From Generative AI to Decision Computation: When Answers Are No Longer Enough
Generative AI can explain how a decision should be made. Decision Computation calculates which decision yields the highest objective value under defined conditions.
Today, ChatGPT, Gemini, and other generative AI systems can analyze, structure, and explain complex management questions within seconds.
But for CEOs, CFOs, and investment committees, the real work often begins exactly where the generic answer ends.
Which of our projects should we actually fund?
Which combination of projects generates the highest value given our budget?
What happens to our portfolio if CAPEX drops by 20 percent?
Which projects should we postpone if engineering capacity becomes a bottleneck?
These questions need no further explanation.
They require a calculation.
This is the transition from Answer Generation to Decision Computation.
Table of Contents
- What Is Answer Generation?
- What is Decision Computation?
- What is the difference between Generative AI and Decision Computation?
- Why is Decision Computation Relevant for CEOs and CFOs?
- 20 Management Questions That Can Be Calculated
- Why Portfolio Decisions Are Mathematical Problems
- Why the decision ultimately rests with management
- Decision Computation in the Boardroom
- Bringing together generative AI and mathematical optimization
- Decision Computation with StratePlan
- FAQ on Decision Computation
What Is Answer Generation?
Answer Generation refers to the ability of generative AI to generate a linguistic response based on information, context, and a question.
For example, a CFO might ask:
“How should I prioritize CAPEX projects with a limited budget?”
Generative AI can provide a well-reasoned answer:
Evaluate projects from an economic perspective. Analyze NPV and ROI. Consider strategic criteria. Assess risks. Take resource constraints into account. Compare scenarios.
This answers the question:
How should I proceed in general?
But it doesn’t yet answer:
Which specific combination of our projects should we choose?
What is decision computation?
Decision computation refers to the mathematical calculation of decision options based on specific data, objectives, and constraints.
Instead of merely explaining how a decision could be made, the actual decision space is calculated.
For example:
- 187 investment projects
- €420 million in available CAPEX
- 31,000 engineering hours
- 14 mandatory projects
- 23 project dependencies
- Multiple plants
- multiple business units
- Strategic criteria
- Multiple planning periods
The question is no longer:
“How does capital allocation work?”
But rather:
“Which combination of these 187 projects satisfies our constraints and maximizes our defined objective function?”
This is a mathematical optimization problem.
What is the difference between generative AI and decision computation?
| Generative AI | Decision Computation |
|---|---|
| explains | computes |
| generates answers | calculates decision options |
| works primarily with language and context | works with decision variables, objective functions, and constraints |
| answers “How could one proceed?” | answers “Which combination best meets our defined goal?” |
| structures knowledge | structures the decision space |
| can explain methods | applies mathematical optimization to concrete data |
| provides arguments and analyses | provides calculated scenarios and portfolio alternatives |
Generative AI answers questions about decisions.
Decision Computation calculates the consequences of specific decisions.
These two approaches do not necessarily compete with each other.
They can cover different levels of the same decision-making process.
Why is Decision Computation relevant for CEOs and CFOs?
Many management decisions are not knowledge problems.
The organization knows its projects.
Finance knows the budget.
Controlling knows the business cases.
Operations knows the resources.
Strategy knows the priorities.
The problem lies in translating all this information simultaneously into a consistent portfolio decision.
The challenge isn’t a lack of information.
The challenge is combining them.
This is precisely why capital allocation, CAPEX planning, and portfolio selection become mathematical problems as the number of projects increases.
20 Management Questions That Can Be Calculated
1. Which CAPEX projects should we finance?
Answer: The combination of projects that best fulfills the selected objective function under the defined budget, resource, dependency, and governance constraints.
2. Which combination of projects generates the highest value with our existing budget?
Answer: Not necessarily the projects with the highest individual values. What matters is the total value of the permissible combination.
3. Which projects should we cut if our CAPEX decreases by 20 percent?
Answer: Not a flat 20 percent from every area. The portfolio should be recalculated based on the reduced total budget.
4. How much value do we actually lose with a CAPEX cut?
Answer: This is determined by the difference between the optimized portfolios at the respective budget levels.
5. Which projects should we finance if we have additional CAPEX?
Answer: The additional combination of projects that generates the highest marginal portfolio value under the remaining constraints.
6. How much additional CAPEX is worthwhile?
Answer: By calculating multiple budget levels, we can determine how much additional portfolio value is generated by additional capital and where the marginal benefit begins to decline.
7. How should we allocate CAPEX across business units?
Answer: Not necessarily according to historical budget allocations. The allocation can be based on the optimal combination of projects across all business units.
8. How should we allocate CAPEX across plants?
Answer: By jointly optimizing the project universe while taking into account plant-specific minimum requirements, capacities, and constraints.
9. Which combination of projects maximizes our NPV?
Answer: The permissible combination with the highest aggregate NPV within the defined constraints.
10. Which combination of projects maximizes our ROI?
Answer: That depends on the exact definition of the portfolio objective function. A simple ranking based on project ROI is not automatically identical to an optimal overall portfolio.
11. What happens if growth becomes more important?
Answer: The strategic weights are adjusted, and the portfolio is recalculated. This reveals which capital allocation aligns with the revised strategy.
12. What happens if our strategic priorities change?
Answer: Different weightings produce different optimal portfolios. This makes the effects of strategic decisions quantifiable.
13. What happens if engineering capacity is reduced by 20 percent?
Answer: Engineering is modeled as a more severe resource constraint. The optimal project combination is then recalculated.
14. Which projects tie up a particularly large amount of capital or resources?
Answer: Portfolio analysis shows which projects consume large amounts of scarce resources and which alternative project combinations are consequently ruled out.
15. Which projects should we start today, and which should we postpone?
Answer: In a multi-year optimization, project start dates, budget utilization, resource requirements, and dependencies are considered collectively across multiple periods.
16. What does our optimal 5-year CAPEX plan look like?
Answer: It is a multi-period optimization that determines which projects should be implemented in which year to best fulfill the defined objective function within the annual constraints.
17. What happens if certain projects become mandatory?
Answer: Mandatory projects are set as fixed constraints. The remaining portfolio is re-optimized around these projects.
18. What happens when there are project dependencies?
Answer: Dependencies are modeled mathematically. For example, if Project B depends on Project A, selecting B without A is not permitted.
19. Is our currently approved CAPEX portfolio optimal?
Answer: This can be determined by comparing the existing portfolio with a mathematically optimized portfolio under the same data, objectives, and constraints.
20. How much more value can we generate with the same projects and the same budget?
Answer: The difference between the existing and optimized portfolios shows the calculated potential for improvement within the defined model.
Same projects. Different combinations. Better results.
Why Portfolio Decisions Are Mathematical Problems
The reason lies in the number of possible combinations.
With N independent projects, each of which can be selected or not selected, there are theoretically:
2^N possible combinations.
| Number of projects | Theoretically possible combinations |
|---|---|
| 20 | 1,048,576 |
| 50 | ≈ 1.13 × 10^15 |
| 100 | ≈ 1.27 × 10^30 |
| 200 | ≈ 1.61 × 10^60 |
However, the management decision does not consist of blindly considering all combinations.
It consists of identifying, among the permissible combinations, those that best satisfy the defined objective function.
To do this, constraints are taken into account:
- CAPEX budget
- Revenue / Expenses
- Engineering capacity
- Human Resources
- Mandatory projects
- Project Dependencies
- Mutual Exclusions
- Strategic Guidelines
- Minimum Investments
- Maximum budgets
- Plant capacities
- Business unit rules
- Planning periods
A project list thus becomes a mathematical decision space.
Why the decision remains with management
Decision computation does not mean that mathematics takes over management decisions.
Mathematics can only optimize within the model defined by the company.
Management determines:
- which goals are relevant,
- which data is used,
- which strategic criteria apply,
- which constraints are mandatory,
- which projects are mandatory,
- which scenarios are considered.
The calculation then answers the following questions:
What are the consequences of these specifications?
This creates a clear distinction:
Management defines the decision-making framework.
Mathematics calculates the decision space.
Management makes the decision.
Decision Computation in the Boardroom
This approach becomes particularly relevant when computation and management discussion are no longer separated in time.
Traditionally, a new management question can trigger a new analysis cycle.
Data is adjusted.
Finance recalculates.
Presentations are updated.
Another meeting is scheduled.
If, on the other hand, portfolio scenarios can be recalculated immediately, the process changes.
The CFO asks:
“What happens if the budget is reduced by 50 million euros?”
Calculate.
The CEO asks:
“What happens if growth is weighted twice as heavily?”
Calculate.
Operations says:
“That engineering capacity won’t be available next year.”
Calculate.
The Investment Committee decides:
“These three projects are mandatory.”
Calculate.
After that, the scenarios can be compared directly with one another.
Question. Calculate. Compare. Decide.
Combining Generative AI and Mathematical Optimization
Generative AI and mathematical optimization serve different purposes.
Generative AI is particularly well-suited for:
- language-based interaction
- Information processing
- Explanation
- Structuring
- Interpretation
- Summary
Mathematical optimization is particularly well-suited for:
- Combinatorial decision problems
- Objective functions
- Budget constraints
- Resource constraints
- Dependencies
- Portfolio Selection
- Scenario Analyses
The interesting architecture therefore does not arise from the question:
Generative AI or mathematical optimization?
But rather:
Which task should be handled by which method?
One possible division of labor is:
Ask.
Management formulates the problem.
Understand.
AI assists with context, structure, and interpretation.
Calculate.
Mathematical optimization calculates the viable portfolio alternatives.
Compare.
Management compares scenarios and trade-offs.
Decide.
The decision rests with the responsible management.
Decision Computation with StratePlan
StratePlan was developed for complex capital allocation, CAPEX, and portfolio decisions.
The initial data can be relatively simple:
- Project ID
- Investment
- Expected value, NPV, or benefit
Depending on the decision-making problem, additional dimensions can be added:
- Strategic criteria
- Resources
- Project dependencies
- Mandatory projects
- Divisions
- Business Units
- Regions
- Planning Periods
StratePlan uses this information to calculate portfolio alternatives based on the defined objectives and constraints.
If the budget, resources, or strategic guidelines are changed, the portfolio can be recalculated.
This transforms static investment planning into an interactive decision-making process.
AI generates answers.
StratePlan calculates decisions.
Don’t just rank projects. Calculate the portfolio.
Don't take our word for it. Calculate it yourself.
FAQ on Decision Computation
What does “Decision Computation” mean?
Decision Computation refers to the mathematical calculation of decision options based on defined data, objectives, and constraints. In a portfolio context, this means, for example, calculating a combination of projects from among many possible projects that best fulfills a defined objective function within budget and resource constraints.
What is the difference between Decision Intelligence and Decision Computation?
Decision Intelligence is the broader approach to systematically improving decisions through data, models, processes, and technology. Decision Computation can be understood as the computational layer within this approach: the specific decision space is mathematically modeled and computed.
What is the difference between Generative AI and Decision Computation?
Generative AI generates answers or content based on context and information. Decision Computation calculates specific decision options based on defined variables, objective functions, and constraints.
Can ChatGPT or Gemini optimize CAPEX projects?
Generative AI can explain methods, interpret data, structure models, and assist in formulating an optimization problem. However, for robust portfolio optimization, the specific company data, objective functions, and constraints must be processed within a suitable mathematical model.
Does mathematical portfolio optimization replace the CFO or the investment committee?
No. Management defines objectives, strategic priorities, assumptions, and constraints. Mathematical optimization calculates the consequences of these specifications. The final decision remains with the responsible management.
Why isn’t a ranking of projects sufficient?
A ranking evaluates projects individually. Portfolio optimization, on the other hand, considers the combination of projects. Due to varying investments, resource requirements, and dependencies, the best combination of projects may differ from the order of a simple ranking.
Can Decision Computation help with budget cuts?
Yes. A portfolio can be recalculated under a reduced budget and compared with the original portfolio. This reveals which combination of projects best maintains the defined target value under the new conditions.
Can Decision Computation account for multiple business units and plants?
Yes. Business units, plants, regions, or other organizational dimensions can be included in a portfolio model as attributes and constraints.
Can Decision Computation be used for multiple years?
Yes. In multi-year optimization, budgets, resources, project durations, and dependencies are modeled across multiple periods. This allows you to decide not only which projects to implement but also when.
What is the goal of Decision Computation?
The goal is not to automate management decisions. The goal is to make the relevant decision space mathematically transparent and to make the consequences of different management directives calculable and comparable.
What is the central question behind Decision Computation?
It’s not just: What should we do?
But rather: What are the concrete outcomes of our projects, our budget, our resources, and our strategic goals?
It is precisely at this point that an answer becomes a calculation.
From Generative AI to Decision Computation.
