Index
What an AI buying agent does and does not do
How to define the buying policy that the agent must follow
Where the agent can search
How it scores and ranks opportunities
The decision sheet received by the buyer
The example of the two SUVs at the same price
Decisions that remain human
How the agent learns from what is bought and what is not
Separating recommendation from authorisation
Dealcar and the buying agent connected to stock
Frequently asked questions
What an AI buying agent does and does not do
The buying agent automates the most time-consuming part of sourcing: searching, filtering, and preliminary evaluation of opportunities. Instead of the purchasing manager reviewing dozens of advertisements on portals, auctions, and professional platforms to find those that fit their buying policy, the agent does this work and presents a sorted list of opportunities with the necessary data to decide.
What the agent cannot do is equally important. It does not physically inspect the car nor detect undeclared damage. It does not know the vehicle's actual history beyond what available reports show. It does not negotiate with the seller. And it must not hide uncertainty: if data is not available (unknown preparation cost, unverifiable maintenance history), the agent must reflect it as a risk, not as zero in the calculation.
The value of the agent lies in reducing search time, sorting opportunities by estimated profitability, and pointing out risks before the buyer invests time in a visit or negotiation.
How to define the buying policy that the agent must follow
The agent can only search effectively if it knows exactly what to look for. The buying policy is the set of rules that defines which cars enter the radar and which are automatically discarded.
An operational buying policy has five building blocks. The vehicle segment: what type of car the dealership is looking for (compact automatic SUV, mid-range saloon, hatchback under 5 years old). Age and mileage limits: for example, maximum 5 years and 90,000 km. The maximum capital per transaction: how much the dealership can pay for a vehicle. The minimum target margin after all costs: if the expected net margin is less than £1,500, the opportunity does not enter the list. And stock turnover targets: if a car profile systematically takes more than 60 days to sell in that specific dealership, it makes no sense to keep buying that profile unless the margin justifies it.
Check which expenses are part of the real cost of a car in a dealership sale.
Exclusions are also part of the policy: versions that do not sell well in the area, cars with a declared accident history, or suppliers with whom you do not want to operate.
With that policy written and configured in the agent, the search is automatic and the results are relevant from day one.
Where the agent can search
The buying agent can operate across several channels simultaneously without the purchasing manager having to open them manually every morning.
Private individuals requesting a valuation through the dealership's website or a specific form are the highest potential margin channel. The agent collects the car's information, cross-references it with the market, and generates a preliminary valuation before the buyer makes any phone call.
The professional network of other dealers and stock exchange platforms between dealerships allow access to vehicles that have not yet reached the mass auction market. The agent can monitor those networks and alert when an opportunity appears that fits the buying policy.
Professional auction platforms (BCA, Openlane, Veiko, Ayvens Carmarket) have accessible data feeds that the agent can process to filter opportunities before the bidding session begins. Instead of reviewing 200 listings, the buyer receives 10 pre-selected ones with the estimated margin already calculated.
Read also the complete guide to car auctions for dealerships.
Saved customer requests are another valuable input: if there are three leads in the CRM looking for a 2021 Toyota RAV4 with under 70,000 km and the agent finds one available in an auction, the connection between active demand and buying opportunity is immediate.
How it scores and ranks opportunities
The agent calculates an estimated margin for each detected opportunity, including all known costs: purchase price, transport from origin, estimated reconditioning based on the profile and available appraisal, warranty if applicable, and the cost of finance during the expected days in stock.
That gross margin is cross-referenced with two data points specific to the dealership: the historical stock turnover of that vehicle profile in that specific business and the active demand in the CRM for that type of car. A car with a gross margin of £2,500 that takes the dealership 90 days to sell generates a much lower daily margin than a car with £1,800 of gross margin that turns over in 30 days.
The agent ranks opportunities by that return on committed capital, not just by unit margin. This completely changes the ranking compared to a selection made solely on purchase price.
Compare market prices in real time with Dealcar's market radar.
The decision sheet received by the buyer
Each opportunity that the agent presents to the purchasing manager arrives with a structured sheet containing all the information needed to make a decision without having to search for additional data.
The sheet includes the vehicle description with the source where it is available, the asking price, and the estimated total cost delivered to the dealership (price plus transport plus estimated reconditioning), the expected retail price range with comparable cars currently published on portals, the estimated margin in both central and conservative scenarios, the days on market of the comparables (if they have been unsold for a long time, the market is saturated), the reason why that opportunity fits the buying policy, any missing data or identified risks, and the deadline if the opportunity has a decision timeframe (closed-date auction, private seller with other visits scheduled).
With that sheet, the buyer knows in two minutes if it is worth proceeding or dismissing the opportunity.
The example of the two SUVs at the same price
Two SUVs available in different auctions, both at an £18,000 hammer price.
The first has comparables on portals at £21,500, giving an apparent gross margin of £3,500 before costs. But that model at that specific dealership takes an average of 95 days to sell, and finance costs during that period add up to an extra £330. The actual net margin is under £2,500 and the capital is tied up for more than three months.
The second has comparables at £20,500, a gross margin of £2,500. But that profile turns over in 35 days at that dealership, there are two active leads in the CRM looking for exactly that model, and finance costs are £120. The net margin is similar but the capital is freed up in a third of the time.
The agent scores the second car higher than the first even though the unit gross margin is lower. The return on capital is better and the probability of a quick sale is higher. Without the agent, that comparison would require manually cross-referencing several data systems. With the agent, it arrives resolved on the sheet.
Read also when to trade-in, keep, or liquidate stock in a dealership.
Decisions that remain human
The physical inspection of the vehicle is always human. No appraisal report or photo replaces seeing the car in person before committing to a significant purchase. The agent can prepare the questions to ask during the visit and point out aspects that the appraisal does not cover clearly, but it cannot substitute that step.
Check what to look for before bidding in a car auction.
The final negotiation is also human. The agent can suggest a maximum bid price based on the target margin, but the conversation with the seller and the decision on how far to go rest with the buyer.
Exceptions to the buying policy always require human authorisation. If the agent detects an opportunity that does not meet some of the established criteria but looks interesting for other reasons, it can flag it as an exception and present it to the purchasing manager to decide, but it cannot approve it on its own.
How the agent learns from what is bought and what is not
The buying agent improves over time if fed with actual transaction results. For each car bought, the actual total cost (purchase price plus all additional expenses), final sale price, actual days in stock, and whether a discount had to be made from the initial price must be recorded.
Comparing that information with the forecast made by the agent at the time of purchase reveals systematic biases: if the agent overestimates the sale price of certain trim levels or undervalues the preparation cost of vehicles with over 100,000 km, that learning adjusts future calculations.
Rejections also provide information. If the buyer systematically rejects opportunities that the agent scores as good, there is a reason: it could be a supplier they do not wish to deal with, a model known for after-sales issues, or a price range that generates problematic buyers. Recording the reason for rejection allows the agent to learn which signals predict a good rejection before the buyer has to explain it every time.
Separating recommendation from authorisation
The agent can detect, score, and propose. The purchase authorisation is always a human decision with traceability: who approved, on what data, and with what margin scenario.
For standard transactions below a defined price threshold, a single approval is sufficient. For high-value cars, those with an uncertain history, or those representing an exception to the buying policy, a second level of review is advisable.
This separation has two benefits. First, it prevents the agent's speed from turning into financial risk: making quick decisions is good, but without control, it can lead to purchases that are later hard to justify. Second, it creates traceability in the buying process, allowing evaluation of both the agent (are its recommendations accurate?) and the human process (who approved what and with what result?).
Dealcar and the buying agent connected to stock
Dealcar's buying agent operates on DealConnect, the platform's sourcing module. It has access to the dealership's current inventory, historical turnover by vehicle profile, and active leads in the CRM, allowing it to cross-reference the three perspectives that give meaning to a buying decision: what the market wants, what that specific dealership sells well, and what is available right now.
You can see how it works at dealcar.io/deal-connect or request a demo at dealcar.io.
Frequently asked questions
Can the agent bid automatically in auctions?
It is not recommended as a starting point. The agent can prepare the sheet for each lot with the estimated margin and maximum bid price, but the execution of the bid should remain in the hands of the buyer. Automatic bidding without human supervision can result in purchases above the maximum profitable price if the logic is not perfectly configured for all possible scenarios.
How does the agent manage costs that it does not know for certain?
Uncertain costs must appear as ranges, not as zeros. If the estimated reconditioning of a car with damage in the appraisal could vary between £400 and £900, the agent must show the margin in the optimistic scenario (£400 reconditioning) and the conservative scenario (£900), not an invented average. The buyer decides which scenario they want to work with.
What happens if two branches of the same dealership want the same car?
An allocation rule must be defined before conflict occurs: by active demand in each branch's CRM, by available capital, by territory, or by rotating turn. The agent applies the rule and records the reasoning for the allocation so it can be reviewed if there is a disagreement.
Can the agent manage communication with the private seller who wants to value their car?
Yes, for the initial data collection stage. The agent can request the registration number, mileage, general condition, and photos of the vehicle, and generate an automatic preliminary valuation. The final offer and negotiation remain human conversations.
