Gains:
- Ability to construct the final pit limit, pushback and production sequence problems on a scenario basis with AI
- Ability to analyze the balance between parameters such as stripping rate, NPV and slope angle with AI support
- Ability to verify the AI suggested plan against optimization software, geotechnical constraint and safety requirement
Once we know where and how much of the ore is, it's time for the most difficult question: how, in what order and to what limits will we dig it? January planning; The block model is the task of coming up with the excavation plan that produces the highest value safely by bringing together costs, metal price, slope safety and time value of money. In open pit (excavation above ground in steps), the final pit boundary, pushbacks (intermediate boundaries that gradually expand the pit) and production sequence are decisive; In the underground, method selection, panel/foot design and support (the support that holds the excavation cavity up) come first. This is an optimization problem where multiple options are compared. AI is strong in scenario generation and parameter analysis; But the final plan, slope angle and safety decision belong to the competent engineer.
Basic concepts
- Block model: Numerical model of an ore deposit divided into small rectangular blocks, with grade, density and type assigned to each block.
- NPV (Net Present Value): The sum of future cash flows discounted to present value; The basic criterion by which plans are compared.
- Stripping ratio: The amount of waste rock removed to reach one unit of ore; determines the economy and the mine limit.
- Ultimate pit limit: The last limit where digging stops economically; It is classically found with algorithms such as Lerchs-Grossmann.
- Sequencing (scheduling): Which block will be excavated in which year; Balances constraints (capacity, blending, slope) by maximizing NPV.
AI is helpful in teaching these concepts and setting up scenarios; however, the final pit and sequence are produced with specialized optimization software (e.g. sectoral planning packages) and geotechnical constraints. An NPV or stripping rate number that AI gives is just an estimate unless calculated with your own block model and cost structure.
Step by step: putting AI into the planning flow
- Clarify inputs. Block model, costs (excavation, transportation, processing), metal price, discount rate, slope angles, capacity. AI: extracts list of missing input.
- Define scenario. Different metal price, different cut-off, different production capacity. AI: builds scenario matrix.
- Set up an account. Code for block value, stripping rate, simple NPV calculation. AI: generates script, interprets unit and assumption.
- Verify with optimization software. The final furnace and sequencing are run in specialist software. AI: aids in result interpretation.
- Apply geotechnical and safety constraint. Slope angle, step height, safety berms. AI: generates checklist; The decision is made by the geotechnical engineer.
- Report. Comparison tables for the decision meeting. AI: draft writer.
Tip: Don't make the AI "suggest" the bevel angle. The slope angle is determined by geotechnical analysis (rock mass, discontinuities, water) and is safety-critical. Use AI here only to compare scenario economics.
The difference of underground planning
In the underground, the problem is mostly method selection (chamber-toe, longwall, lower/upper floor caving etc.), opening stability, support and ventilation. Here safety is even more dominant: ceiling/wall stability, risk of collapse, gas and ventilation. AI method can produce comparison table and cost scenario; But the decision on method and support rests with the geotechnical and mining design engineer. An AI inference such as “a tall leg is suitable for this bed” is merely a hypothesis that sparks discussion.
three mini cases
Case 1 — Stripping rate scenario. In an open coal mine, the team wants to see how the ultimate limit changes at 3 different levels of coal price. They have the AI install a script that calculates the block value and scraping rate; They run it on their own data. When the price drops by 20%, they see that the economic limit narrows and the total mineable reserves decrease from 4.1 Mt to 3.3 Mt. AI has accelerated calculus; The final boundary and decision were made with planning software and engineers.
Case 2 — Incorrect NPV assumption. An intern asks the AI “is this mine profitable”; AI gives an example NPV. The intern puts this in the presentation. However, AI did not know the discount rate, real costs and production profile; The number he gives is completely fake. The correct approach is to calculate NPV with its own cash flow statement. Lesson: AI doesn't need to be asked "is it profitable?" The AI is asked how to set up a cash flow account, the numbers come from you.
Case 3 — Blending constraint. In a forge, the plant requires feed below a certain SiO2 limit. The team tells the AI to write a script that calculates collation scenarios from different block sources; They see that the border will hold if which regions are mixed in which month. The planning engineer verifies the result in the software and adjusts the production sequence accordingly. AI has accelerated constrained scenario arithmetic; The constraint and decision were the engineer's.
Copiable prompt templates
LACK OF INPUT CHECK "Role: You are the open pit planning assistant. What inputs do I need for the final pit boundary and production sequence? Create a complete checklist: block model areas, cost items, economic parameters, geotechnical constraints, capacities. Write in one sentence why each item is needed."
NPV CALCULATION SKELETON "Set up a simple annual cash flow and NPV calculation in Python. Inputs: annual ore tonnage, grade, metal price, yield, unit costs, discount rate. Interpret ALL assumptions upfront. DO NOT make up the number; I will provide the inputs. Finally, tabulate the cash flow and NPV year by year. Add an option that changes the price by ±20% for sensitivity."
SCENARIO COMPARISON"Set up a table comparing results for the following scenarios: [scenario1: price
GEOTECHNICAL/SAFETY CONSTRAINT CHECK "Produce a geotechnical and safety checklist for a draft open pit plan: slope angle verification, step height, safety bar, ramp slope, drainage, monitoring. Add note 'who makes this decision / what analysis is required' for each item. Emphasize that this list cannot be DECISIONED by AI, geotechnical engineer approval is required."
Weak prompt / Strong prompt
WEAK PROMPT: "Give me the best pit plan for this mine."
STRONG PROMPT: "Role: You are the planning assistant. I will work with my own block model, costs, and geotechnical slope angles. Give me (1) a list of missing inputs, (2) a code skeleton for the stripping rate and block value calculation, (3) a tabular structure comparing three price scenarios. YOU determine the final pit boundary; note that this will be generated by the optimization software and geotechnical constraint."
Comparison table: decision and AI role
decision
decisive
AI role
verification
Final pit limit
Economics + geotechnics
Scenario/account
planning software
slope angle
Geotechnical analysis
Checklist only
geotechnical engineer
Production sequence
Capacity + NPV
arithmetic helper
Software + engineer
blending
Facility constraint
Scenario account
Metallurgy + planning
underground method
Design + security
comparison chart
mining design engineer
Common mistakes
- Making the AI ask “is it profitable/what is the best plan?” AI does not know your cost and discount data; Calculate NPV with your own table.
- Have AI suggest the slope angle. Security-critical; determined solely by geotechnical analysis.
- Focusing on a single scenario. Price and cost are uncertain; Sensitivity analysis is essential.
- Sacrificing geotechnical constraint to economy. A steeper slope means less rust, but the safety margin is non-negotiable.
- Blindly accepting the optimization output. The software is only as good as the quality of your input; manually logic check the results.
Attention: When economy and security conflict in planning, security wins. Even if the AI suggests a steep slope that seems economically better, the geotechnical safety limit will not be gone below.
In summary
January planning; The block model is an optimization job that combines economy, security and time. AI input control, stripping rate and NPV calculation skeleton is a powerful aid in scenario comparison and blending arithmetic. However, the final pit boundary is determined by the optimization software, and the slope angle and underground method are determined by the geotechnical/design engineer. Don't ask AI "is it profitable" or "what is the best plan"; Generate numbers from your own data and never sacrifice security for economy.
Application task
Set up an annual cash flow and NPV calculation using the "NPV calculation framework" template with your own (or sample) block model and cost data; Change the price by ±20% and see the sensitivity. Then chart the three scenarios with the "Scenario comparison" template. Finally, create a safety checklist for your plan draft with the “Geotechnical/safety constraint check” template and note who made the decision for each item.
checklist
- [ ] I didn't ask AI for NPV, I calculated it with my own cash flow statement.
- [ ] I determined the slope angle through geotechnical analysis, I did not have it suggested by AI.
- [ ] I compared at least three scenarios (price/cut-off/capacity).
- [ ] I put geotechnical/safety constraints before economics.
- [ ] I produced the final furnace boundary with planning software and passed it through logic checking.
- [ ] I left final plan approval to the competent engineer.