Marina Rakhimzhanova · Venture Building

Real AI value for real business

I identify market opportunities, bring specialist teams together and test ideas in real projects. AIStart360 provides the methodology; JOIN LAB connects people for collaborative work.

Bring a business challengeJoin an experiment
For investors and strategic partners
Real customersReal processesMeasurable outcomes
MARINA RA LAB / 02 ACTIVE

Projects in focus

Human × AI × Business

01HONORClient case · retail021 MLN KMOur own laboratory
METHOD7 STAGES
OUTPUTEVIDENCE
REGIONCENTRAL ASIA

01 / WHY

Business has enough AI presentations. What it needs is evidence.

Companies launch assistants and automation, yet often cannot answer which customer problem AI solves, which metric should change, or whether the solution deserves to scale after the pilot.

“

We do not start with technology. We start where a customer or a business loses time, money, trust, or an opportunity to grow.

01 / BUSINESS SERVICES

01You need a plan

Point A analysis

Your challenge
You need to understand what constrains growth and where to start.
What we will do
We will review the data, market and seven GRI areas, then choose the main constraint.
What you will get
A baseline map and a 30 / 60 / 90-day plan with a first experiment.
Explore Point A analysis

02You have a process and data

AI pilot

Your challenge
You have a workflow and data, and need to test the value of AI.
What we will do
We will compare one AI use case with the current workflow on quality, time and costs.
What you will get
Test results and a recommendation to continue, change or stop.
How a pilot works

02 / GRI × GTM GROWTH SYSTEM

6 AI-powered
growth levers

GRI helps identify a constraint. GTM helps choose an action. In a pilot, we test whether that action delivers value.

AIStart360 methodology · Marina Rakhimzhanova ↗
GRI03 / ACTIVATE

Growth levers

Choose a business challenge and see how to test it.

  1. 01Faster salesthe path to payment
  2. 02Repeat salesreturning customers
  3. 03Order value and purchase frequency
  4. 04Partner and sales-agent network
  5. 05Customer reviews and referrals
  6. 06Cash flowMore predictable receipts
Target economics from the GTM modelLTV / CAC ≥ 3CAC payback 1–2 monthsRepeat revenue 40–60%MoM growth 15–25%
Explore the GRI & GTM library →

04 / METHOD

From a business challenge to proven impact.

Four phases. Each produces a clear deliverable and a decision about the next step.

  1. 01
    About 7 days

    Point A analysis

    We analyse the challenge, customer journey and business baseline.

    • Interviews with the owner and team
    • Data, constraints and growth priority
    DeliverableChallenge brief and baseline metrics
  2. 02
    1 day · intensive

    GRI working session

    We connect GRI findings to a growth hypothesis and a test plan.

    • Key constraint and growth lever
    • Hypothesis, metric and team roles
    DeliverableGrowth roadmap and first pilot plan
  3. 03
    From 21 days + support

    Business pilot

    We test one use case with real customers and business processes.

    • Prototype, implementation and short cycles
    • Baseline comparison and cost tracking
    DeliverableMeasured effect and a pilot decision
  4. 04
    Planning guide · 9–12 months

    Repeatable growth

    We embed effective solutions in processes, the team and management.

    • Standards, ownership and monitoring
    • Economics and a scaling plan
    DeliverableA repeatable system with clear economics

Timings are planning guides. Scope, metrics and team involvement are agreed before starting. Each phase informs the decision to proceed.

Seven steps within the methodology
  1. 01

    Challenge

    Where does the customer or business lose value?

  2. 02

    Diagnosis

    Journey, data, constraints, and baseline metrics.

  3. 03

    AI hypothesis

    What does AI do, and what remains human?

  4. 04

    Prototype

    The smallest solution capable of testing behavior.

  5. 05

    Pilot

    Work with real customers and business processes.

  6. 06

    Measurement

    Compare the outcome with the baseline.

  7. 07

    Scale

    A playbook to improve, stop, or scale the solution.

SUCCESS CRITERION

An experiment succeeds not when the AI works, but when a customer or financial outcome changes.

Discuss your starting point

05 / RA LAB / DECISION LIBRARY

A library for your business challenge.

We will help you find a practical guide, leaders’ experience and research to test your idea.

Open the library ↗Explore leadership decisions →
01

Solve a business problem

Sales, retention, order value and cash flow

02

Examine a company decision

Charts, diagrams, alternatives and leadership lessons

03

Test your hypothesis

An action plan, a metric and a next-step criterion

06 / OWN LABORATORY · TRAVELTECH

1 MLN KM: researching the travel experience.

Our own lab: we test demand, tour selection and the path to payment.

1 MLN KM project illustration: travellers, mountains and a route02 / TRAVELTECH

Owned project · validating demand and business model

1 MLN KM — building an intelligent travel ecosystem

Developing a curated travel marketplace connecting travellers, vetted organisers and confirmed sales. The first pilot segment is women’s curated travel for ages 33+.

An independent project and an owned laboratory: testing demand and a commission model through travel, product and technology expertise.

AI advisorRoutesGeodataCommunity
Explore the lab →Discuss your challenge ↗

07 / EXPERIMENTS

Growth hypotheses. Tested in reality.

We investigate where AI changes customer behaviour, team performance and business economics. Every experiment begins with a real problem and supports a decision: scale, iterate or stop.

Research agenda: proposed hypotheses, not completed experiments or validated outcomes. An implemented feature does not prove its economic impact.

EXP 01Proposed hypothesis

RetailTech · HONOR

Better decisions. Better sales?

Can context-aware AI recommendations increase paid conversion without compromising margins or recommendation quality?

Primary metric
Paid conversion
HONOR case study →
How we will test it
Supporting metrics
Gross profit per visitor · Decision time
Business problem
Customers struggle to choose a suitable kit for their activity, weather and body measurements. Recommendations must help a purchase, not just sound convincing.
Comparison and test design
We will compare ordinary selection with AI-assisted selection on comparable HONOR traffic. Before launch, we will define the allocation unit, baseline conversion, sample size and rules for recording online and offline payments. We will account for inventory, seasonality and discounts; without a valid comparison, we will not attribute growth to AI.
AI and human roles
AI proposes a kit using verified product data. A consultant checks complex recommendations, availability and fit for the customer's needs.
Guardrails
Recommendation errors, returns and margins. Stop if errors or returns increase beyond the agreed limits.
Decision after the test
We expand selection only when paid conversion improves meaningfully while quality and economics hold. If evidence is insufficient, we extend observation or change the hypothesis.
EXP 02Proposed hypothesis

Customer lifecycle · HONOR

From first purchase to lifetime value.

Can AI personalisation grounded in real customer needs increase profitable repeat purchases?

Primary metric
Repeat purchase rate
HONOR customer system →
How we will test it
Supporting metrics
Incremental margin · Reactivation cost
Business problem
A history of purchases and fittings does not itself bring customers back. We need a useful offer at the right moment.
Comparison and test design
We will compare similar buyer cohorts receiving personalised offers or standard service. We will define the repeat-purchase window in advance and account for returns. Observation will last 90–180 days or longer, depending on the purchase cycle.
AI and human roles
AI proposes the next relevant contact. A staff member checks the offer and contact frequency. Communications require appropriate customer consent.
Guardrails
Unsubscribes, complaints, returns and discount costs. Unprofitable discount-driven purchases do not count as success.
Decision after the test
We identify which personalised offers create additional profitable purchases and justify their costs.
EXP 03Proposed hypothesis

TravelTech · 1 MLN KM

From inspiration to trusted itineraries.

Can AI turn travel inspiration into a verified itinerary that users are willing to adopt?

Primary metric
Validated itinerary acceptance
1 MLN KM laboratory →
How we will test it
Supporting metrics
Planning time · Critical error rate
Business problem
Interest in a destination still leaves questions about time, budget, accessibility and reliable information.
Comparison and test design
Once a working prototype is confirmed, compare independent planning and AI assistance on identical tasks. Count an itinerary as accepted only if the user selects it and it passes material constraint checks.
AI and human roles
AI assembles and helps revise the plan. A person checks safety, constraints and important local facts; material claims need traceable sources.
Guardrails
Critical factual errors and infeasible constraints. Accepting an itinerary does not prove a completed trip or revenue.
Decision after the test
We decide whether the use case is useful enough for further development and a separate monetisation test.
EXP 04Proposed hypothesis

GRI × GTM · AI economics

Which AI is worth the investment?

Can diagnosing a business constraint help select AI use cases with measurable economic value after all incremental costs?

Primary metric
Incremental profit contribution
Methodology and library →
How we will test it
Supporting metrics
Payback period · Pilot cost
Business problem
Choosing technology before defining the business problem can produce working automation with no clear economic contribution.
Comparison and test design
We record the baseline, main business constraint and one GTM hypothesis. The manager approves the budget, primary metric and stopping rules. Where feasible, we use a control group or phased rollout. We include integration, data, AI usage, training and human review in costs.
AI and human roles
AI supports analysis and proposed actions. A manager owns the budget, measurement quality and final decision; an accountable specialist checks financial conclusions.
Guardrails
Data quality, operational risk and budget overruns. Where incremental impact cannot be estimated reliably, report uncertainty rather than attributing all process revenue to AI.
Decision after the test
We scale, change or stop the selected use case. One pilot does not establish the superiority of the entire GRI methodology; long payback periods require continued observation.
EXP 05Proposed hypothesis

Human × AI · B2B and services

When does AI make humans better?

Can the right division of work between AI and people improve productivity while preserving customer trust and service quality?

Primary metric
Quality-adjusted tasks per hour
Business challenges in the library →
How we will test it
Supporting metrics
Resolution time · Error rate
Business problem
Automation may speed up a response while creating errors, repeat contacts and additional work for the team.
Comparison and test design
We will compare ordinary work with AI-assisted work on similar B2B or service tasks. Before the test, we will define the quality standard, task complexity and cases requiring a human handoff.
AI and human roles
AI prepares a response or action; a person checks consequential decisions. People own customer commitments, exceptions and final execution.
Guardrails
Errors, repeat contacts, customer satisfaction and policy violations. Speed without quality is not an improvement.
Decision after the test
We identify what can be automated, what to leave to specialists and where human review is mandatory.
NEXTOpen call

Your business challenge could be our next experiment.

Where is your company losing customers, time or profit? Together, we can define a testable hypothesis, meaningful metrics and the right pilot approach.

Propose a business hypothesis →

The outcome determines the observation window

A 21-day test can assess a bounded workflow change. A 90-day business pilot includes baseline measurement, implementation and economic evaluation. Repeat purchases may require 90–180 days or more. These are planning horizons, not promises of results.

What we will publish after testing

Baseline, comparison method, period, sample size, costs, limitations and the decision. Before launch, we define the primary metric, minimum meaningful effect and stopping rules. A null or negative result can still support a useful decision.

08 / AISTART360 · JOIN THE LAB

Strong people.Bold hypotheses.Real results.

We bring specialists together around complex business challenges. We test ideas in real projects and build solutions whose value can be demonstrated.

Explore experimentsOffer your expertise
We start with a challenge, a role and a way to contribute.
  1. 01Business challenge

    We find where value is lost

  2. 02People × hypothesis

    We connect expertise and a testable idea

  3. 03Test → value

    We measure the effect and make a decision

HONOR · 1 MLN KM · YOUR CHALLENGE

Have a challenge for the team?

Bring a business challenge

09 / FOUNDER

Marina Rakhimzhanova

Venture Builder · Founder of AIStart360

Portrait of Marina Ra
AI × GTMCentral Asia

We integrate GRI and GTM diagnostics, sales, analytics, and AI into a unified system for measurable business growth.

We have worked across retail, SaaS, FMCG, banking, pharma, logistics, and technology projects in Central Asia and Eastern Asia.

+300%year-over-year growth in one project
×2revenue growth to ≈ ₸2.5B
25people in the sales team built
Our role is to connect AI with real business constraints and turn experiments into a repeatable growth system.

10 / KAZAKHSTAN × UZBEKISTAN

Regional ecosystem

Astana HubTechnology parkIT Park UzbekistanIT ecosystemMOSTBusiness & capitalAlmaUUniversityAI & Digital BridgeTechnology forumCEVFVenture forum

Regional organisations and events. Inclusion does not imply a partnership with RA LAB.

Discuss a partnership

Choose a location around your business challenge.

We are developing opportunities to work together across six locations. These are options for future programmes, not a required route through every country.

Learning and professional resources
IxDFIxDFGTM AllianceGTM AllianceSITE GenevaSITE GenevaAIStart360AIStart360
11 /Kazakhstan and Central Asia contextWe account for the region’s market, teams and customer behaviour.Read moreShow less

AI IN CENTRAL ASIA

AI implementation grounded in Kazakhstan’s real business context.

We work with companies in Kazakhstan and Central Asia to move from AI interest to a measurable operating result—through customer evidence, workflow redesign and controlled field experiments.

01

AI readiness × growth

We diagnose the customer, data, process and organizational conditions that determine whether an AI use case can create value.

02

90-day business pilot

We select one high-value hypothesis, define a baseline and KPI, test the Human × AI workflow and make a Scale / Iterate / Stop decision.

03

GTM and customer experience

We connect AI to a real customer mission, adoption mechanism, commercial metric and go-to-market plan.

Almaty · Astana · Kazakhstan · Uzbekistan · Central Asia

12 / FROM AI HYPOTHESIS

Do not implement AI blindly. Prove the value first.

Discuss your challenge
Marina Ra invites you to connect
Discuss your challenge
Discuss your challenge