From mathematics to strategic decision-making intelligence

The algorithmic basis of mAInthink

The technological basis of mAInthink was not created at short notice, but is the result of decades of scientific work at the interface of mathematics, algorithms and computer science.

Dr. Igor Kadoshchuk, who has been working on mathematical processes, optimization algorithms and computer-aided decision-making at university since the 1980s, plays a central role in this.

Scientific background: mathematics meets computer science

Dr. Kadoshchuk held a chair at the Moscow Institute of Physics and Technology (MIPT ), where he worked as a professor of computer science. His research and teaching focused in particular on

• mathematical optimization
• Algorithmics
• combinatorial decision problems
• computer-aided modeling of complex systems

In the course of these many years of work, a central insight emerged:

Mathematical methods and computer technology can be arranged in such a way that in complex project and investment decisions it becomes objectively recognizable which path is the best.

Project optimization as a mathematical problem

Project, portfolio and investment decisions ultimately consist of IDs, variables, restrictions and numbers. The problem lies not in the availability of data, but in the combinatorics.

Even with just a few projects, the number of possible combinations grows exponentially:

• 5 projects → 32 combinations
• 10 projects → 1,024 combinations
• 20 projects → over 1,000,000 combinations
• 50 projects → astronomical orders of magnitude

Traditional tools (e.g. Excel, simple scoring models or linear approximations) are generally unable to fully map this complexity, but inevitably work with simplifications.

Hybrid AI & multithreaded computing

mAInthink therefore uses hybrid AI approaches that combine classical mathematics, heuristic optimization and algorithmic search with modern multithreaded computing architecture.

This enables us to achieve an accuracy of 97% to 99.99% in real project and investment scenarios and to perform very complex project calculations very quickly, which conventional tools typically cannot achieve at this depth and speed.

Why not 100%?

If you theoretically want to achieve 100% accuracy, this means that every single possible combination would have to be calculated completely, precisely and without shortcuts.

Even with today's technical possibilities, this would mean a calculation time of around 75,000 years in large portfolio scenarios. The reason for this is not a lack of software, but the exponential increase in the decision space as the number of projects and the density of restrictions increase.

Example: Why is computing time exploding?

Imagine you have a portfolio with many projects and sub-projects. Each decision (project yes/no, sub-package A/B/C, sequence, budget limits, dependencies, risks) massively increases the number of possible combinations. From a certain size upwards, a search space is created that can no longer be fully enumerated with conventional computer architectures without the computing time growing to impracticable dimensions.

This is precisely where mAInthink's strength lies: We use hybrid AI and parallelized computation to deliver solutions with very high accuracy in practically relevant time - instead of theoretical perfection in millennia.

Looking to the future: quantum computers

Quantum computers would not replace this approach, but would further accelerate it. With increasing industrial availability, certain classes of optimization problems could be processed much faster. Based on the mathematical logic already established, mAInthink would once again be considerably faster.

Conclusion

mAInthink stands for scientifically sound decision-making intelligence - the result of decades of mathematical work and consistently enhanced by modern AI and computing technology.

It's not gut feeling that decides. Not simplified models. But calculable reality.

FAQ - Algorithmic project & investment optimization at mAInthink

Frequently asked questions

Who is Dr. Igor Kadoshchuk?

Dr. Igor Kadoshchuk is a mathematician and computer scientist who has been working scientifically on algorithms, mathematical optimization and computer-aided decision-making since the 1980s. He held a chair at the Moscow Institute of Physics and Technology (MIPT) and was a professor of computer science there.

What is the central finding of his research?

That mathematical methods and computer technology can be combined in such a way that it is possible to objectively calculate which investment path is best for complex project and investment decisions - regardless of subjective assessments.

Why are traditional tools such as Excel unsuitable for this?

Traditional tools work with simplifications, linear assumptions or isolated evaluations. They cannot fully calculate the exponentially growing number of project combinations, dependencies and restrictions.

What does "hybrid AI" mean at mAInthink?

Hybrid AI combines classical mathematics, heuristic optimization methods, algorithmic search and modern AI methods with parallel (multi-threaded) computing architecture. This allows very large decision spaces to be analyzed efficiently.

What accuracy does mAInthink achieve?

In real scenarios, mAInthink achieves an accuracy of around 97% to 99.99%. This represents the technically and economically optimal ratio between computing time and decision quality.

Why is 100% accuracy not aimed for?

A complete calculation of all possible combinations would require - depending on the scenario - up to 75,000 years of computing time. Such perfection is technically possible, but not practical or economically viable.

What is a simple example of this complexity?

Just a few projects with dependencies, budget limits, risks and alternatives create an exponential search space. Each additional variable multiplies the number of possible combinations.

What role do quantum computers play?

Quantum computers could speed up these calculations considerably in the future. The mathematical models remain the same, but the calculation of many states takes place in parallel. mAInthink is architecturally prepared for this.

For which use cases is mAInthink particularly suitable?

For portfolio optimization, investment decisions, project prioritization, budget allocation, strategic planning and scenarios with high complexity and many dependencies.

Comparison: classic tools vs. mAInthink

Criterion	Classic tools (e.g. Excel)	mAInthink
Calculation model	Linear, simplified	Hybrid: mathematics + AI + algorithms
Number of projects	Limited practicability	Scales to very large portfolios
Dependencies & restrictions	Manual or highly simplified	Fully integrable
Combinatorial depth	Very limited	Exponential decision spaces
Computing time	Fast, but incomplete	Fast with very high accuracy
Accuracy	Subjective / heuristic	97 % - 99,99 %
Future viability	Limited	Prepared for quantum computing

Why real decision costs are almost always higher than computing costs

In practice, the greatest economic damage is rarely caused by computing costs - but by wrong decisions: incorrectly prioritized projects, overoptimistic business cases or portfolios that look good on paper but are not viable under real-world restrictions.

This is exactly where mAInthink comes in: Mathematically-based optimization and hybrid AI are used not only to select "a good project", but also to determine the best investment path under budget, risk and dependency conditions. The crucial point here is:

Computing time costs minutes - misallocations cost months, years and often seven-figure sums.

Typical cost blocks that classic tools underestimate

Cost block	What is often missing in classic tools	Typical impact in reality	How mAInthink addresses this
Capital commitment	Capital is considered a "given"; opportunity costs are missing	Money is tied up even though a better way exists	Optimizes portfolio and sequence under budget constraints
Management & team capacity	Resources are modeled roughly or statically	Bottlenecks, delays, overload, project backlog	Considers capacities, dependencies and timing
Dependencies	Sub-projects are evaluated in isolation	"Good" projects fail because preliminary work is missing	Calculates optimal chains (predecessors/constraints) with
Risk & uncertainty	Risk is managed as an overhead or text field	Budget and deadline explosion, ROI collapses	Risk and scenario parameters are integrated mathematically
Implementation sequence	Sequence is decided "from experience"	Cash flow and ROI are realized later than necessary	Finds the sequence with maximum effect and minimum blockage
Opportunity costs	Not visible because only project ROI is considered	Missed market windows, missed economies of scale	Compares investment paths and shows lost benefits
Change costs & rework	Changes are not managed as a cost model	Rework, replanning, additional trades/partners	Evaluates robustness: solutions that generate less "rework"

Specific example: "quick decisions" are often expensive

A classic scenario from portfolio practice:

• Project A appears to be a TOP project because the ROI is highest on paper.
• However, project A ties up critical resources and budget early on.
• This delays two smaller projects (B and C), which together would deliver a faster cash flow and a more stable risk structure.

The result: Project A wins in Excel - in reality, the portfolio loses time, cash flow and flexibility.

mAInthink not only calculates "which project looks best", but also which investment path achieves the best overall effect under real restrictions.

Key point for decision-makers

Calculation time is a cost factor - wrong decisions are a multiplier.

Data in. Maximum ROI out. Without prompts. Without strategy creation.

The approach of mAInthink and the algorithmic solution StratePlan is deliberately clear and practical:

The customer delivers their project strategy - we deliver the optimization.

To do this, mAInthink only requires the customer's existing planning data, e.g. as:

• XLS / Excel files
• JSON files

No prompts, no text-based AI instructions and no semantic interpretations are required. StratePlan works based on data and models - not prompt-driven.

Important principle: Strategy comes from the customer

mAInthink does not create a project strategy. This is a conscious and central design decision.

Why? Because the CEO, CFO, project manager or C-Level:

• know their markets
• understand their risks
• can assess regulatory, political and operational framework conditions

No AI can or should replace this market and context knowledge.

Our task is different:
We validate the existing strategy - and usually optimize it.

Validation & optimization instead of rethinking

In over 95% of cases, it turns out that existing project or investment strategies are

• are not optimally prioritized mathematically
• Do not fully consider dependencies
• Opportunity costs are overlooked

The result:

An optimization of typically 10 % to 60 %+
in terms of ROI, cash flow timing or risk structure - without changing the content of the strategy.

Dynamic markets = dynamic strategy

Markets change. Budgets change. Risks shift.

This is why the strategy creator can

• upload an adjusted plan
• integrate new assumptions or restrictions
• reflect a changed market situation

StratePlan then recalculates the optimized or validated strategy.

In this sense, StratePlan is a kind of business GPS:
Regardless of whether it is a price adjustment, market change or new framework conditions - the system calculates the best starting position for well-founded CEO decisions at all times.

Why the "ROI doesn't stand up to reality" argument doesn't work

A common argument is that optimized ROIs can shrink in reality due to external circumstances.

This is correct - but applies to any method, including traditional tools.

The decisive difference:

Scenario	Classic planning	StratePlan optimization
Calculated ROI (planning)	7 %	35 %
External influences during implementation	-4 %	-8 %
Real ROI after implementation	3 %	27 %

Both approaches are subject to market changes. The difference is the starting point.

Even if part of the optimized ROI is lost due to external factors, the result usually remains well above the level of classic calculations.

Conclusion

StratePlan does not replace a strategy - it makes it better.

Your strategy remains your strategy.
Our algorithms ensure that you get the most out of it under real restrictions.

Data in. Maximum ROI out.

External studies confirm the paradigm shift

Leading economic and research institutes have independently come to a clear clear conclusion: traditional financial and planning models systematically reach their systematically reach their limits when it comes to complex investment decisions.

Studies by McKinsey & Company, Bain & Company and the OECD show that companies with data- and model-based capital Significantly better results than those that rely on isolated project valuations or linear isolated project evaluations or linear Excel models.

Dr. Igor Kadoshchuk 's research on NP-Hard Financial Management Problems provides the mathematical Many real investment decisions are exact optimization problems, which cannot be fully solved using classical methods.

Selected References

• McKinsey & Company (2023). Optimized Capital Allocation Report.
• PwC (2022). Risk Management Strategies for Competitive Advantage.
• Kadoshchuk, I.T. (2021). Experimental Algorithms for Solving NP-Hard Financial Management Problems.
• Kadoshchuk, I.T. (2018). The vertices we choose.
• Harvard Business Review (2021, 2023).
• MIT Sloan Management Review (2023).
• World Economic Forum (2022).

Sources & links

1. World Economic Forum (2023) - How artificial intelligence will transform decision-making
   https://www.weforum.org/stories/2023/09/how-artificial-intelligence-will-transform-decision-making/
2. World Economic Forum (2025) - Investment companies can use AI responsibly to gain an advantage
   https://www.weforum.org/stories/2025/02/ai-redefine-investment-strategy-generate-value-financial-firms/
3. World Economic Forum (2025) - Artificial Intelligence in Financial Services (PDF Report)
   https://reports.weforum.org/docs/WEF_Artificial_Intelligence_in_Financial_Services_2025.pdf
4. Bain & Company (2025) - The Future of Financial Planning Is Autonomous
   https://www.bain.com/insights/the-future-of-financial-planning-is-autonomous/
5. SSRN (2023) - The Role of Artificial Intelligence in Financial Decision... (Abstract/Download Page)
   https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4628237
6. Academic PDF (secondary reference) - Data-driven decision making (PDF host)
   https://prosiding.areai.or.id/index.php/ICEAT/article/download/54/79/322
7. Al-Basaer Journal (secondary research PDF) - AI integration in financial planning (PDF download)
   https://www.albasaer.org/index.php/abjbr/article/download/21/6/192

StratePlan in practice: What traditional planning cannot achieve

Many companies and organizations have good data, experienced decision-makers and established planning routines. Despite this, sub-optimal portfolios, delayed implementation and unnecessary capital commitment regularly occur. The reason is rarely a lack of information - but rather the limits of classic tools and thought models in the face of high complexity.

1) Project and financial planning is a calculation problem - not a gut feeling

Traditional financial planning often fails due to structural factors: fragmented decisions, uncoordinated priorities, Simplifications in models and emotional or politically driven individual decisions. In complex project landscapes the result is not "wrong", but rarely optimal.

This is precisely where StratePlan comes in: It maps decisions as a calculable model and optimizes capital and project allocation under real restrictions.

2) Why optimization quickly becomes "NP-Hard"

Real project and investment decisions are rarely linear. As soon as dependencies, budgets, capacities, timing, risks and Alternatives (e.g. project variants) come together, the search space grows exponentially. Many of these problem classes are NP-hard - this means that an exact calculation of all combinations is theoretically possible, but often not realistic in practice.

Consequence: If you still want to calculate "completely exactly", you pay with impractical computing time.

3) Why 100% accuracy does not make sense in practice

A 100% calculation would require every possible combination to be completely enumerated and evaluated. From a certain Order of magnitude, this becomes pure theory. That is why StratePlan relies on a combination of mathematical methods and experimental/hybrid optimization algorithms that deliver very high levels of accuracy in practice - with a practicable Computing time.

The result: decisions are not calculated "somehow faster", but in a depth that classic tools typically do not typically achieve.

4) Hybrid algorithms instead of Excel logic or prompt AI

StratePlan is not a generative text AI. It does not interpret prompts and does not generate "probable answers". The system works on the basis of data and models and combines

• classical mathematical optimization
• algorithmic search and heuristics
• scalable parallelization (multi-threaded computing)

This results in an optimization system that calculates consistently - instead of "guessing".

5) Speed is a success factor - not just a convenience feature

In dynamic markets, it is not only the best portfolio that counts, but also the right timing. StratePlan accelerates decision-making by quickly calculating complex scenarios and enabling iterative optimization.

Practical benefits: Opportunities can be exploited before framework conditions change again.

6) StratePlan as a validation and optimization layer (strategy remains at C-level)

A central principle: mAInthink does not create a project strategy. A CEO, CFO or project manager can do this better because they know the markets, Goals, political framework conditions and operational constraints.

The customer delivers their strategy as a data model - typically as XLS or JSON:

• Data in: Project list, budgets, dependencies, restrictions, targets
• Value out: validation, prioritization, optimal sequence, budget allocation, scenario results

In practice, it is very often the case that good strategies can be measurably improved through optimization (e.g. through better sequencing, recognition of hidden opportunity costs or more robust structuring against risks).

7) Iteration instead of a one-off plan: StratePlan as a "business GPS"

Markets, costs, supply chains, interest rates and political conditions change. Therefore, a strategy does not have to be "perfect once", but must be continuously updated.

In this sense, StratePlan is a business GPS:

• Adapt strategy
• upload new file
• recalculate
• again obtain an optimized starting point for decisions

In this way, planning remains capable of acting even in the event of course changes and new constraints.

8) ROI is dynamic - this applies to all methods (the difference is the starting point)

A typical counter-argument is that optimizations can shrink in reality due to external circumstances. This is true - but it applies for every planning method, including classic tools. The decisive factor is the starting point.

Example	Classic planning	StratePlan-optimized
Calculated ROI (plan)	9 %	42 %
External influences during implementation	-5 %	-10 %
Real ROI after implementation	4 %	32 %

Both approaches experience deviations from reality. The difference is that a higher, optimized starting position usually remains above the result of usually remains above the result of classic calculations.

9) "Zero hallucinations" - because StratePlan calculates instead of interpreting

StratePlan does not hallucinate because it does not "answer" text-based. It does not generate free interpretations, but calculates Results from a defined data model (IDs, numbers, restrictions). This means that the output is deterministically traceable and can be checked internally.

Contact us now