Index
Informational chatbot vs digital salesperson: what is the difference in practice
The three levels of conversational AI on the web
When is a chatbot enough and when do you need an agent
How the system should behave according to the website section
The discovery conversation: how to guide the buyer in four questions
How to display cars without overwhelming the buyer
What to do when the car is no longer available
What data needs to reach the salesperson
The evolutionary ladder: how to add autonomy in phases
How to measure if the chatbot is generating real value
More than 750 dealerships already use Dealcar for their daily operations
Frequently Asked Questions
Informational chatbot vs digital salesperson: what is the difference in practice
An informational chatbot explains the dealership's opening hours, sends the link to the catalogue and collects the visitor's name and telephone number. It may be well-built and respond fluently, but it does not know what cars are in stock, cannot filter by budget and cannot schedule a visit directly.
A digital salesperson does different things: it understands what car the visitor is looking for, filters the actual inventory in real-time, displays two or three options with a specific reason for each one and offers an appointment in a real slot in the calendar. When the visitor closes the tab, the summary of the conversation (vehicle viewed, budget, part-exchange car) reaches the CRM or the salesperson.
The difference is not in the fluency of the conversation or the prettiness of the widget: it is in what tools the system has connected and with what autonomy it can act.
Also check out how to automate WhatsApp Business in a dealership with AI.
The three levels of conversational AI on the web
There are three levels of capability that are worth distinguishing before evaluating any proposal or buying any product.
The FAQ chatbot responds within a predefined conversation tree. It can resolve hours, location, documentation needed to buy a car or questions about warranties. It does not check stock in real-time nor can it create records in the CRM. It is the simplest level to implement and has the lowest operational risk.
The co-pilot assists the salesperson without acting autonomously. It can consult the stock to prepare a recommendation, draft a response for the salesperson to review and send, summarise a call or suggest the next step. The final decision is always human. It is useful when the team wants speed but does not want to give autonomy to the system.
The agent interprets the intent, consults tools (stock, calendar, CRM), decides the next step within the dealership's rules and executes actions: displays filtered cars, creates an appointment in the actual calendar, adds a note in the CRM, initiates a follow-up via WhatsApp. It requires more integration, more testing and more supervision. In return, it can complete real work without human intervention.
Also check out what AI agents for dealerships are and what they can do.
The same visitor who asks "I want an automatic SUV, I am part-exchanging my car and I can come on Saturday" gets very different answers depending on the level. The chatbot collects the details. The co-pilot prepares three options and a draft response for the salesperson. The agent searches for the available automatic SUVs, requests the details of the part-exchange car, checks Saturday's calendar and books the slot if there is availability within the rules.
When is a chatbot enough and when do you need an agent
The correct question before choosing the level is not "which technology is better?" but "what needs to be done by the end of the conversation?".
If the answer is to inform the visitor and capture their name and telephone number so that the salesperson can call them, a well-configured chatbot can be sufficient. The operational risk is low, the implementation is fast and the maintenance is simple.
If the answer is to identify the visitor, show them the cars that match their criteria in the actual stock, book a visit in the calendar and register the interest in the CRM, an agent level is needed. This level requires integrations with stock, calendar and CRM, clear permissions on what the system can do autonomously, tests before activating it in production and continuous supervision.
A chatbot can create appointments if it has an integrated form. An agent creates them by consulting real availability and blocking the slot. The boundary does not depend on the product's commercial name: it depends on how much it interprets, what tools it uses and with what autonomy it acts.
Most frequently, when a dealership says "the chatbot doesn't work", they have a level 1 chatbot installed expecting level 3 results. The correct product depends on the process you want to automate, not on the name in the provider's proposal.
How the system should behave according to the website section
The chatbot or agent should not behave the same on all pages of the website. The context of the page the visitor is on defines what the most relevant action is.
On a specific vehicle detail page, the visitor already has a car in mind. The system should know which page the visitor is on and start there: "Do you want to check availability, calculate a finance quote or see similar alternatives?" Those three options are more useful than a generic "how can I help you?".
On the stock listing, the visitor is exploring. The system can ask about budget, body type and fuel to filter the results and show the most relevant ones.
On the valuation or "sell your car" page, the visitor wants to know how much their car is worth. The system can open the valuation file by collecting the registration plate and mileage, and explain what the process is from there.
On the finance page, the visitor wants to know what their monthly payment would be. The system can collect the price of the car they are interested in, the term and if they have a deposit, and hand over to the salesperson with that scenario prepared.
On the contact page or the home page, the system can discover budget, usage and body type before recommending.
Page context reduces the number of questions needed and improves the relevance of the response. A visitor who has spent 90 seconds looking at a Seat Arona detail page does not need to be asked if they are looking for a car.
The discovery conversation: how to guide the buyer in four questions
Instead of opening with "how can I help you?", the system can propose four paths from the start: buy a car, sell or value theirs, consult finance or ask a specific question. This initial structure reduces the time to the first useful action.
If the visitor wants to buy, four questions are usually enough to be able to show specific options: approximate budget, main use of the car (city, mixed road, long journeys), preferred body type (hatchback, SUV, estate) and fuel or gearbox preference.
You don't have to ask all four at once. If the visitor is already on an automatic SUV page, the body and gearbox questions are already answered. Only budget and use are needed to refine.
The goal of each question is to reduce uncertainty with the fewest possible turns. If the system already has enough context to show two relevant options, it has to show them instead of keeping on asking.
How to display cars without overwhelming the buyer
A list of 15 cars that fit the visitor's criteria is not a recommendation: it is a catalogue. The visitor has to do the work of comparing by themselves, which is exactly what they were doing before opening the chat.
The system should present two or three options with a specific reason for each. "This Corolla fits your budget and its average consumption is 4.8 l/100 km, the lowest of the three. This León has 20,000 km less, which can reduce the maintenance cost in the coming years." This justification differentiates the digital salesperson from the portal search engine.
The recommendations have to be explainable and honest. The system should not invent equipment that is not in the vehicle details, nor show cars that are not actually available in stock at that moment. If there are two units of the same model with different specifications, it must indicate the difference.
The visitor must be able to refine: "do you have anything cheaper?" or "is there an automatic one?" are natural continuations that the system must handle without starting from scratch again.
What to do when the car is no longer available
The worst mistake of a chatbot connected to an outdated feed is confirming availability of a car that has already been sold. If the visitor arrives at the page of a specific car and the system says "yes, it is available" when in fact it was sold that morning, the disappointment during the visit is hard to recover from.
Also check out how to keep stock synchronised in real-time across digital channels.
When the car is reserved or sold, the system has to say so clearly and propose two alternatives: similar cars available right now, or an alert for when a similar unit comes in. That second option converts a visit without a car into captured demand that the buying team can use to prioritise sourcing.
Feed synchronisation with actual inventory has to be in real-time or with a maximum delay of minutes. A feed updated once a day can generate incorrect confirmations for hours.
What data needs to reach the salesperson
The value of the chatbot is not in the conversation: it is in what it leaves registered in the CRM when the visitor closes the tab. If the system does not transfer useful context to the salesperson, it is just a customer service tool, not a sales tool.
The six pieces of data that must reach the CRM after each conversation are: the vehicle that initiated the interest and the alternatives the visitor viewed, the budget and purchase timeframe if they mentioned them, if they have a car to part-exchange, the questions that remained unanswered, the preferred channel and time for the next contact, and whether an appointment was created or not.
Read also how to manage leads in a dealership to increase sales.
With those six pieces of data, the salesperson can prepare the follow-up call without having to ask again which car the visitor was interested in.
The evolutionary ladder: how to add autonomy in phases
Activating an agent with full autonomy from day one is the most common mistake in conversational AI implementations. The level of autonomy has to grow progressively, validating each phase before moving on to the next.
Read also how to implement AI in a dealership without stopping operations.
The first phase is response and capture: the system answers FAQs and collects the visitor's name and contact details. No stock or calendar integrations.
The second phase adds stock checking: the system can answer availability questions by consulting the inventory in real-time. It doesn't schedule or create records yet.
The third phase adds scheduling with confirmation: the system can propose appointment slots, but the appointment is confirmed only when a salesperson approves it. The system informs the visitor that someone will confirm shortly.
The fourth phase adds direct execution: the system can create the appointment in the actual calendar if the slot is available and within the configured rules, and register the lead in the CRM autonomously.
The fifth phase adds multi-channel follow-up: if the visitor did not convert on the web, the system can initiate a follow-up via WhatsApp the next day with the context of the conversation.
Each phase is activated when the previous one is reliable. Activating phase 4 without phase 3 working well generates double bookings, duplicate records and errors that the team has to resolve manually.
How to measure if the chatbot is generating real value
The number of conversations started or messages sent is not a value metric: it is an activity metric. What measures whether the chatbot is generating value for the business is the complete conversation funnel.
The six indicators that need to be visible are: conversations with buying intent detected (out of total conversations, how many the system identifies as active buyers), vehicle recommendations viewed (how many visitors get to see at least one specific car), contact details shared (name and telephone or email), appointments created, appointments attended, and sales where the chatbot was the first point of contact.
If many visitors ask about availability but do not leave their details, there is a problem in the next step: the response does not generate enough trust or the system asks for too much data at the wrong time.
The correct test is to install the chatbot on the vehicle detail pages and on the stock listing, keep the existing form active in parallel for 30 days, and compare identified leads, appointments and assisted sales between those who interacted with the chatbot and those who did not. The number of messages is not included in the comparison.
Also check out how to use web analytics in a dealership to make decisions.
More than 750 dealerships already use Dealcar for their daily operations
The Dealcar website includes the chatbot integration with real-time stock: when the visitor asks about a car, the system consults the current inventory, not a feed with hours of delay. The leads generated on the web reach the same CRM as leads from WhatsApp and portals, with the context of the conversation linked to the file.
You can see how it works at dealcar.io/web-concesionarios or request a demo at dealcar.io.
Frequently Asked Questions
Should the chatbot appear as soon as the page loads?
Not always. It can be triggered by intent: time on page, comparing cars, search with no results or clicking "help". Activating it immediately on all pages can interrupt a visitor who is still exploring and generate more widget closes than useful conversations.
When does it make sense to ask for the telephone number?
After providing a valuable answer: saving a selection of cars, proposing an appointment or answering a specific question. Asking for details in the first message turns the chatbot into another capture form, which is exactly what the visitor hopes not to find.
How does the chatbot affect the website's SEO?
The chat does not replace crawlable content. Well-structured vehicle pages, category pages and blog content are what rank. The chatbot is an interaction layer for the visitor, not for the search engine. A poorly implemented chatbot that hides page content can indeed harm SEO; one that adds a conversation layer without removing existing content does not.
Can a chatbot create appointments directly?
Yes, if it has calendar integration. The difference with an agent is not whether it can create the appointment, but whether it does so by consulting real availability and blocking the slot, or if it simply sends the data to the salesperson to create it manually.
Index
Informational chatbot vs digital salesperson: what is the difference in practice
The three levels of conversational AI on the web
When is a chatbot enough and when do you need an agent
How the system should behave according to the website section
The discovery conversation: how to guide the buyer in four questions
How to display cars without overwhelming the buyer
What to do when the car is no longer available
What data needs to reach the salesperson
The evolutionary ladder: how to add autonomy in phases
How to measure if the chatbot is generating real value
More than 750 dealerships already use Dealcar for their daily operations
Frequently Asked Questions
Informational chatbot vs digital salesperson: what is the difference in practice
An informational chatbot explains the dealership's opening hours, sends the link to the catalogue and collects the visitor's name and telephone number. It may be well-built and respond fluently, but it does not know what cars are in stock, cannot filter by budget and cannot schedule a visit directly.
A digital salesperson does different things: it understands what car the visitor is looking for, filters the actual inventory in real-time, displays two or three options with a specific reason for each one and offers an appointment in a real slot in the calendar. When the visitor closes the tab, the summary of the conversation (vehicle viewed, budget, part-exchange car) reaches the CRM or the salesperson.
The difference is not in the fluency of the conversation or the prettiness of the widget: it is in what tools the system has connected and with what autonomy it can act.
Also check out how to automate WhatsApp Business in a dealership with AI.
The three levels of conversational AI on the web
There are three levels of capability that are worth distinguishing before evaluating any proposal or buying any product.
The FAQ chatbot responds within a predefined conversation tree. It can resolve hours, location, documentation needed to buy a car or questions about warranties. It does not check stock in real-time nor can it create records in the CRM. It is the simplest level to implement and has the lowest operational risk.
The co-pilot assists the salesperson without acting autonomously. It can consult the stock to prepare a recommendation, draft a response for the salesperson to review and send, summarise a call or suggest the next step. The final decision is always human. It is useful when the team wants speed but does not want to give autonomy to the system.
The agent interprets the intent, consults tools (stock, calendar, CRM), decides the next step within the dealership's rules and executes actions: displays filtered cars, creates an appointment in the actual calendar, adds a note in the CRM, initiates a follow-up via WhatsApp. It requires more integration, more testing and more supervision. In return, it can complete real work without human intervention.
Also check out what AI agents for dealerships are and what they can do.
The same visitor who asks "I want an automatic SUV, I am part-exchanging my car and I can come on Saturday" gets very different answers depending on the level. The chatbot collects the details. The co-pilot prepares three options and a draft response for the salesperson. The agent searches for the available automatic SUVs, requests the details of the part-exchange car, checks Saturday's calendar and books the slot if there is availability within the rules.
When is a chatbot enough and when do you need an agent
The correct question before choosing the level is not "which technology is better?" but "what needs to be done by the end of the conversation?".
If the answer is to inform the visitor and capture their name and telephone number so that the salesperson can call them, a well-configured chatbot can be sufficient. The operational risk is low, the implementation is fast and the maintenance is simple.
If the answer is to identify the visitor, show them the cars that match their criteria in the actual stock, book a visit in the calendar and register the interest in the CRM, an agent level is needed. This level requires integrations with stock, calendar and CRM, clear permissions on what the system can do autonomously, tests before activating it in production and continuous supervision.
A chatbot can create appointments if it has an integrated form. An agent creates them by consulting real availability and blocking the slot. The boundary does not depend on the product's commercial name: it depends on how much it interprets, what tools it uses and with what autonomy it acts.
Most frequently, when a dealership says "the chatbot doesn't work", they have a level 1 chatbot installed expecting level 3 results. The correct product depends on the process you want to automate, not on the name in the provider's proposal.
How the system should behave according to the website section
The chatbot or agent should not behave the same on all pages of the website. The context of the page the visitor is on defines what the most relevant action is.
On a specific vehicle detail page, the visitor already has a car in mind. The system should know which page the visitor is on and start there: "Do you want to check availability, calculate a finance quote or see similar alternatives?" Those three options are more useful than a generic "how can I help you?".
On the stock listing, the visitor is exploring. The system can ask about budget, body type and fuel to filter the results and show the most relevant ones.
On the valuation or "sell your car" page, the visitor wants to know how much their car is worth. The system can open the valuation file by collecting the registration plate and mileage, and explain what the process is from there.
On the finance page, the visitor wants to know what their monthly payment would be. The system can collect the price of the car they are interested in, the term and if they have a deposit, and hand over to the salesperson with that scenario prepared.
On the contact page or the home page, the system can discover budget, usage and body type before recommending.
Page context reduces the number of questions needed and improves the relevance of the response. A visitor who has spent 90 seconds looking at a Seat Arona detail page does not need to be asked if they are looking for a car.
The discovery conversation: how to guide the buyer in four questions
Instead of opening with "how can I help you?", the system can propose four paths from the start: buy a car, sell or value theirs, consult finance or ask a specific question. This initial structure reduces the time to the first useful action.
If the visitor wants to buy, four questions are usually enough to be able to show specific options: approximate budget, main use of the car (city, mixed road, long journeys), preferred body type (hatchback, SUV, estate) and fuel or gearbox preference.
You don't have to ask all four at once. If the visitor is already on an automatic SUV page, the body and gearbox questions are already answered. Only budget and use are needed to refine.
The goal of each question is to reduce uncertainty with the fewest possible turns. If the system already has enough context to show two relevant options, it has to show them instead of keeping on asking.
How to display cars without overwhelming the buyer
A list of 15 cars that fit the visitor's criteria is not a recommendation: it is a catalogue. The visitor has to do the work of comparing by themselves, which is exactly what they were doing before opening the chat.
The system should present two or three options with a specific reason for each. "This Corolla fits your budget and its average consumption is 4.8 l/100 km, the lowest of the three. This León has 20,000 km less, which can reduce the maintenance cost in the coming years." This justification differentiates the digital salesperson from the portal search engine.
The recommendations have to be explainable and honest. The system should not invent equipment that is not in the vehicle details, nor show cars that are not actually available in stock at that moment. If there are two units of the same model with different specifications, it must indicate the difference.
The visitor must be able to refine: "do you have anything cheaper?" or "is there an automatic one?" are natural continuations that the system must handle without starting from scratch again.
What to do when the car is no longer available
The worst mistake of a chatbot connected to an outdated feed is confirming availability of a car that has already been sold. If the visitor arrives at the page of a specific car and the system says "yes, it is available" when in fact it was sold that morning, the disappointment during the visit is hard to recover from.
Also check out how to keep stock synchronised in real-time across digital channels.
When the car is reserved or sold, the system has to say so clearly and propose two alternatives: similar cars available right now, or an alert for when a similar unit comes in. That second option converts a visit without a car into captured demand that the buying team can use to prioritise sourcing.
Feed synchronisation with actual inventory has to be in real-time or with a maximum delay of minutes. A feed updated once a day can generate incorrect confirmations for hours.
What data needs to reach the salesperson
The value of the chatbot is not in the conversation: it is in what it leaves registered in the CRM when the visitor closes the tab. If the system does not transfer useful context to the salesperson, it is just a customer service tool, not a sales tool.
The six pieces of data that must reach the CRM after each conversation are: the vehicle that initiated the interest and the alternatives the visitor viewed, the budget and purchase timeframe if they mentioned them, if they have a car to part-exchange, the questions that remained unanswered, the preferred channel and time for the next contact, and whether an appointment was created or not.
Read also how to manage leads in a dealership to increase sales.
With those six pieces of data, the salesperson can prepare the follow-up call without having to ask again which car the visitor was interested in.
The evolutionary ladder: how to add autonomy in phases
Activating an agent with full autonomy from day one is the most common mistake in conversational AI implementations. The level of autonomy has to grow progressively, validating each phase before moving on to the next.
Read also how to implement AI in a dealership without stopping operations.
The first phase is response and capture: the system answers FAQs and collects the visitor's name and contact details. No stock or calendar integrations.
The second phase adds stock checking: the system can answer availability questions by consulting the inventory in real-time. It doesn't schedule or create records yet.
The third phase adds scheduling with confirmation: the system can propose appointment slots, but the appointment is confirmed only when a salesperson approves it. The system informs the visitor that someone will confirm shortly.
The fourth phase adds direct execution: the system can create the appointment in the actual calendar if the slot is available and within the configured rules, and register the lead in the CRM autonomously.
The fifth phase adds multi-channel follow-up: if the visitor did not convert on the web, the system can initiate a follow-up via WhatsApp the next day with the context of the conversation.
Each phase is activated when the previous one is reliable. Activating phase 4 without phase 3 working well generates double bookings, duplicate records and errors that the team has to resolve manually.
How to measure if the chatbot is generating real value
The number of conversations started or messages sent is not a value metric: it is an activity metric. What measures whether the chatbot is generating value for the business is the complete conversation funnel.
The six indicators that need to be visible are: conversations with buying intent detected (out of total conversations, how many the system identifies as active buyers), vehicle recommendations viewed (how many visitors get to see at least one specific car), contact details shared (name and telephone or email), appointments created, appointments attended, and sales where the chatbot was the first point of contact.
If many visitors ask about availability but do not leave their details, there is a problem in the next step: the response does not generate enough trust or the system asks for too much data at the wrong time.
The correct test is to install the chatbot on the vehicle detail pages and on the stock listing, keep the existing form active in parallel for 30 days, and compare identified leads, appointments and assisted sales between those who interacted with the chatbot and those who did not. The number of messages is not included in the comparison.
Also check out how to use web analytics in a dealership to make decisions.
More than 750 dealerships already use Dealcar for their daily operations
The Dealcar website includes the chatbot integration with real-time stock: when the visitor asks about a car, the system consults the current inventory, not a feed with hours of delay. The leads generated on the web reach the same CRM as leads from WhatsApp and portals, with the context of the conversation linked to the file.
You can see how it works at dealcar.io/web-concesionarios or request a demo at dealcar.io.
Frequently Asked Questions
Should the chatbot appear as soon as the page loads?
Not always. It can be triggered by intent: time on page, comparing cars, search with no results or clicking "help". Activating it immediately on all pages can interrupt a visitor who is still exploring and generate more widget closes than useful conversations.
When does it make sense to ask for the telephone number?
After providing a valuable answer: saving a selection of cars, proposing an appointment or answering a specific question. Asking for details in the first message turns the chatbot into another capture form, which is exactly what the visitor hopes not to find.
How does the chatbot affect the website's SEO?
The chat does not replace crawlable content. Well-structured vehicle pages, category pages and blog content are what rank. The chatbot is an interaction layer for the visitor, not for the search engine. A poorly implemented chatbot that hides page content can indeed harm SEO; one that adds a conversation layer without removing existing content does not.
Can a chatbot create appointments directly?
Yes, if it has calendar integration. The difference with an agent is not whether it can create the appointment, but whether it does so by consulting real availability and blocking the slot, or if it simply sends the data to the salesperson to create it manually.
