AI-salesperson is changing the nature of a commercial website
AI-salesperson is changing the nature of a commercial website
For a long time, a commercial website was designed as a self-service storefront. It could be very good: a clear catalog, strong product cards, convenient filters, reviews, recommendations, a cart, and checkout. But all of this belonged to the platform itself. A full-fledged salesperson inside the website was typically absent.
This was not an accident, but a consequence of web economics.
Most websites grew around already existing incoming traffic—primarily search traffic. A user formulated a query, found the necessary page, and arrived at the website with a certain intent. Therefore, for decades, web development learned mainly not to interfere with formed demand: to show the product, provide information, help compare, and accept payment.
In such a model, the differences between types of retail spaces were almost erased. A website was thought of simply as a "store website," and conversion as a task of UX, content, and marketing.
However, physical retail does not work this way. A kiosk in a train station passage, a brand boutique, a neighborhood grocery store, and a specialized roller skate shop are different retail environments. They require different salespeople, different qualifications, and different service economics.
For the first time, an AI-agent makes it possible to bring this distinction back to the web.
The website and the salesperson are now two different parts of a single internal system
Once a visitor arrives, two entities exist within our system:
The marketplace — the entire passive apparatus of the website: assortment, price, navigation, product cards, content, trust, forms, payment, delivery, interface, and purchase rules.
The AI-salesperson — an active subject who works with a specific person: notices their state, asks questions, builds hypotheses, explains, adapts behavior, and guides the interaction toward a commercial result.
Therefore, a "selling website" and a "website with a salesperson" are not the same thing. A landing page, a quiz, recommendations, and personalization can be very effective tools, but they are not salespeople in themselves. This boundary is separately examined in "A selling interface is not a salesperson".
The historical reason why websites remained self-service points for so long is discussed in detail in "A commercial website grew out of a self-service model based on incoming demand."
External traffic and internal sales must be separated
SEO, advertising, brand, social media, recommendations, and partner channels answer the question: who came and how much it cost to bring them. This is the external contour—the digital equivalent of choosing a location for a store.
After entry, a different contour begins: what the marketplace itself and the salesperson operating within it do with the visitor who has already arrived.
These contours are economically connected, but they must not be mixed. Poor internal conversion cannot be endlessly treated by purchasing new traffic, and a good salesperson cannot be credited with audience quality created by advertising or SEO.
Read more: "External traffic and internal sales are different contours".
Why websites need to be distinguished by storefront type again
When there was no salesperson inside the website, the typology of the storefront could be almost ignored: the user mostly interacted with a self-service interface.
With the advent of the AI-salesperson, this becomes a mistake. You first need to understand what retail situation the website itself creates, and only then choose the agent's behavior.
The working model identifies four basic types of commercial websites as marketplaces:
- Ready-demand transactional site. The visitor already knows what they want; the main task is to conduct the transaction quickly and cheaply.
- Weak-demand capture platform. Traffic exists, but purchase intent is weak; attention must be turned into interest and a transaction.
- Repeat-relationship platform. The primary value is created through customer return, memory, and LTV.
- Complex-choice expert platform. The buyer wants to solve a problem, but is unable to independently and reliably choose a product or service.
This typology and its connection to salespeople are analyzed in detail in "Four types of commercial websites require different AI-salespersons."
AI-salespersons also have different roles
We use four working archetypes:
- Scoop / Unloader — cheaply and quickly serves already formed demand;
- Hunter — turns weak intent into a transaction;
- Owner — builds repeat relationships and optimizes LTV;
- Expert — reduces uncertainty and the risk of a complex choice.
They do not form a "bad → good" scale. These are different production roles. An Expert can be a wonderful salesperson and at the same time economically meaningless on an exact-SKU website. A Scoop can be a weak consultant, but an ideal cheap serving agent where the buyer has already decided everything.
Read more: "Four archetypes of a salesperson: Scoop, Hunter, Owner, and Expert".
The main question is the match between website and salesperson
Now you need to design not the "smartest agent," but a pair:
marketplace type × AI-salesperson type.
Basic compliance looks like this:
| Website type | Main task inside the storefront | Basic salesperson role |
|---|---|---|
| Ready demand | Do not interfere and quickly checkout | Scoop / Unloader |
| Weak demand | Create interest and lead to a deal | Hunter |
| Repeat relations | Maintain trust and increase LTV | Owner |
| Complex choice | Reduce uncertainty and risk of error | Expert |
This does not mean a rigid, permanent binding. A single website can contain multiple retail situations, and a single AI-agent can switch roles. But for each situation, you need to understand what function is required and how much it should cost.
The cost of the salesperson is part of the website's economics
The most powerful AI-salesperson is not always the best one.
If a person has come to buy a precisely known model, an expensive Expert may simply turn into a very expensive cash register. If the product is complex and a choice error is costly, a cheap Unloader saves on inference and at the same time loses much more gross profit.
Therefore, a salesperson should be chosen not by maximum intelligence, but by economic return. Read more: "The cost of the salesperson must match the retail location".
The salesperson must prove their incremental effect specifically
You cannot attribute all purchases after communicating with AI to the agent's merit. Some people would have bought even without them.
The right question is:
how much more does a website with a salesperson earn compared to the exact same website without a salesperson?
You need to account for growth in conversion, average check and LTV, reduction in returns and human workload—and subtract the cost of the model, infrastructure, errors, and extra friction.
Read more: "An AI-salesperson must create incremental profit".
Why this architecture is becoming important right now
Organic traffic is becoming less guaranteed, and paid acquisition of relevant visitors is getting more expensive. Therefore, businesses face two parallel tasks.
The first is to expand their own attention capture surface: create additional thematic resources, knowledge bases, and touchpoints with the audience.
The second is to increase the efficiency (ROI/throughput) of the traffic already acquired: improve both the marketplace itself and the salesperson inside it.
This connection is analyzed in "Increasing traffic costs raise the value of internal efficiency".
One of the options for expanding the external surface is "Thematic advisory websites increase their own attention capture surface". At the same time, a thematic advisor and a salesperson inside a specific store are different roles; see "AI-consultant and AI-salesperson are different commercial roles".
What changes in the very nature of a commercial website
Before AI, a developer had to pre-code most of the buyer's journey: which page to show, what filter to provide, what text to write, where to put a CTA, which form to open.
Now part of this behavior can be executed dynamically by a salesperson. The agent is capable of deciding for themselves what to clarify, what to explain, which option to exclude, when not to interfere, and when to actively intervene.
Therefore, a commercial website ceases to be merely a statically designed self-service point. It becomes a retail environment within which a dedicated adaptive salesperson operates.
This is precisely the central shift of the entire model.
Reading map
If you only need the big picture, this concept is enough.
For deeper reading:
- A commercial website grew out of a self-service model based on incoming demand
- A selling interface is not a salesperson
- External traffic and internal sales are different contours
- Four types of commercial websites require different AI-salespersons
- Four archetypes of a salesperson: Scoop, Hunter, Owner, and Expert
- The cost of the salesperson must match the retail location
- An AI-salesperson must create incremental profit
- Increasing traffic costs raise the value of internal efficiency
- Thematic advisory websites increase their own attention capture surface
- AI-consultant and AI-salesperson are different commercial roles
An applied version for an independent specialist can be found at solopreneur.prof. A futuristic interpretation of the shift is at futurist.expert.