Colorado

Service Center

• First State Auto Group owns the Crossroads Hyundai location.
• In 2025, they possess a privacy policy that allows them generous capabilities with marketing. Unsurprisingly, we are signing things like this every day.
• I went to the dealership because of an outstanding service recall, and a check engine light. I did not feel that I had options.
• Within 30 minutes of my signature given, an AI agent purporting to be a human-named employee at Crossroads Hyundai solicited me for a “YES” so that the dealership could buy my car from me.
• The terms are not visible, requiring an “INFO” reply.
• The LLM agent replies instantly to non-Yes/No/Info replies with a second attempt to pressure for a “YES”.
• Requesting employee information from the LLM caused it to go silent.
• Confronting a trio of service center employees with the privacy policy, none could enumerate what the policy allowed for.
• Employees are trained to dodge the question of speaking with the named LLM “rep”. It requires a second question to directly expose it as AI.
• Their privacy policy makes no mention of AI specifically, therefore I pressed to know what else they can do with the terms as defined by them. None could suggest anything more than “marketing”.
• I left this interaction with the feedback that another opportunity to consent to such a service would be an improvement.

Email to Service Center

• A few months later, I planned a subsequent visit to the dealership to follow up on work they had recommended in the previously documented visit.
• On a Friday, I submitted a web request to the Service Center. Not believing that I had a way to see the prior service log, I elected to use the web form to describe what I remembered about the work. I described my needed service incorrectly, and I knew this–I was presenting my errors for them to be corrected.
• Enjoying the opportunity to remind them who I was, I ended the note by requesting that I kindly not be put into an AI privacy nightmare again, because, as I recounted in a summary, an AI had tried to buy my car without anyone knowing.
• Two emails arrived on Monday morning: 1) A human salesman trying to buy my car, and 2) another LLM service bot who had been routed to my spam box two minutes later.
• The AI agent claimed the name “Ashley James” and used the personal pronoun “I” alongside the apparent job title of “Internet concierge”.
• The AI agent repeated my broken description of what service I needed, verbatim. It asked in a kindly way to get started on it.
• My emails with it confirmed its LLM nature as quickly as the auto-reply.

Crossroads Hyundai, Day 1

• I called the dealership, only to find the phone ringing with no answer. It was not closing time, but it was near enough that I could imagine needing to ring them a few times if I wanted an explanation.
• Instead, the connection is dropped after a robot announces that the voice mailbox is full.
• I called repeatedly while I drove to the location before closing time.
• When I asked what phone I was actively ringing in front of them, they did not know. They could not reveal or find out.
• I asked them my real question: Who is Ashley James.
• The employees do not know the term LLM, so I clarify in my request if it is an AI.
• Ashamed like the last batch, they admit it is.
• They have no knowledge of the conversation it had with me. There is no human QA at all.

Crossroads Hyundai, Day 2

• The morning of Tuesday, I arrived again to ask questions.
• They still had no knowledge of the conversation that I had with their supposed representative.
• I ask them if they can in fact access the log.
• He is able. I confirm this without looking at it myself by soliciting him to answer about the contents: he confirms I used vulgarities in my “conversation” with the unmonitored chat ai.
• When asked who names the bots, he reveals “Impel” as the operator of the service.
• I discover that these desk employees represent themselves to me as employees of First State Auto Group. This was not expected, nor does it align easily with declared ownership by 2Autogroup, which also owns the next door Mazda dealership.
• I have notes about who authorizes this decision at Crossroads Hyundai. He was not in the state at the time of my visit. Nor was the human sales agent present who I confirmed had a specific desk nearby with a second busy individual.
• The dealership is a multi-state operation which authorizes the use of the service known as “impel.ai”.
• impel.ai features copious testimonials that are universally doctored. They portray a telling clarification in each, in the vein of “I just love Amber (Sales AI)!”
• All promotional material for this service purports to label its sales assistants, but I do not know if they never actually label them, or if it is up to each dealership.
• However, the doctored labels hang a hat on why the label is needed: The service works so well specifically because the humans believe there is a human power agent who is magically not available to come out and say hello.
• Employee testimonials featured on impel.ai’s materials also clearly indicate that they’re so bummed they have to tell humans that their LLM hoax can’t see them when they close on a deal.

Supplemental:

1. It is not lost on me that Bojangles Michel’s desk ownership is hearsay. We have no first-hand claim with metadata that the desk is anything but a social engineering decoy.

Kansas

7050 W Frontage Rd, Merriam

• It appears the owner of Crossroads Hyundai bought the Reed Hyundai dealership exactly a month ago.
• The reviews of the location preceding this announcement are very bad. My intuition says that the new owner is specifically scouting dealerships with terrible service, and planning to apply the LLM hoax to fix it.
• The new owners also got a Toyota dealership.
• I recorded this interaction.
• My hunch was a bullseye. The representatives here were at first pretending to not engage with the Impel angle, preferring to reduce it instead to “lead generation”.
• As soon as I brought up Impel being equal to AI sales agents, everyone admitted we were talking about the same thing.
• The entire interaction, I was counter-monitored by a representative’s Meta Ray Bans glasses. Should he review his footage, he will note my direct eye contact with the camera next to his brow.
• We were able to discuss the legitimate needs of a high-pressure business to respond to queries quickly and efficiently.
• Trying the question again about who names the AI, a satisfyingly direct answer came forward: “Kate is Kate.” The representatives are expressly interested in treating the LLMs as self-sufficient beings.
• My concerns enumerated to them, despite concessions about business needs, were 1) that Impel’s own testimonials depend on the LLM hoax–the customers are happy because they think there is a person, and 2) there is literally no QA/QC.
• The representatives are explicitly uninterested in monitoring the LLMs. They explicitly perceive the job of an LLM to be unsupervised in the success case.
• I expressed interest a Quality Control initiative. The response to the idea implied I was being idealistic.
• We agreed on much, but we were not concerned with the same things.
• The Meta Ray Bans observed my Hyundai and license plate as I drove away, with the representative feigning interest in the sky while I waved.

impel.ai has not returned requests for a product demo.

Additional contact is being arranged.

You can reach Autumn Ryan to be heard about this subject at [email protected]. Do not transmit sensitive or private information if it’s unsuitable for others to have.