A B2B Website for Technical Buyers: How We Built LIZAR’s Platform

Avatar
Author

LIZAR Systems designs, manufactures and assembles identity document issuance systems in Europe, with headquarters in Lisbon and production in Pombal. Its customers are ministries, mints, civil registries, security printers, banks and fintechs. These buyers read the spec sheet before they read any marketing copy.

The old website did not reflect that. LIZAR had grown from one flagship machine into an ecosystem of hardware, software and AI modules. The website still presented it as a hardware supplier.

In September 2026 we launched the new platform in five languages. This article explains the decisions we made and what any B2B company with a technical product can apply to its own website.

A form captures intent. A conversation captures the reasoning behind it.
5 languages
Portuguese, English, Spanish, French and German, from a single content structure

+70%
Increase in qualified leads after launch, according to LIZAR

6 parameters
Collected by the assistant in each conversation: deployment model, volume, document type, infrastructure, geography and timeline

0 fields
The static contact form was replaced by a conversation

What was wrong with the old website?

LIZAR’s brief was direct: a website that explains a complex product family without oversimplifying it, serves as a reference throughout a procurement process and turns anonymous visits into qualified technical conversations. Three problems stood in the way.

  • The portfolio had outgrown the story. DRACO, VARAN and SMAUG handle card personalisation, passport personalisation and portable power for mobile issuance. Around them sit software modules for machine control, fleet management, AI quality control, chip encoding, immutable logging and field verification. Presented as a list, it looked like complexity.
  • Technical information was locked inside sales conversations. Every question needed a person on the other side. That slowed early evaluation and put off buyers who prefer to research before talking to anyone.
  • Incoming enquiries said very little. A name, an email and one sentence do not tell a sales engineer the volume, the document type or the deadline. Every lead started from zero.

How do you structure a B2B website for technical buyers?

The structural decision was to present document issuance as a complete chain instead of a catalogue.

Ecosystem
A flow, not a list
Visitors follow the path of a credential: source data, machine software, chip encoding, laser engraving, AI inspection, audit logging and field verification. Each step links to the module that handles it.

Spec sheets
Specifications as data
Hardware pages show the figures that decide a shortlist, always in the same order. Anyone comparing DRACO with VARAN reads the same rows in the same sequence.

Solutions
Organised by context
Distributed and mobile issuance, national ID and civil registration, migration and humanitarian contexts, critical infrastructure, banking and payments. Each buyer finds an entry point in their own vocabulary.

Compliance
Stated just as clearly
Certifications and standards appear on the product page itself, separated into what is native and what comes from partners.

The ecosystem page works as an index. Each module page works as a standalone document for visitors who arrive straight from search.

What does the AI assistant do and not do?

The assistant, built on the Claude API, replaces the contact form. It answers from the platform’s own product documentation, so it handles specification questions without improvising. A visitor can ask whether a machine works with polycarbonate or how chip encoding fits into an existing PKI.

After answering, the assistant steers the conversation towards the parameters a sales engineer needs. An eight-field form asking for the same information would be abandoned. A conversation that gives something back with every answer is not.

The assistantDoesDoes not
SpecificationsAnswers from published documentationSpeculate beyond what the documentation says
CommercialCollects volume, deadline, geography and contextQuote prices or commit to delivery dates
ComplianceStates the published certificationsSuggest certifications the company does not hold
Out of scopeSays it does not know and offers a demoInvent an answer to keep the conversation going
For a company selling to public buyers, the assistant’s restraint is worth more than the breadth of the conversation.

At the end of each qualified conversation, the extracted parameters, a summary and the full transcript go to LIZAR’s back office as a structured package, through an authenticated webhook. Nobody retypes anything. A ministry assessing a national programme and a student researching laser engraving are separated before anyone spends time on either.

Do you have a technical product and a website that does not help sell it? Let’s talk about structure, content and AI-powered lead capture.

Talk to our team →

Why do five languages need a single source of truth?

With teams in Europe, Latin America and Africa, language parity was an operational requirement. The known risk of a multilingual website is drift: a specification updated in one language and forgotten in the others.

The content model was built so that each specification lives in one place and appears on every page and in every language where it is used. When a machine’s capacity or a certification status changes, it changes once. In a public procurement process, a published figure that no longer matches the machine becomes a serious problem long before anyone notices it. The assistant replies in the language the visitor writes in.

What results did the new platform deliver?

The most visible change is in what reaches the sales team. Qualified leads rose by around 70 per cent. Qualification now happens before a person steps in. A conversation that ends at two in the morning in Bogotá is already structured in the back office when the Lisbon team starts work.

Filipe Castro, CEO of LIZAR Systems, describes the change from the sales team’s side. Before, every enquiry from the website had to be qualified by hand. Today leads arrive more complete and better qualified, with no extra work for the team. Follow-up also starts sooner.

There is a second, less obvious gain. The conversation logs show what buyers actually ask. Three patterns emerged in the first weeks.

Deployment flexibility comes up early. Buyers ask about the deployment model before asking about machine capacity. Distributed issuance is the argument that lands first.

AI questions are about auditability. For AI quality control, the questions centre on being able to explain a rejection. A public authority needs to justify the decision, not just record it.

Passport demand is higher than expected. A steady share of visits arrives looking for passport personalisation. That changed how prominently VARAN features on the site.

None of this would have surfaced through a form. The logs now shape LIZAR’s content plan.

What can you apply to your own B2B website?

  1. Publish specifications as data. The same fields, in the same order, for every product. Technical buyers compare line by line.
  2. Organise by use context as well as by product. A bank and a ministry arrive with different problems. Give each one a way in.
  3. Replace the form only if the conversation knows what it is talking about. An assistant without grounding documentation invents. Connect it to published content and nothing else.
  4. Set boundaries before you set the tone. Prices, deadlines and certifications are the answers an assistant must never improvise.
  5. Read the conversations. Repeated questions show where the documentation falls short. It is the best content plan there is.

What makes this kind of project different from a corporate website?

A corporate website presents the company. A website for technical buyers has to work as reference documentation through months of evaluation. That changes how the project is planned.

AspectCorporate websiteWebsite for technical buyers
ContentPresentation copy written onceStructured specifications, kept as data and updated with the product
ValidationMarketing sign-offTechnical review of every published figure, because one error can exclude the company from a tender
LanguagesTranslation at the end of the projectA multilingual model from the start, with one source for each specification
ContactGeneric formStructured qualification, connected to the sales team’s back office
SuccessVisits and time on pageQualified leads and technical conversations that move forward
The difference is not in the design. It is in treating the website as part of the sales process.

Which metrics should you track after launch?

On a B2B website with an assistant, the usual traffic metrics say little. These say more about whether the website is doing its job.

Qualified leads per month. Enquiries that arrive with volume, document type, geography and deadline. This is the indicator the sales team feels first.

Completed conversations. The share of conversations that reach the end of qualification. If many stop halfway, one question is putting visitors off.

Unanswered questions. The times the assistant says it does not know. Each one points to a gap in the documentation.

Most frequent questions. The topics that keep coming up in the logs. They guide the next content to publish.

Time to first human contact. How long passes between the conversation and the team’s reply. Good qualification only pays off if follow-up is fast.

These metrics only exist because the conversation is structured. A form would never provide them. Reviewing them monthly with sales and product turns the website into a source of market insight, not just a channel.

Which materials should you gather before starting?

A website for technical buyers depends on information that is usually scattered across the company. Gathering it before the project saves weeks of back and forth.

Up-to-date spec sheets. One per product, with the same fields. They are the basis of the content model and of the assistant.

A list of certifications. With the scope of each one. Note what is native and what comes from partners.

Questions from the sales team. The doubts customers raise in meetings. They are the first draft of the assistant’s answers.

A glossary per language. The technical terms used in each market, so translation does not invent vocabulary.

Frequently asked questions

B2B websites

What makes a B2B website different from a consumer website?
B2B buyers have a longer decision process, involve several people and compare specifications before talking to a supplier. A good B2B website gives them that information without forcing a sales contact.
How should a website for technical products be structured?
With specifications presented as data in the same format across products, pages organised by use context as well as by product, and a content model that keeps each figure in one place.

AI assistants for lead capture

Does an AI assistant replace the sales team?
No. It handles first contact and initial qualification. It then hands a documented conversation to the person who will speak to the customer. Prices, commercial terms and technical scope stay with the team.
How do you stop an AI assistant from inventing answers?
By connecting it only to published documentation and setting explicit boundaries. In LIZAR’s case, the assistant does not quote prices, commit to deadlines or mention certifications the company does not hold. When it does not know, it says so and offers a demo.
Which metrics show whether a B2B website is working?
Qualified leads per month, completed conversations, questions the assistant cannot answer, the most frequent questions and time to first human contact. They say more than visits or time on page.
Does a multilingual B2B website need separate content teams per language?
Not if the content model keeps each specification in one place. Translation then covers the narrative text, while figures and certifications update once and appear correctly in every language.
Can an AI assistant work in several languages at once?
Yes. In LIZAR’s case, the assistant replies in the language the visitor writes in, based on the same product documentation. Because each specification lives in one place, the answer is the same in every language.
Caixa Mágica Software, technology partner of LIZAR Systems

Caixa Mágica Team
Caixa Mágica Software is a Portuguese software company with more than 20 years of experience in custom software, applied AI and digital identity. It is LIZAR Systems’ technology partner and a member of the Claude Partner Network.

Custom software and AI · Caixa Mágica Software
A website that works throughout the buying process
We build platforms for companies with technical products and demanding buyers: specifications as data, several languages from a single source and AI assistants that qualify without inventing.