turn off AI Follow up first (then AI Reply as well if you want to silence automated replies).

Appraisals flow

The appraisals workflow manages trade-in valuations within the CRM.

How it works:

• A rep or manager initiates an appraisal from the customer record or the New Appraisal quick action on the Home page.
   
• The rep completes the appraisal form — vehicle details, condition, mileage, and any additional notes.
   
• On submission, the CRM creates the appraisal record and assigns it an upload status.
   
• Where an appraisal-provider integration is configured for the dealership, a background job uploads the appraisal to that provider, and the status on the record updates to reflect the result of the upload.
   
• The appraisal record is visible on the customer's Appraisals tab and in the CRM Appraisals list page (/crm/appraisals).

Deals Desk flow

The Deals Desk in the Fullpath CRM is a visibility layer only — all deal creation and editing happens in Fullpath Desking (powered by Transact).

How it works:

• A rep or manager initiates a deal from the customer record using the Desk a Deal action.
   
• The rep selects a vehicle of interest. Fullpath packages the customer and vehicle data and sends it to Transact.
   
• The deal is built and managed entirely within Fullpath Desking.
   
• Deal stage and status sync back to the CRM, where the deal appears in the customer's Deals tab and in the Deals Desk list page.

Fullpath Desking generates a short-lived exchange token to authorize the handoff between systems. Managers can view active deals from the Overview page's Active Deals section, which surfaces leads currently in the Proposal stage along with their current deal status.

Management oversight

Managers use the Overview page as their primary performance dashboard. It provides:

• Key Insights — AI-generated highlights and recommendations for the store, refreshed regularly.
   
• Top Stats — incoming leads, scheduled appointments, appointments shown, vehicles sold, and total gross profit for the selected date range.
   
• Performance vs. Pace widget — projects end-of-month performance against manually set unit and gross goals, based on current pace (calculated using business days in the current month only; no historical pattern analysis).
   
• Pipeline Snapshot — a live count of all leads across the six pipeline stages for the entire store.
   
• Active Deals — a snapshot of active deals (leads currently in the Proposal stage), each shown with its current deal status; the full list is available on the Deals Desk page.
   
• Upcoming Appointments — a snapshot of the next upcoming appointments across all reps.

Reports are restricted to manager-role users for the initial beta phase. The Team Performance report provides a pivot table across metrics including active leads, total sales, close rate, appointment show rate, and average response time.

AI decision transparency

Every task created by the Task Builder includes a full explanation of why it was generated — the specific lead signals, shopper behavior, and contextual factors that triggered the recommendation. This explanation is:

• Visible in the task panel when a rep opens a task.
• Logged in the Task Activity Log on the customer's Activity tab.
• Preserved permanently, so managers can review AI reasoning for any past task.

Important to note

• "Customer" vs. "shopper" — use "customer" when referring to CRM records and the people sales reps work with. Use "shopper" only in the CDP context. These are linked but distinct entities.
   
• Group-level access is not supported in the current beta. The CRM operates on a single-rooftop basis. Multi-store group functionality is coming soon.
   
• Service data is not currently included in the Version 1 CRM release. Service appointments, DMS service records, and service leads are not part of the pipeline, event logs, task generation, or reporting in V1. Service support is planned for a future release.