I turn AI pilots
into P&L impact.

// 02 — PRACTICE

Pilots don't move EBIT. Rewired operating models do.

Operating-model redesign

AI sticks when the operating model changes around it: who acts on which signal, at what cadence, with what budget authority. The NBA closed loop and the Digital Reps Center are operating-model builds, not tools.

Decision-grade, causal analytics

MMM, uplift and A/B before scale-up — true incrementality, not attribution — decide what gets budget. This is how I make AI pay: grounded in causal measurement, not dashboards. It's what separates an EBIT line from a pilot graveyard.

AI I still build myself

The hands-on part is real: I build the data and agentic tooling, not just the strategy — including a self-built MCP data-hub that serves governed commercial datasets straight into AI assistants. New capability gets a working artifact, not a slide.

Governed by default

Fourteen years in regulated pharma make model risk, data governance and human oversight the way I ship, not an afterthought — EU AI Act high-risk instinct that transfers to any regulated commercial domain.

// 03 — SELECTED WORK

Employers are named; client-level engagements are anonymized under NDA — every metric keeps scope, baseline and role.

// 05 — ABOUT

Fourteen years turning commercial questions into systems that ship — Next-Best-Action, marketing-mix modeling, omnichannel and NLP — across Bayer, AstraZeneca and Akrikhin. I set the AI & Commercial-Excellence strategy and lead the cross-functional teams that deliver it — data scientists, business partners, digital reps, analysts and CRM specialists, 5 to 30 strong — and I still build the systems and own the P&L myself. From model to outcome, every claim carries scope, baseline and my role.

// OPERATOR A. OGARKOV SERBIA (CET)

Role

AI & Advanced Analytics Leader

Tenure

14 yrs

Domains

NBA · MMM · Omni · NLP

Stack

Python · SQL · Power BI · Veeva · Salesforce

Scale

20k+ customers · 20+ launches

Teams

5–30 · cross-functional

Based

Serbia · CET · global markets

EB1-A

Extraordinary ability (2024)

// 06 — RECOGNITION

• 2024 EB1-A — Extraordinary Ability classification U.S. Citizenship & Immigration Services
• 2024 Top Digital Experts 2023 ↗ “I Am a Leader” association
• 2024 Author — «Интеллектуальный маркетинг» (in Russian) ↗ RIPOL Classic · ISBN 978-5-386-15246-8
• 2024 Awards jury ↗ RB Digital Awards
• 2023 Expert of the Year — Marketing Digitalization Techno National Business Awards
• 2022 Growth & Leadership Award — AI, Big Data & Digitalization AstraZeneca, CIS
• 2016 LIFE Award — Leadership Bayer

// 07 — DUE DILIGENCE

01 Does your pharma experience transfer to other industries?

All of it — the mechanisms were never pharma-specific. Next-Best-Action, proven as per-HCP touch ranking, becomes per-account outreach for B2B, fintech and SaaS. Marketing-mix modeling becomes media and revenue elasticity for CPG and retail. The omnichannel hub, proven across 20k+ HCP relationships, becomes an engagement hub for any distributed sales org; NLP call-tagging becomes sales and service call intelligence; regulated pharma promotion maps onto fintech, insurance and healthtech go-to-market.

02 Are you still hands-on, or a leader who stopped building?

I stopped choosing between the two. I set the AI & Commercial-Excellence strategy, lead cross-functional teams of 5 to 30 — data scientists through CRM specialists — and stay the builder of record: I can architect what I ask teams to ship, and I own the P&L for it. Current proof: a self-built MCP data-hub — governed commercial datasets, served straight into AI assistants.

03 Most AI pilots stall. Why do yours reach the P&L?

Because the model was never the hard part. AI reaches the P&L when the operating model is rewired around it — signal, cadence, budget authority — and when causal measurement decides what scales: MMM, uplift and A/B before any scale-up. Attribution flatters; incrementality pays. On the record: Next-Best-Action + marketing-mix modeling at a top-5 pharma in CIS — +7% incremental Rx, visited vs not-visited control, indexed 100 → 107; role: AI lead / builder.

04 Doesn't fourteen years in a regulated industry make you slow?

It makes me fast where it counts: shipping systems that survive scrutiny. In regulated pharma an ungoverned model doesn't ship late — it doesn't ship at all. So model risk, data governance and human oversight are how I build: an EU AI Act high-risk instinct (documentation, validation, human-in-the-loop) that carries into any regulated commercial domain. Proof of pace — the Digital Reps Center shipped inside that regime as a staffed, running commercial function, 200k interactions a year.

05 Your clients are anonymized. What can you actually show?

The NDA hides the names, not the apparatus. +35% HCP retention — omnichannel + NLP across 20k+ HCPs at a top-5 pharma in Russia/CIS, vs pre-omnichannel baseline, indexed 100 → 135; role: product owner / project lead. +183% sales — the five-year national expansion at a top-5 pharma in CIS, attributed vs pre-program baseline; role: analytics & program lead. Employers are named — Bayer · AstraZeneca · Akrikhin; clients stay descriptors. Charts are indexed, published endpoints only. Each case page prints the full Challenge → Approach → Measurable Outcome.

06 Where are you based, and what are you working on now?

Based in Serbia (CET), operating across global markets. Current seat: Head of HCP Excellence at Akrikhin (Polpharma). On the record before that: independent consulting in pharma data & analytics (2024–2025), AstraZeneca (2017–2024), Bayer (2012–2017) — fourteen years end to end. Email: hello@datadrivengrowth.tech.