Skip to content

An AI-powered studio CRM: from company analysis to mockup, email and order

This is our own CRM, and we use it every day. It checks the public websites of Lithuanian companies, uses AI to rate their design and suggests whom to approach and with what offer. It then builds a new-site mockup based on the niche’s competitors, drafts the email, queues it and sends it by set rules, and it also tracks replies and tasks. The database holds about 45,000 rated websites and over 100 mockups. By our estimate, routine work now takes about 95% less time. We build systems like this for other companies too.

  • ~45,000rated websites
  • 100+mockups
  • 10–20emails a day, by set rules
  • ~95%less time on routine work (by our estimate)
Studio CRM overview: email queue by day and what to do next

The brief

To offer a company a new website you have to find the company, open its site, check speed, SEO and GDPR, look at competitors, build a mockup, write the email, remember the follow-up, read the reply, and make sure neither we nor our partner write to the same company twice. By hand, one company takes several hours, and tracking dozens of emails and tasks in different places is hard.

The solution

Company and website analysis. The system takes open data from the Lithuanian Register of Legal Entities and Sodra, finds .lt websites and the email addresses companies publish there, and checks whether each site works. Scripts take desktop and mobile screenshots and measure Google Lighthouse speed, SEO against 7 criteria and GDPR (cookie banner, privacy policy). They also check since when the design has stayed unchanged and whether Google flags the site as unsafe. AI rates the design from the screenshot and suggests what the company needs: a new site, SEO or just fixes.

Whom to contact. A 0–100 rating combines need, niche, company size and contact. The system skips competitors, blogs and companies in liquidation. AI assigns each site’s niche.

Mockup based on competitors. An AI agent reviews 3–5 strong competitors in the niche and 1–2 foreign leaders, notes what is worth adopting, and builds a mockup of a new site from the company’s own facts, in its brand colours. It fixes the problems found, and the mockup is checked automatically. AI generates any missing images. The mockup goes online with a view counter.

Email and queue. The email shows the site on desktop and mobile, Lighthouse scores, the main problems and a link to the mockup. Every email is checked automatically before sending. The system then sends emails itself: only Tuesday to Thursday, at times suited to the niche, 10–20 a day, one every 9–14 minutes, with one follow-up after 6 working days.

Replies and tasks. The system reads the mailbox and sorts replies, bounces, opt-outs and auto-replies. A person sees who viewed the mockup and what to do next. Client requests become tasks: you pick a scenario in a form, and an AI agent works through the steps, asks questions and drafts the email. Only a person can publish or send.

Working together. We and our partner each work on our own computers. Only the company registry on the server is shared: a company is reserved before work starts, each user connects with their own key, and a company that opted out is closed for both.

Integrations and modules

  • Lithuanian Register of Legal Entities (JAR), Sodra – open data
  • Google Lighthouse, Google Web Risk
  • Claude API, OpenAI API
  • IMAP / SMTP, DKIM
  • a shared company registry on the server (PHP + SQLite API)

Result

One system from the company registry to the order: analysis, mockups, emails, replies, tasks, finances. About 45,000 rated websites, over 100 mockups. By our estimate, routine work takes about 95% less time. Partners work separately, and the same company never receives a second offer. Duration: since October 2026, still being developed. Budget: internal project.

An idea, not delivered work

What we would add with AI today

AI models and agents used, and for what. Claude Opus – design rating from screenshots, a recommendation and a sentence for the email. Claude Haiku – each site’s niche from its title, description and opening text. Claude Code (agent) – competitor analysis, mockups, tasks from the form, client email drafts, commercial proposals, a weekly SEO review of our own websites. OpenAI gpt-image – images for mockups. Sending, scheduling, Lighthouse measurements and inbox sorting run on scripts and rules, without AI. Emails and GDPR. We write only to companies, using public registry data and the email addresses companies publish on their own websites. Every email can be answered with “Ne” (No), and we stop writing. Opted-out and bounced addresses go on a permanent do-not-contact list. Emails per day are capped, and sending stops if too many bounce. What a human checks. A person writes and sends all replies to clients. Only a person publishes individual mockups and proposals. We review each mockup before the email, and automatic emails can be viewed and withdrawn before they go out. What else such a system can automate (possibilities, not yet built in this system): AI could sort replies by meaning (asking for a price, asking for a call, declining) and draft an answer with prices from the price list. After a conversation the system could issue an invoice in the accounting software, move reminders into a calendar, transcribe calls and link them to the client, and email a weekly report.

Studio CRM

Screenshots

  • Studio CRM: a new task for the AI agent – a form with six scenarios
    Studio CRM: a new task for the AI agent – a form with six scenarios
  • Studio CRM: the email queue with statuses (demo data)
    Studio CRM: the email queue with statuses (demo data)
  • Studio CRM: incoming mail sorted into replies, opt-outs, bounces and auto-replies (demo data)
    Studio CRM: incoming mail sorted into replies, opt-outs, bounces and auto-replies (demo data)
  • Studio CRM: from the company registry to the email, rating and candidates (demo data)
    Studio CRM: from the company registry to the email, rating and candidates (demo data)

FAQ

Frequently asked questions

Is this an off-the-shelf product?

No, it is our own CRM, and we use it every day. We build systems like this for other companies too.

Where do the company email addresses come from?

We use public registry data and only the email addresses companies publish on their own websites. We write only to companies.

How do I stop receiving emails?

Every email can be answered with “Ne” (No), and we stop writing. Opted-out and bounced addresses go on a permanent do-not-contact list, and a company that opted out is closed for our partner too.

Does AI reply to clients by itself?

No. Incoming mail is sorted by scripts and rules, without AI, and a person writes and sends all replies to clients. Only a person publishes individual mockups and proposals.

Enquiry

A similar solution: process automation and a CRM with AI

Tell us what you want to automate or connect – we will assess it and suggest where to start.

+370 670 99265Mon–Fri 9:00–18:00

info@saitai.ltWe reply within 1 working day

Alaušo g. 1C, LT-08424 VilniusMeetings by prior arrangement by phone only

Enter a number

— required field. We use your data only to reply to your request.