Automotive AI Agents Graduate From Chatbot to Coworker

When the Dealership's AI Starts Closing Deals and Cutting Checks

At its first Tekion One conference, held June 15 to 17 at Resorts World Las Vegas, Tekion put AI agents in charge of dealership work that used to require a person at a desk. The new agents engage buyers on price and availability, catch the missing paperwork that stalls a finance deal, and read, code, and help pay vendor invoices. Tekion calls the coordinating piece T1 Pro, which it describes as "a unified AI interface and agent orchestration layer" inside its Automotive Retail Cloud. The sell has moved from software that answers questions to software that does the job.

The distinction matters here, because it changes what a dealer is buying. A chatbot responds. An automotive agent acts: hand it an instruction like "engage this lead and book the appointment" or "code and pay these invoices," and it returns finished work instead of a draft. Over the past year the AI automotive pitch shifted from chat to action, and this launch is the most complete version of that move so far. Founder and chief executive Jay Vijayan, the former Tesla chief information officer who started the company, put it bluntly: "AI isn't a trend car buyers are waiting on. It's already shaping how they shop, decide, and transact."

The dealership just hired a few agents

The news here is the coordination, more than any single bot. T1 Pro sits over a dealership's data, its customers, deals, repair orders, inventory, and accounting, and routes work to specialized agents that can each take an action rather than only surface an answer. Three are new. Salesperson AI handles lead engagement, responding on pricing, availability, and vehicle details, then pushing toward a booked appointment. F&I Manager AI reviews a deal before it leaves the desk, catching missing documents, incomplete checklists, and compliance gaps that delay funding. Accounts Payable AI reads an invoice, recommends how to code it, matches it against a purchase order, flags anything that does not reconcile, and routes it for approval. They join three agents Tekion shipped earlier, Scheduler AI, Technician AI, and Service AI, all reachable through the same interface.

Put together, the suite lets a manager point at a whole automotive workflow, "qualify this lead and set the appointment" or "process this batch of invoices," and expect the work done. That is a different promise from the lead-routing and FAQ bots dealers have leaned on for years, and it carries different risk. A bot that drafts an email is forgiving. An automobile agent that books a customer, edits a compliance file, or moves a payment is operating where a mistake has a cost attached. The capability is the easy part now; deciding how much latitude to give it is the work.

Agents that spend money, not just talk

The most consequential part of the launch sits in accounting. Alongside the Accounts Payable agent, Tekion tied in Brex to embed corporate cards and spend management, branded Tekion Spend, directly inside its platform. The combination lets an agent both code an invoice and move the money to pay it, with card issuance, category and amount limits, and automatic reconciliation handled in the same system.

The manual cost this targets is real. Tekion cites dealer groups processing more than 100,000 checks a year, many of them to settle invoices smaller than what it costs to cut the check. Automating that is where the easy soft-dollar savings live, and it is also where an automated automotive back office stops being a convenience and starts touching cash. Once an agent can approve and pay, the control issue shifts. It is no longer only whether the agent's answer is correct. It is who set the spending limit, who can review what the agent did afterward, and what stops a misfire from writing a real check to the wrong vendor.

Brex chief business officer Art Levy described Tekion Spend as putting "real-time spend controls" and "automated reconciliation directly into the platform dealers use every day." That instinct is sound: spend controls belong next to the agent doing the spending. The harder part is making those controls legible to the people who carry the risk, the controller who has to close the books and the finance director who has to answer for a compliance file.

What the showroom floor already knows

A launch is a promise. The dealers already running AI are a better guide to what these agents will and will not do, and their reports are more mixed than any keynote.

Take Drive Point Auto Group, which runs seven dealerships across Ohio and started leaning on AI in 2025. Partner Kyle Pisani uses it to compile reports across sales history, incentives, and inventory, and to analyze recorded management calls for coaching. In one case the software surfaced that a store was collecting payment before repair orders were closed, an error Pisani says he would not have thought to go looking for. The upside is concrete. One tool the group uses, Drive Solutions, flagged up to $80,000 a month in missed finance-and-insurance profit at some stores, money sitting in deals the team had simply not worked.

Much of that value comes from unglamorous work: pulling clean data out of the dealer management system, the CRM, and the digital-retailing tools, then making sense of it. Drive Point also runs Vinessa, Cox Automotive's customer-communications assistant, and Fullpath, a customer-data platform Cox rolled out in June, to keep that information in one place. The lesson for a dealer evaluating Tekion's agents is that an agent is only as good as the records it reads.

The same dealers are candid about where AI breaks, and it is rarely the model itself. The constraints they name are data quality, workflow fit, and the handoff between an agent and a human. Pisani's preference is telling: "I would rather see AI handle the whole thing or a human handle the whole thing." A customer interaction that an agent half-finishes and then dumps on a salesperson, with no context carried over, can be worse than one no agent touched. His other recurring note is that vendors need to become more "dealer-esque," to learn how a store actually runs before assuming a dealership will reshape itself around an AI product.

None of that argues against agents. It sets the conditions under which they pay off: clean data, a process the agent genuinely fits, and a handoff that passes along the context, not just the conversation.

Before you hand over the keys

Strip away the branding and the dealership-AI decision in 2026 comes down to four things a general manager or controller can carry into any pitch.

First, orchestration. Can you coordinate several agents and route work between them, including back to a person, the way T1 Pro coordinates Tekion's agents? A single capable bot is useful. A system that knows which agent handles what, and when to stop and ask a human, is what makes the output dependable.

Second, spend control. Before an accounts-payable agent ever cuts a check, can a human set per-agent budgets, spending caps, and an approved-vendor list? Embedded cards make payment easy. The discipline is deciding, in advance, what an agent may spend and on whom.

Third, auditability. Is every agent action recorded, down to the request and the response, so accounting and finance can show exactly what happened and why? An agent that acts without a reviewable record is a compliance gap waiting to be found.

Fourth, portability. Are you committing to one vendor's platform and one underlying model, or can the dealership keep its own choice of agent and model and wire it into the systems it already runs? Tekion's path, deep integration inside a single automotive cloud, is a legitimate one, and for a store that wants everything in one place it may be the right call. The trade-off is being tied to one company's roadmap and one model's limits.

That second path is where a horizontal option fits. AgentPMT, a model-agnostic iPaaS for AI agents, is built around it: run whatever agent and model you prefer, then govern it with a budget system that sets per-agent spending limits, tool and vendor restrictions, a real-time audit feed you can filter down to full request and response payloads, and human approval steps you can drop anywhere in a dealership workflow. Tekion and a platform like that are not the same kind of product. One is a vertical system tuned for automotive retail; the other is horizontal, meant to put the same controls around any agent a dealership chooses to run. A dealer weighing the agentic pitch does not have to accept the closed version to get orchestration and oversight.

For years, most talk about artificial intelligence and cars pointed at the windshield: self-driving, robotaxis, the autonomous future. The change that arrived this month is quieter and closer to the books. Agents that engage a buyer, clear a finance file, and pay a vendor are shipping now, and the open work for dealers is governance, not capability. Before you let an agent talk to a customer or pay a supplier, settle three things: who sets its budget, who can audit what it did, and whether you are comfortable being tied to one vendor's platform to get there. Decide that first, and the agents start to look like the coworkers the marketing promised. Skip it, and you have handed the keys to something you cannot yet see clearly.

Sources

• Tekion Unveils New AI Agents and T1 Pro, Its AI Interface and Agent Orchestrator, Business Wire (via Yahoo Finance)
• Tekion deploys 4 new additions to agentic AI suite, forms partnership to help dealers manage expenses, Auto Remarketing
• Adapting AI to auto dealership operations is a work in progress, WardsAuto