Innokenty Bodrov

AI implementation for consequential workflows

I lead the product side of AI implementation: choose the right workflow, define AI and human decisions, establish quality checks and work with engineering to turn a pilot into daily practice.

Where I take ownership

I connect the business problem, AI quality and delivery. I lead product decisions alongside engineering, ML and operations teams.

Find a workflow worth changing

Map the workflow, case volume, cost of error and available data. The result is a priority use case or a reasoned decision not to use AI.

Make decision quality testable

Define what AI may decide, what stays with people, which failures matter and how the team will accept a pilot.

Carry the product into real use

Shape data and integration requirements, lead priorities and acceptance with the delivery team, and assign owners for the workflow and post-launch checks.

How AI entered transaction monitoring

A result from my employed role. The figure is author-reported; its source and limits are explained in the case.

65%

Finom · employed product role. I led product work on an AI transaction-monitoring assistant: analyst workflow, data requirements, decision logic and rollout with engineering and operations teams.

By my account, 65% of suspicious-transaction monitoring cases were handled fully automatically. This does not measure model accuracy or savings; no public Finom report with this figure is linked.

Where this experience carries over

Direct AI work, adjacent product delivery and my own products answer different buyer questions. They are separated here.

AI in a regulated product

Production product role; individual client outcomes are confidential.

Sumsub · RegTech

AI quality in digital identity

Ground truth, precision and recall trade-offs, human review and post-launch monitoring for European markets.

View role in portfolio

Adjacent systems and own products

They show delivery and adoption depth, but are not client AI outcomes.

Stenn · FinTech

Decisioning across 15+ data sources

Trade-finance decisioning platform; the author-reported improvement in decision SLA and accuracy is 35%.

Read the case
First-party project · AI

People and AI agents at work

Shared context, decision rights and evidence checks in my own product portfolio.

Explore the system
Founder product · learning

Learning and AI products

2K+ paying learners and products with hundreds of monthly users, from first-party records.

See the products

The portfolio shows context, my role, the source of each figure and the limits of each result.

Open the portfolio

From idea to controlled rollout

Four decisions connect value, AI boundaries, quality and work after launch. The diagrams below illustrate possible deliverables; they are not client materials.

Artefact · ACurrent decision map

Roles, exceptions, delays, volume and the cost of error become one validated view of the workflow.

Artefact · BSolution boundaries

Explicit decision rights, integrations, data and out-of-scope protect both the user and the delivery team.

Artefact · CEvaluation and control plan

Quality is defined before the build through a test set, failure modes, thresholds, escalation and acceptance.

Artefact · DPilot and control plan

Agree the pilot scope, acceptance criteria, workflow owner and post-launch quality checks.

For products built with AI agents, I use TDPD to connect source evidence, user scenarios, tests and human acceptance.

Explore the TDPD method

Start with one AI workflow

In 1–2 weeks we decide whether a pilot is warranted. Pilot delivery and my ongoing role are scoped separately from that decision.

For AI integrators

Bring me into presale or after it, under your brand or alongside your team. You lead the client and implementation; I clarify the process, scope, risks and acceptance criteria.

Discuss a client project
PartnerClient · technology · implementation
InnokentyDiscovery · scope · evaluation · acceptance
ClientDecisions · access · adoption

Start with one paid pilot with a fixed scope.

For startup founders

Have an AI prototype, early B2B customers or a product decision your team cannot yet make? I help define the next experiment, prepare a client pilot and work alongside founders and engineering.

Explore startup support

From systems analysis to product leadership

I moved from systems and business analysis into product leadership, platform products and production AI. Eight years of process architecture and 60+ API and system integrations across FinTech and RegTech taught me not to separate a promising idea from how decisions are actually made.

At Sumsub, I am responsible for identity and user verification products in highly regulated European markets. This includes flows for fintechs, banks, gambling and iGaming, crypto and payment processing: different ways to verify identity and payment methods, work with European identity documents and digital identity wallets. This work involves eIDAS requirements and ETSI standards.

In my own practice, I am building an AI operating system for a product portfolio and teaching senior leaders to use AI as a working system. In client work, I lead discovery, product decisions, evaluation and implementation readiness with engineering and ML teams.

SumsubFinomStennMTS

Tell me where AI should help

I read every enquiry myself. If there is a fit, I will suggest a 25-minute call to agree the next step.

innokenty@innokentybodrov.com

For a first conversation, your name, email and a few lines about the workflow are enough. The enquiry goes to my CRM; I will read it and reply myself.

Back to top