Technology & Data

Digital transformation is a fundamental change in how an organization creates value using digital technologies. It is not the automation of existing processes (that is "digitization"), nor merely converting them into electronic format (digitalization)—it is a rethinking of the business model in t...

Three levels of transformation: 1. Digitization — converting analog information to digital (paper archive → PDF) 2. Digitalization — using digital technologies to improve existing processes (electronic document management) 3. Digital Transformation — creating fundamentally new value and business ...

Changing customer expectations: Amazon, Netflix, Uber have redefined the customer experience. Now clients expect instant, personalized, 24/7 service from all companies, including banks, insurers, and retailers.

Competitive pressure: every industry faces threats from "native" digital players (digital natives), who are not burdened by outdated infrastructure.

Modern AI is not a “thinking robot” from science fiction, but a set of statistical methods that enable computers to detect patterns in data and make predictions.

Machine Learning (ML): algorithms that learn from data without explicit programming. Supervised learning, unsupervised learning, reinforcement learning.

Deep Learning: neural networks with many layers. A revolution in computer vision (object recognition), speech recognition, and language processing.

Generative AI (GenAI): models that create new content — text (GPT-4, Claude), images (Midjourney, DALL-E), code (GitHub Copilot), video (Sora).

Large Language Models (LLM) are large neural networks trained on enormous volumes of text. GPT-4 (OpenAI), Claude (Anthropic), Gemini (Google), Llama (Meta) are the main models in the market.

Capabilities of LLMs: text generation (at the human level); answering questions; analysis and summarization of documents; translation; code writing; creative thinking.

Why LLMs are a breakthrough: before LLMs, AI performed well with structured data (images, numbers). LLMs opened AI to unstructured text—80% of corporate information is unstructured.

Content generation: marketing texts, product descriptions, reports, template documents. McKinsey estimates: AI automates 20-40% of office tasks.

Before building a strategy—assess the current state. Maturity models: MIT CISR Digital Maturity Model (4 stages: beginner → fashionista → conservative → digerati), McKinsey Digital Quotient.

Key dimensions: digital capabilities (technology, data, processes); organizational readiness (culture, talent, leadership).

"Technology for technology's sake": "Let's implement blockchain" without understanding what problem it solves.

"Too many projects at once": running 30 initiatives in parallel → none are completed.

Cloud computing is the provision of IT resources (servers, storage, applications) over a network (typically the internet) using a pay-as-you-go model.

Before the cloud: the company purchased its own servers, maintained them, incurred fixed costs. With the cloud: resources are like electricity—you pay for what you use.

IaaS (Infrastructure as a Service): rental of virtual servers, storage, networks. Control over the OS and above. AWS EC2, Azure VMs, Google Compute Engine. For whom: IT teams, developers.

PaaS (Platform as a Service): a platform for developing and deploying applications without managing the infrastructure. AWS Elastic Beanstalk, Google App Engine, Heroku. For whom: developers who do not want to manage servers.

Data-driven is a culture and practice of making decisions based on data analysis, rather than intuition or experience. Amazon: Bezos introduced a “culture of metrics”—every decision that can be measured must be measured.

Stages of data analytics: 1. Descriptive: what happened? (Reports, dashboards, KPI) 2. Diagnostic: why did it happen? (Root cause analysis, drill-down) 3. Predictive: what will happen? (Forecast models, ML) 4. Prescriptive: what should we do? (Optimization, AI recommendations)

North Star Metric: one key metric reflecting value for customers. Airbnb — “number of nights booked.” Facebook — “daily active users.” All other metrics are subordinate to it.

Funnel: for e-commerce: Impressions → Clicks → Add to Cart → Purchase → Repeat. Conversion at each step is the key to optimization.

Formulas

The human brain processes visual information 60,000 times faster than text. Good visualization turns tables into insights. Bad visualization leads to misconceptions.

Power BI (Microsoft): Most widely used in the corporate segment. Integration with Excel, Office 365, Azure. Price — $10/user/month (Pro). Rich connectors to various data sources.

Tableau (Salesforce): The strongest visualization tool. Interactive dashboards, drag-and-drop. More expensive than Power BI. Popular among analysts.

Looker (Google): Modern BI based on LookML (modeling language). Metric logic is stored in the code, not in reports — a single version of the truth. Integration with BigQuery.

General Data Protection Regulation (EU, 2018) is the most influential personal data law. It applies to any company processing data of EU residents, regardless of the company’s jurisdiction.

Data Subject Rights: right of access; right of rectification; right of erasure (“right to be forgotten”); right to data portability; right not to be subject to automated decisions.

Fines: up to €20 million or 4% of global turnover (whichever is greater). Meta — €1.2 billion fine (2023) for illegal data transfer to the US.

Five principles of ethical AI (OECD): human values and rights; transparency and explainability; reliability; accountability; inclusiveness.

1. Problem Definition: The ML task must correspond to the business objective. “Reduce customer churn” → “predict the probability of churn in the next 30 days” (classification).

2. Data Collection and Preparation: 80% of ML project time. Data cleaning, handling missing values, feature engineering (creating features from raw data).

3. Model Selection and Training: simple models (logistic regression, decision tree) → ensembles (Random Forest, XGBoost) → neural networks. Rule: start with simple.

4. Model Evaluation: metrics depend on the task: Accuracy, Precision, Recall, F1 (classification); RMSE, MAE (regression); AUC-ROC (probabilistic classification).

API (Application Programming Interface) is an interface that allows applications to communicate with each other. Like an electrical socket: a standard interface for connecting different devices.

REST API is the most common standard. It works via HTTP: GET (retrieve data), POST (create), PUT/PATCH (update), DELETE (remove). Data are in JSON format.

Example: Uber does not build its own maps. It uses the Google Maps API. Stripe does not construct its own payment infrastructure—they use the APIs of banks and payment networks.

Open Banking: banks, due to regulatory requirements (PSD2 in the EU), have opened APIs for fintechs. This enabled the creation of account aggregators, PFM-applications, payment initiators without their own banking license.

Fintech (Financial Technology) refers to companies that use technology to provide financial services. The broad definition covers: payment services, lending, investments, insurance (insurtech), wealth management (wealthtech), regulatory technology (regtech).

First wave (2008–2015): Post-crisis — a crisis of trust toward banks opened the door for fintechs. Emerged: Stripe (2010), Robinhood (2013), Betterment (2010), TransferWise (Wise, 2011).

Second wave (2015–2020): neobanks (Revolut, N26, Monzo), P2P lending, insurtech. The banking experience is shifting from “branch” to “mobile app”.

Third wave (2020–): embedded finance, BaaS (Banking as a Service), DeFi (Decentralized Finance), AI-first finance.

Blockchain is a distributed ledger, in which records are: immutable, transparent, decentralized, cryptographically secured.

Mechanism: transactions are grouped into blocks → each block contains the hash of the previous one → changing any block changes all subsequent ones → the entire network verifies → impossible to forge.

Proof of Work vs Proof of Stake: Bitcoin uses PoW (mining) — huge energy expenditure, slow. Ethereum switched to PoS (staking) — 99% reduction in energy consumption.

Bitcoin (2009, Satoshi Nakamoto) is the first decentralized cryptocurrency. Limited issuance (21 million coins) → argument as "digital gold" (hedge against inflation). Institutional investors: Spot Bitcoin ETF (BlackRock, Fidelity, approved by SEC in January 2024). MicroStrategy is the largest co...

Traditional card payment: Customer → Acquirer (merchant’s bank) → Payment system (Visa/Mastercard) → Issuer (customer’s bank) → Back. This process takes fractions of a second, but settlements — T+2 days.

Participants: Visa/Mastercard — “rails” (do not hold money, only process information); issuer banks (issue cards); acquirer banks (accept payments); processors (technical layer).

Real-Time Payments: instant interbank transfers. The Bank of Russia’s system (SBP), UPI in India (8+ billion transactions/month), FedNow in the USA (2023), SCT Inst in the EU.

Open Banking Payments: payment directly from the buyer’s account without a card — via the bank’s API (A2A payments). Cheaper for the merchant, no interchange fee.

RegTech (Regulatory Technology) refers to technologies that enable the fulfillment of regulatory requirements more efficiently and cheaply. The increasing regulatory burden (AML, GDPR, Basel III) creates enormous demand.

Global spending on compliance: $270 billion per year (2022). Large banks spend 15–20% of operating expenses on compliance.

KYC/AML automation: identity verification (eKYC) via AI (document recognition, liveness check, biometrics). Providers: Jumio, Onfido, Sumsub. KYC speed: 30 minutes → 2 minutes with automation.

Transaction Monitoring: ML models detect suspicious transactions (fraud patterns, structuring, laundering). Reducing false positives is a key task.

Wealthtech is the application of technology in wealth management. Traditionally, it has been a closed, elite industry. Technologies democratize access to investments.

Robo-advisor: an automated platform that builds and manages a diversified portfolio based on the client's risk profile.

Model: the client answers a questionnaire → algorithm determines the portfolio (usually ETFs) → automatic rebalancing → all of this is inexpensive ($0 or 0.25%/year).

Major players: Betterment ($35 billion AUM), Wealthfront ($27 billion). In the UAE: Sarwa, StashAway.

Average cost of a data breach (IBM 2023): $4.45 million. In addition to direct losses—reputational damage, regulatory fines (GDPR), loss of clients. Major breaches: Equifax (147 million clients, $700 million fine), Sony Pictures (2014, $100 million in losses).

Changing Threat Landscape: attacks are no longer limited to large corporations. Ransomware targets hospitals, small businesses, municipalities.

Phishing: fraudulent emails/messages imitating legitimate organizations. 91% of cyberattacks begin with phishing. Spear phishing — targeted phishing against a specific person.

Ransomware: the program encrypts data, demands a ransom. Colonial Pipeline (2021): $4.4 million in ransom, gasoline panic on the U.S. East Coast.

There is no absolute security. The task is to manage risks: reduce the probability of an attack, minimize the damage from a successful attack, ensure recovery.

Identify: inventory of assets, understanding risks. "What are we protecting?"

Protect: controls to reduce risks. MFA, encryption, staff training, access management.

Detect: monitoring to identify incidents. SIEM (Security Information and Event Management) — aggregation and analysis of logs.

The principle of “privacy by default”: data protection must be built into the system from the start, not added later. Seven principles of Ann Cavoukian: proactivity; privacy by default; embedded into design; full functionality; end-to-end security; visibility and transparency; respect for the user.

Encryption: at-rest (data in storage) and in-transit (data in transmission). AES-256 for storage, TLS 1.3 for transmission. The “golden rule”: encrypt everything possible.

Tokenization: replacement of sensitive data (card number) with a random token. Even in the event of a leak, the token is useless without the key.

Anonymization and pseudonymization: reversible (pseudonymization) and irreversible (anonymization) removal of identifying features. GDPR distinguishes: anonymous data is not regulated, pseudonymized data is regulated.

In the cloud, security is a shared responsibility: the provider is responsible for the security of the "cloud" (physical data centers, hypervisors, network infrastructure). The client is responsible for security "in the cloud": configuring services, data, access, applications.

The main mistake: believing that AWS/Azure "do everything for us." The biggest cloud leaks are the result of incorrect client configuration (for example, an open S3 bucket).

Misconfiguration: accidentally opening public access to data. Capital One (2019): $150 million fine due to incorrect AWS IAM configuration.

Excessive privileges: every service and user should have only the minimum necessary rights (IAM policies).

Not "if" an attack happens, but "when". Understanding this changes the approach: from "prevent at any cost" to "prevent as much as possible + prepare for responding".

1. Preparation: CIRT (Computer Incident Response Team); response plan; playbooks for typical incidents; communication chains; contacts for insurer, lawyer, PR.

2. Detection and Analysis: sources: SIEM alerts, user complaints, external notifications (partners, regulators). Assessment: scope, type, severity.

3. Containment: immediate (isolate infected systems — disconnect from network) and long-term (determine root cause, eliminate attack vector).

PropTech (Property Technology) — the application of technologies for transforming the real estate market: search, purchase, management, financing of properties.

PropTech 1.0 (1990-2008): transferring listings to the Internet. Zillow (2006), Rightmove, CIAN. Digitization, not transformation.

PropTech 2.0 (2008-2020): platforms, marketplaces, sharing economy. Airbnb — redefined short-term rentals; WeWork — offices by subscription; Opendoor — iBuying (instant purchase of homes via algorithm).

Real estate investment: tokenization (fractional shares); crowdfunding (Fundrise, RealtyMogul); data and analytics (CoStar, MSCI Real Assets).

A smart building (Smart Building) is a building that uses IoT sensors and automated systems to optimize: energy consumption, comfort, security, and operational efficiency.

Components of a smart building: BMS (Building Management System) — the heart of the smart building, it integrates all systems; HVAC (heating, ventilation, air conditioning) — the largest energy consumer; lighting (automatic based on occupancy and daylight); access and security; elevator monitoring.

IoT sensors collect data: occupancy sensors (occupied/vacant — optimize HVAC and lighting); CO2 and air quality sensors; consumption meters (electricity, water, heat); temperature and humidity sensors; visitor counters.

Data → analytics → actions. Example: CO2 sensors show that the conference room is constantly occupied, although according to bookings — only 3 hours. Optimization: eliminate booking requirement, manage HVAC based on actual occupancy.

Tokenization is the representation of ownership rights to a real asset (real estate) in the form of digital tokens on a blockchain. Each token = a share in the asset. The investor purchases tokens and receives a share in the rental income and value appreciation.

Accessibility: Instead of $1 million to purchase an apartment, you can invest $1,000. Democratization of real estate investment.

Liquidity: Tokens are traded on secondary markets (vs real estate — an illiquid asset, sale takes months). This is a key advantage.

Diversification: You can invest $50,000 in 50 properties in different countries.

Traditional valuation: the appraiser analyzes comparable transactions, applies adjustments, and issues a subjective opinion. Takes days, costs $500-5000.

AVM (Automated Valuation Model): machine learning algorithms analyzing thousands of factors—location, area, floor, transportation accessibility, proximity to infrastructure, historical transactions, market trends.

AVM accuracy: for standard properties—within 5-10%. For unique (luxury, commercial) properties—it is lower. Zillow Zestimate is the most famous AVM, but had a notorious failure (iBuying program).

By analogy with hotel Revenue Management: rental price changes depending on demand, season, market occupancy, and property characteristics.

Virtual Reality (VR) enables virtual visits to properties. Applications: (1) Virtual Tours — a buyer/tenant from another country “walks” through an apartment wearing a VR headset; (2) Virtual Staging — demonstration of unconstructed space furnished; (3) Construction — virtual walkthroughs of buil...

Business Effect: Sotheby's International Realty: buyers using VR tours are 40% more likely to make an offer without a physical viewing.

AR (Augmented Reality) allows “overlaying” digital information onto the real world. An application for buyers: point your phone at a building → see data about residents, transaction prices, characteristics.

Metaverses (Decentraland, The Sandbox, Horizon Worlds) — digital worlds where you can buy “land plots” in the form of NFTs.