Work, shown clearly.

Company knowledge, agent roles, enquiry handling and operational data. See the specific work behind each project, including the distinction between implementation, testing and technical design.

Smart Wines: from a slow shop to connected operations

I combined a WooCommerce redesign with performance work, technical SEO, catalogue cleanup and safer links to invoicing and daily operations. The result is documented with measurements from specific checks, not estimates.

Implementation · WooCommerce · SEO · Fakturoid

Database
765 → 189.47 MB
Homepage JS
59 → 29
Public warnings
27/32 → 0/106

CliqSales · Turnkey AI team setup

An AI team built around the company’s work

For Pumpido and Win Win House, I followed the seven-week CliqSales setup method: company and product knowledge, agent roles, marketing material, CRM workflows, integrations and handover documentation. The scope was adapted to each company; the Viking project focused on technical assessment and design.

E-commerce · AI team and sales workflows

Pumpido

An AI team with product knowledge and clear responsibilities

For Pumpido, I prepared company knowledge and configured five agent roles. The work then extended to marketing, CRM and customer care, from a company presentation to an email sequence and review request.

pumpido.sk ↗

Specific work completed

  • Company knowledge for agents

    I prepared Brand DNA and Product DNA covering the brand, tone of voice, pool equipment, water treatment, heating and smart controls, and organised the material in a shared Data Center.

  • Five roles, not one general-purpose chat

    I configured agents for management, marketing, sales, operations and technical administration, with individual responsibilities, context and review rules.

  • Presentation and ten product videos

    I prepared an HTML company presentation and configured a ten-video production series. The agent team completed vertical product videos with Slovak voice-over, captions and the company logo, and the outputs passed internal quality review.

  • Seven-email follow-up

    I added seven CRM emails for days 0, 1, 3, 5, 8, 11 and 14, with consent, email-suppression and unsubscribe checks. I removed a duplicate step and verified internal test delivery.

  • Post-purchase review request

    I configured a review-request workflow linked to post-sale follow-up, with a Heureka link, eligibility checks, audit tags and re-entry disabled to prevent repeat requests.

  • Work and dependency overview

    I created an operations dashboard showing tasks, integration status and blockers. It explicitly distinguishes a saved snapshot from live data.

Prepared workflow

  1. CRM contact
  2. Consent and sending rules
  3. 7 follow-up emails
  4. Post-sale care and reviews

Deliverable and verification

The deliverables include a configured AI team, shared knowledge and specific sales workflows with safeguards. Internal tests verified delivery of all seven emails.

Scope: The email sequence remained a draft with no customer campaign, and the videos were not published to social media. The review workflow awaited a complete customer test and shop integration. Fireflies-to-wiki integration remained a separate open item.

Hospitality and events · CRM and operations

Win Win Coffee / Win Win House

From an event enquiry to a sales agent

For Win Win House in Trnava, I configured an AI team for marketing, sales and operations. A key part was structured event intake: the CRM stores the details and a sales agent prepares an internal summary for follow-up.

winwincoffee.sk ↗

Specific work completed

  • Brand and service knowledge

    I prepared Brand DNA and three service areas: breakfast, all-day dining, and the café, bar and event spaces. I set Slovak communication, the company profile and shared wiki material.

  • AI roles and boundaries

    I configured five agents: Steve for management, Marc for marketing, Alex for sales, Meg for operations and Elon for technical administration. The sales agent processes enquiries internally; a person confirms prices and binding reservations.

  • Marketing and event material

    I prepared a Brand Kit, company presentation and event-planning material for Trnava, and adapted the landing page, questionnaire and thank-you page to the brand and Slovak language.

  • Event form and CRM pipeline

    I built a form for event type, date, guest count, budget, catering and additional requirements, plus a seven-stage CRM pipeline. Tests verified contact-field mapping and opportunity creation.

  • Automatic handoff to Alex

    I connected questionnaire submission to the CRM and an authenticated webhook to the AI team. A test confirmed agent intake and a saved summary without an automatic email, SMS or reservation confirmation.

  • Protected dashboard and read-only data

    I created a login-protected dashboard and adapters for Dotypos, CRM, Meta Ads and FIO. Point-of-sale and banking access is read-only, and the overview distinguishes available data, an empty result and a source error.

Verified test journey

  1. Event questionnaire
  2. CRM contact and opportunity
  3. Handoff to Alex
  4. Internal enquiry summary

Deliverable and verification

A test verified the questionnaire → CRM → AI agent → internal summary chain. The dashboard API is protected; anonymous data access was rejected during verification.

Scope: The public event funnel remained intentionally unpublished. The agent does not confirm reservations or send customer replies on its own. Dashboard freshness requires ongoing checks; this is not a revenue-growth claim.

Sales and service · Technical analysis and phased plan

Bazény Viking / VWD

A plan to connect jobs, inventory, technicians and AI

For Bazény Viking and VWD, I assessed a detailed brief for the IQ system in Knack and its integrations. I separated native-system work from external development and identified prerequisites for implementation.

bazeny-viking.sk ↗

Specific work completed

  • Breaking the brief into work packages

    I analysed requirements for sales, installation, service, inventory, invoicing, attendance and the customer portal, separating IQ changes from a mobile app, CRM and external integrations.

  • Secure integration design

    I assessed Knack APIs and webhooks and proposed a server-side integration layer, covering key protection, repeated events and which system should be authoritative for each data type.

  • Inventory-design review

    I reviewed dispatch notes, technician-held material, stock reservations, counts and serial numbers, and flagged that unused material must not be restocked without confirmation of its physical return.

  • Field-work and invoicing assessment

    I assessed mobile requirements for photos, material, mileage, signatures and offline work, along with SuperFaktura integration, separating native Knack features from mobile or middleware development.

  • Phases for CRM, AI and further automation

    I placed AI roles, CRM synchronisation, marketing bots and meeting transcripts within the wider plan, proposing implementation order, technical pilots and the inputs needed to price individual phases.

Proposed implementation order

  1. Existing IQ audit
  2. API and integration layer
  3. Inventory and invoicing
  4. Mobile operations and AI

Deliverable and verification

The output is a technical assessment, responsibility split and phased plan, with specific decisions for management and a framework for smaller, verifiable work packages.

Scope: This reference describes analysis and design, not a completed inventory system, mobile app or production AI team. Implementation depended on an existing-application audit and integration verification.

More work

DeerVision logo

DeerVision

For DeerVision, I designed and built a new presentation website for mental coach Marian Jelínek Jr. The work covered service architecture, copy, responsive implementation, analytics and technical SEO.

deervision.cz ↗
Cover photo for the Rodič není Trenér project

A Parent Is Not a Coach

For this expert project for parents of young athletes, I created a publishing system and a local tool for managing Herohero from Codex and Claude. Sensitive actions such as uploads, profile changes and publication still require explicit human confirmation.

40+

More than forty people guided through working with an AI team

During seven-week programmes, participants learn to use an AI team in everyday work, assign tasks and review its outputs. The number describes programme participants, not forty separate production deployments.