How to build a knowledge base for a website AI consultant

An AI consultant on the website can answer customers around the clock, explain services, help choose the right option and transfer a potential client to the manager.

But the quality of his answers directly depends on the information about the company.

If the knowledge base is prepared correctly, the AI employee understands what the company does, what services it provides, what conditions apply and what issues require human participation.

Let's figure out what information should be prepared before connecting an AI consultant and how to do it without complex technical work.

What is a knowledge base for an AI consultant

The knowledge base is a set of information that AI uses to respond to site visitors.

• Description of the company;
• List of services;
• Characteristics of goods;
• Prices and tariffs;
• Terms of delivery;
• Payment methods;
• Guarantees;
• Answers to frequently asked questions;
• Work rules;
• Documents and instructions.

In fact, this is information that the manager usually uses when communicating with the client.

The AI consultant accesses this data and uses it during the dialogue.

What documents can be used

It is not necessary to create a special document to begin with.

Most of the necessary information usually already exists inside the company.

Site pages

The following are often already published on the website:

• Services;
• Prices;
• Contacts;
• Working conditions;
• Information about the company;
• Answers to popular questions.

This is a good basic source of information.

Price lists

If the cost of services or goods is contained in a separate price list, it can also be used as a data source.

It is only important to make sure that the prices are up to date.

Commercial offers

Commercial offers often contain useful information:

• Composition of the service;
• Stages of work;
• Terms;
• Additional options;
• Advantages of the company.

FAQ

If managers constantly answer the same questions, it makes sense to collect these answers separately.

For example:

• How long does the delivery take?
• Is there a guarantee?
• Can I pay by bank transfer?
• Do you work with legal entities?

Such questions are especially useful for training an AI consultant.

How to structure information correctly

AI is also able to work with plain text, but well-organized information usually gives more accurate answers.

The best option is to divide the data by topics.

About the company

Briefly describe:

• What the company does;
• In which cities does he work;
• How many years has it existed;
• Which clients he works with.

Services

For each service, it is desirable to specify:

• Name;
• What is included;
• Estimated cost;
• Terms;
• Restrictions;
• Additional conditions.

Payment

Specify:

• Available payment methods;
• Whether prepayment is required;
• Whether payment by invoice is possible;
• Whether you work with VAT.

Delivery

If the company sells goods, add:

• Delivery regions;
• Cost;
• Terms;
• Transport companies;
• Self-delivery conditions.

Guarantee

Separately fix:

• Warranty period;
• What it applies to;
• Which cases are not guaranteed.

Do not upload conflicting data

One of the most frequent problems is different versions of the same information.

For example, the website has one price, and the old commercial offer has another.

Or in one document the production period is 10 days, and in the other - 20.

Before connecting the AI, it is advisable to check such discrepancies.

If there are several variants of conditions, it is better to explain them explicitly.

For example: "The standard production period is from 10 to 15 working days. For non-standard products, the term is calculated separately".

This format is much more understandable for both the client and the AI.

Do I need to download all the company information?

No.

An AI employee should be given only the information that is needed for his role.

For example, an online store consultant does not require internal accounting documentation.

And an AI employee of a construction company does not need to know information that has nothing to do with working with a potential customer.

The more precisely the role of AI is defined, the easier it is to control the quality of its answers.

What to do if there is no exact answer

An AI consultant should not come up with missing information.

If the exact data is unknown, the correct scenario is to inform the visitor about it and offer to pass the question to the company employee.

For example: "The exact cost depends on the project parameters. I can help you leave a request so that the specialist can calculate the price."

Such an answer is more useful than a random or outdated figure.

Do I need to update the knowledge base regularly?

Yes.

Especially if they change frequently:

• Prices;
• Shares;
• Assortment;
• Terms of delivery;
• Tariffs;
• Terms;
• Contact details.

If the company has changed the working conditions, it is advisable to update the relevant information for the AI consultant.

How to understand that the knowledge base is well prepared

You can do a simple test.

Ask the AI the same questions that customers usually ask.

For example:

• What do you sell?
• How much does the service cost?
• What are the deadlines?
• How is the payment made?
• Is there a guarantee?
• Do you work in my city?
• What is included in the price?
• How to leave a request?

If the answers are accurate and correspond to the real work of the company, the knowledge base is prepared correctly.

How NEXSTA simplifies setting up an AI employee

At NEXSTA, an AI employee is created for a specific company and a specific task.

He can advise site visitors, answer questions about goods and services and help translate the dialogue into an inquiry to the company.

When setting up, it is important not to turn AI into a universal chat on any topic.

His task is to help the client within the business.

FAQ

Do I need to write a knowledge base from scratch?

Usually not. You can use the existing materials of the company: website, price lists, description of services and answers of managers.

The more documents, the better?

Not always. Up-to-date and understandable information is more useful than a large number of old or contradictory files.

Can AI answer questions that are not in the knowledge base?

He can use the general context of the conversation, but the specific conditions of the company should be based on real business data.

Is it necessary to constantly monitor AI?

After starting, it is useful to periodically review the dialogues and supplement the information on questions that visitors often ask.

Result

A good knowledge base for an AI consultant does not require hundreds of pages of documentation.

To begin with, it is enough to collect basic information about the company, services, prices, terms and rules of work.

The main thing is that the data is relevant, consistent and consistent with the real work of the business.

So the AI employee will be able to respond to site visitors faster and help the company process requests even when the manager is not available.

You can try an AI employee for the site in NEXSTA.