DealerAgent

Add-on · One flat price for the store

AI agents

Six specialized dealership managers share one Command Center and one Work Case — drafting quotes, repair estimates, customer replies, SOP answers, and exception reviews while a person approves every store change.

$299 / month flat for the account — all users & rooftops

The AI agents add-on is a flat $299 a month for the entire account. Every user and every rooftop on your account gets the full six-agent team without per-seat or per-token add-on fees.

AI Command Center · Work case case-1

Riley Skiff

Stock U-26018 · Selling price $64,500 · Linked Repair Order RO-3001

3 of 5 steps approved
Derived next action: Pat Owner (Owner / GM) — Review Karen's case-grounded customer draft (message.draft)

Step 4 inspection · message.draft

Karen: Case-grounded customer draft

“Hi Riley, your quote on stock U-26018 ($64,500) and $486 delivery prep estimate are ready to review whenever you are.”

Cited dealership records
Approved Quote #Q-884 ($64,500) · Approved Estimate #EST-3001 ($486)
Server-side guardrail
Outbound SMS and email routes stay hard-gated; a person reviews and sends.

Immutable audit ledger (3 approved)

  • quote.propose · Pat Owneraud-101
  • estimate.draft · Pat Owneraud-104
  • service_task.create · Pat Owneraud-107

Shared Work Cases

Instead of scattering alerts across four tabs, related work lands on one AI Work Case. Each case links the customer, the stock or customer-owned unit, the Quick Quote, the repair order, assigned staff tasks, message drafts, and cited evidence — exposing one clear Next Action with a named owner and due date.

Human approval & separation of duties

Server-side permissions enforce four risk tiers. Agents read scoped records and run math immediately, log low-risk internal follow-ups with idempotent keys, pause medium-risk CRM writes in Awaiting Approval, and hard-gate outbound texts, emails, financing, and pricing. No agent or requester can approve its own proposal.

Cross-agent handoffs & signal fusion

Sally's approved quote, Sven's pre-delivery or seasonal estimate, Karen's grounded customer message draft, and Jim's cited GM exception stay on the same case timeline. When signals conflict — such as high repair spend paired with a customer text saying they plan to keep their unit — confidence drops automatically instead of pushing unwanted sales outreach.

Server-enforced guardrails

What runs automatically vs. what waits for a person

Permissions are enforced on the server, not just hidden in the interface. Approval cannot be inferred from casual chat text, and every decision writes an immutable audit entry.

  • 1 · Read & analyze

    Runs immediately within role permissions

    Pre-arrival service briefs, unanswered SMS episode checks, model conversion math, and natural-language ⌘K lookups. Confidential dealer cost stays hidden from customer-facing outputs.

  • 2 · Low-risk internal support

    Logged with idempotent replay protection

    Scheduling a rep follow-up after a quote is approved, answering a staff policy question with section citations, or logging an undocumented SOP gap for leadership.

  • 3 · Medium-risk CRM writes

    Halts in Awaiting Approval until a person confirms

    Saving a Quick Quote, writing labor and parts lines to a service estimate, assigning a technician task, or opening a GM exception task. Requesters cannot self-approve.

  • 4 · Outbound & financial gates

    Hard-blocked from silent execution

    Sending customer SMS or email, submitting lender financing, placing supplier POs, quoting sold units, or pricing below dealer costline. Ambiguous names or stock units always pause to ask.

Included in the $299 / month add-on

The six AI managers

Each manager owns a defined department role, reads live dealership records within user permissions, and hands context cleanly across shared Work Cases.

See the AI Command Center in action

Modules