Yossi ElmaliahCPA · Group CFO · Founder & CFO, FinPro

AI in Practice

Finance judgement, built into systems that run every day.

I am a CPA and Group CFO with more than twelve years leading finance in international, multi-entity groups. Over the past year I have used AI to design and build four systems that run my own businesses. I defined the logic, the data model and the controls. AI tools were the engineering team I directed.

4business systems designed and built with AI
33 / 246finance components and sub-components in one live knowledge engine
70%improvement in communication with workers and employers
50%improvement in operating efficiency at WorkForce

Advisory platform · finpro.com.cy

FinPro: diagnostic engine and knowledge library

A CFO's diagnostic judgement, turned into a structured and repeatable system. It shows a business owner where the finance function is weak and what to fix first, and it turns that insight into qualified demand for my practice.

  1. Knowledge engine33 components · 246 sub-components
  2. Smart assessmentsQuick and in-depth
  3. Automated findingsWhere it hurts, what to do
  4. CRMEvery lead, tracked
  5. Content libraryLinked to LinkedIn

What I built

  • A website presenting my services, with a library of professional content I write and publish.
  • Smart questionnaires that assess core finance processes and return an automatic diagnosis: where the weakness is and what the company should do to strengthen it.
  • An in-depth assessment of several dozen questions that maps the whole company, so an owner or CEO can see exactly where the financial and operational weaknesses sit.
  • A CRM behind the website that shows me every potential client as they arrive.
  • LinkedIn publications linked to the library content, so I can measure what creates real interest.

Under the hood

  • A live knowledge engine of 33 finance components and 246 sub-components. It is the single source of truth behind every questionnaire, every result and every market case analysis.
  • Four connected databases, for knowledge, tools, clients and marketing, linked by one client ID from first visit to signed client.
  • Benchmarks by ten industries and company size, managed through an admin module with review before publication and full version history.
  • Campaign attribution: tags are captured at first touch and attached to whatever the visitor submits, so content performance is measured in leads, not likes.
  • A content workflow built on my own AI skills: drafting, legal review against EU and Cyprus advertising rules, and visual design.
Claude CodeVercelMongoDBStripe
FinPro assessment screen: the ERP and accounting component, with a graded maturity question on scalability
The assessment engine. Each area is scored as component, sub-component and graded maturity level, drawn live from the knowledge engine.
FinPro admin console: the campaign links builder, with destination, channel, source and campaign fields and a list of saved LinkedIn links
Campaign links in the admin console. Every post, ad or email gets a tagged link, so each lead arrives showing the campaign that brought it.

Operations · workforce.cy

WorkForce: recruitment CRM with a daily AI agent

A recruitment business that runs on a process rather than on memory. No lead is lost, every conversation starts from a ready answer, and my time goes where it creates the most value.

  1. Smart intake formsWorkers and employers
  2. CRMDuplicates blocked at entry
  3. Automatic matchingCandidate to employer
  4. Daily AI agentNext best action
  5. WhatsApp templatesSent after my approval

What I built

  • A CRM that records worker and employer leads, with alerts and reminders so nothing falls through the cracks.
  • Automatic matching between candidates and employers.
  • A ready message template for every common question, connected directly to WhatsApp.
  • A daily AI agent that advises me where to invest my time. Nothing is sent without my approval.
  • Automatic logging of media publications, which makes monitoring and control simple.
  • A mobile app, so I can access the data and run the business from anywhere.
70%improvement in communication with workers and employers
50%improvement in overall operating efficiency

Under the hood

  • Public smart forms that write straight into the CRM, so each record is entered once, at source.
  • A matching engine that scores each pair on the factors that decide a real placement: nationality, live-in or live-out, city, availability, childcare and experience.
  • Role-based access: an outside partner sees only the candidates shared with her workspace.
Claude CodeNext.jsVercelGoogle SheetsWhatsApp
WorkForce daily workspace: scored candidate and employer matches ready to action, and the people who need attention today, each with an approve-and-WhatsApp action
The daily workspace. Matches scored and ready to action, and the people who need attention today, each one a click from an approved WhatsApp message.

Finance control · all my companies

Client & Billing System

One controlled flow across all my companies: know the client, sign the agreement, then bill and collect. The routine work is done by the system, and the approval stays with me.

  1. Onboarding & KYCCompany and director documents
  2. AgreementGenerated and e-signed
  3. Billing profileServices and terms
  4. Automatic invoiceCalculated and sent
  5. CollectionMonitored with alerts

What I built

  • All my companies managed in one system, with a client account for each customer, its agreements and its services.
  • Full onboarding, including a Know Your Client process and upload of company and director documents.
  • Agreements generated automatically and signed by the client inside the system.
  • Automatic invoices, including the calculations, sent to each client, with collection monitoring and alerts.
  • Stage-by-stage alerts, and connections to the other systems I am building.

Controls built into the system

No billing before KYC and signatureThe control sits in front of the money, not in a reminder.
Credit note onlyAn issued invoice is never edited or deleted.
Gapless numberingOne sequential series per company, assigned at issue.
Full audit trailEvery change is recorded.
Client record showing status, outstanding and overdue balances, unmatched payments and the next invoice
The client record. Outstanding, overdue, unmatched payments and the next invoice, read at a glance before picking up the phone.

Under the hood

  • Integrated with Google Drive and Gmail to file and send invoices, with EU VAT number checks (VIES) and European Central Bank exchange rates.
  • Specified in full first, then built phase by phase with an AI coding agent.
Claude CodeNext.jsPostgreSQLGoogle WorkspaceE-signature

Client service · my first client product

Cash Flow System

Cash-flow management for small and mid-sized groups, delivered as an ongoing service. My judgement sits on top, and the system carries the routine work, from learning the client to delivering the monthly management pack.

  1. From billingNo double entry
  2. 6-stage onboardingThe client as a group
  3. Maturity assessmentWhat to fix, and where
  4. Cash-flow modelForecast vs actual
  5. Management packExceptions and required actions

What I built

  • The client arrives directly from the billing system, and onboarding starts from there.
  • Onboarding spread over six stages, through which the system learns the client as a group of companies: entities, banks, contacts, report types and more.
  • At the end of the assessment, the system determines the client's maturity level automatically and, where something needs fixing, shows exactly where.
  • Every task runs on a tight, agreed timetable with the company.
  • Direct, automated contact with the company's different managers, with feedback on the reports and information they provide.
  • Once the client is ready, a full cash-flow model built from its own reports, with forecast compared to actual.
  • A management reporting pack that highlights exceptions and the actions the company must take.

Under the hood

  • Connected to the Client & Billing System, so a new client flows in without being entered twice.
  • An operations agent for each client that handles the technical work, while I review and decide.
  • A cash-flow model with a KPI dashboard, traffic-light status and a minimum cash buffer alert.
  • Supported by client guides I wrote, such as a 13-week forecast readiness checklist and cash-flow warning signs.
Claude CodeAI agentsAutomated workflows
Cash Flow onboarding assessment: a six-step progress map from Contact to Result, with 8 of 28 questions answered
The onboarding assessment. Six steps, from the contact person to the result: what can be built, what is missing, and what the client needs to provide.

One connected architecture

Designed as one client journey, not four separate tools

The systems hand work to each other. A prospect who finds me through FinPro becomes a client in the billing system and is onboarded into the cash-flow service without anyone typing the same data twice.

AttractFinProContent, assessments and CRM bring in and qualify prospects.
Contract and billClient & BillingKYC, signed agreement, invoicing and collection under control.
DeliverCash FlowOnboarding, the model and the monthly management pack.

WorkForce runs a separate business on the same principles: one source of truth, automation for routine work, and human approval before anything leaves the system.

How I work with AI every day

A personal operating system for a finance leader

Beyond the systems above, AI is part of how I learn, plan and communicate every day.

Email and calendar agentConnected to my inbox and calendar, with automatic meeting summaries.
Daily brief, weekly reviewA structured start to each day and a weekly review of what moved and what did not.
Personal knowledge baseA growing database of what I learn, used to improve how I work.
Skills, connectors and MCPCustom AI skills and integrations that automate recurring work across my tools.
ResearchClaude, ChatGPT and Perplexity for market, regulatory and technical research.
LinkedIn contentProfessional posts written, legally reviewed and illustrated through my own AI workflow.

Build

Claude CodeClaude CoworkCursor

Research and content

ClaudeChatGPTPerplexity

Data

MongoDBPostgreSQLGoogle Sheets

Platform and integrations

VercelGoogle WorkspaceWhatsAppStripeMCP

See it live

The work, in public

FinPro websitewww.finpro.com.cy WorkForce websitewww.workforce.cy Yossi on LinkedInlinkedin.com/in/yos-fin FinPro on LinkedInlinkedin.com/in/finpro-transformation