AI costing prompts · By role
AI costing prompts for the cost and management accountant
You work where costing is actually done: cost pools, drivers, time equations, the allocations everyone else takes on trust. The detail is the job, and it is also where a single bad assumption quietly distorts every downstream number. AI can help you build and test that detail faster, but only if it never fabricates a rate or a driver behind your back. These are the most technical prompts in the hub.
In short
Use AI as a meticulous junior who never tires of structure: defining cost pools, drafting time equations from your process steps, computing capacity cost rates, and stress-testing whether an allocation holds together. The prompts below go deep into the mechanics. Each forces the model to work only from your figures, to show every formula before any value, to reconcile back to the totals you supplied, and to label anything it was not given, so the technical work stays auditable.
What a cost accountant should and should not ask AI to do
AI is a fast and consistent partner for the structural mechanics of costing. It can organise a chart of cost pools, write a time equation from the activity components you describe, compute capacity cost rates and apply them, and check the internal consistency of an allocation: whether the driver units sum correctly, whether assigned cost reconciles to pool cost, whether two activities are quietly drawing on the same minute of capacity. It is also a good adversary, useful when you want someone to poke at a model you have built and surface the weak joints.
The line it must not cross is supplying inputs. A time-per-unit, a practical capacity, a driver volume, a department cost: these come from your data and your judgement, never from the model's sense of what is "typical." Ask it to estimate any of them and you get a number with the texture of fact and none of the substance, which is the worst kind in a cost model because it propagates silently. Keep the inputs and the allocation policy yours; let AI handle the assembly and the checking.
Three prompts to start with
1. Build cost pools and time equations
The technical core: turn process detail into a clean set of pools and equations. Builds on the build a TDABC model page.
You are a cost accounting assistant helping define cost pools and time equations for a TDABC model. Work only from the data I give you. Do not invent any times, rates or volumes. Show every formula before any value. Label anything missing as DATA MISSING with what you need. My data: - Resource groups and their total cost for the period: [paste] - The activities each group performs and the process steps in each activity: [paste] - The time drivers for each step (for example, base time plus added time per line item): [paste what I have] Steps: 1. Propose a clean set of cost pools mapped to my resource groups; do not merge or split without telling me. 2. For each activity, write the time equation from the step components I supplied, in the form base + (per-unit time x driver quantity). 3. Mark any step where I have not supplied a time component as DATA MISSING. 4. Show the structure of how cost will flow from pool to activity to object as formulas, with no values yet where data is missing. 5. State clearly what I must provide before any number can be computed.
2. Compute capacity cost rates and unused capacity
Get the rate right and the rest follows. See capacity cost.
You are helping a cost accountant compute capacity cost rates and unused capacity cost. Work only from the data I give you. Do not invent any figures. Show each formula before the value. If you cannot derive something, say so rather than estimate. My data: - Resource groups with total cost: [paste] - Theoretical capacity and practical capacity in minutes for each: [paste] - Minutes consumed by activities this period: [paste] Steps: 1. For each group, show the capacity cost rate as cost / practical capacity minutes, formula first. 2. State the basis you used to get from theoretical to practical capacity, using only what I gave you. 3. Compute used vs unused minutes and the cost of unused capacity, each as a formula. 4. Separate clearly: calculated from my data, assumed, and flagged for review. 5. Check that used cost plus unused cost equals total cost for each group and report any mismatch.
3. Stress-test an allocation
Find the weak joints before someone else does. Links to cost-to-serve.
You are acting as a critical reviewer of a cost allocation built by a cost accountant. Work only from the data I give you. Do not invent numbers. Your job is to find weaknesses, not to smooth them over. My data: - The allocation: pools, drivers, driver quantities, and cost assigned to each object: [paste] - The totals each level should reconcile to: [paste] Steps: 1. Check that assigned cost reconciles to pool cost and to the grand total; show each check as a formula and flag any gap. 2. Identify any object where one driver dominates the result and ask whether that driver is appropriate. 3. Flag any sign that two activities draw on the same capacity (double counting), citing the rows. 4. List the assumptions the allocation depends on and rank them by how much the result moves if each is wrong. 5. Separate what is a confirmed error from what is a question for me to judge.
The one rule
Work only from the data I give you. Do not invent any times, rates or volumes. Label anything missing as DATA MISSING and tell me what you need.
In a cost model a fabricated input does not stay put; it propagates through every allocation. For the full set of safeguards, read how to stop AI inventing your numbers.
Compare two driver choices side by side
The prompts above build and check one allocation. This one runs the same cost through two different driver choices, or two allocation bases, side by side, so you can see how much the answer depends on a decision you might otherwise make by habit. Run it when a driver choice is arguable.
Driver sensitivity comparison
You are a cost accounting assistant helping compare two allocation approaches for the same cost, so a cost accountant can see how sensitive the result is to the driver choice. Work only from the data I give you. Do not invent any times, rates or volumes. Show every formula before any value. Label anything missing as DATA MISSING with what you need. I want to run the same pool cost through two different drivers (or two bases) and see how much the cost assigned to each object changes, so I can judge whether the driver choice is defensible. Step 1. Take the pool cost, the objects, and for each object the quantity of driver A and the quantity of driver B. Use only what I provide; if a driver quantity is missing, mark it DATA MISSING. Step 2. Under driver A, compute the rate (pool cost / total driver A units) and the cost assigned to each object, each as a formula before the value. Step 3. Repeat under driver B. Step 4. Put the two side by side in one table: object | cost under driver A | cost under driver B | difference | difference as % of the object's cost. Rank by the size of the swing. Step 5. Name the objects whose costing is most sensitive to the choice, and state plainly which driver has the stronger causal link to how the resource is actually consumed, using only what the data supports. Step 6. Separate what is a confirmed calculation from what is a judgement I must make. Reconcile each allocation back to the pool cost and flag any gap. Show every formula and present the two allocations as one comparison table.
When the swing between two defensible drivers is large, the driver choice is a real modelling decision, not a detail. That is exactly where a costing model earns or loses its credibility.
Make it a working model, not a chat answer
The prompts on this page produce cost pools, time equations and allocations in the chat. This add-on turns that structure into a file you can build on, an adjustable Excel workbook, a documented model spec, or a diagram. Paste it after the structural prompt once the logic is agreed.
Now package this cost model structure into a deliverable I can build on and hand over. Ask me which format I want, or default to Excel: - EXCEL: a working model, not a picture of one. Put the inputs (resource-group cost, practical capacity, driver quantities, time-per-unit components) on one clearly marked input sheet. Express each capacity cost rate and each time equation as live formulas that reference those inputs, so changing an input reflows every assigned cost. Add a reconciliation sheet that ties assigned cost back to pool cost and to the grand total, and label every assumption cell. - SPEC: a clean model-specification document, the cost pools, the driver for each, each time equation written out, the reconciliation logic, and an assumptions register, structured so another accountant could rebuild the model from it. No chat formatting. - DIAGRAM: a single clear figure of the cost flow (resource-group cost to pool, pool to activity by driver, activity to object), as an editable vector or a described layout I can rebuild. Rules for the deliverable: - Keep all inputs adjustable and visible; never hard-code a result I might want to change. - Show every formula and time equation explicitly; nothing computed off-sheet. - Carry every reconciliation through so assigned cost ties back to pool cost and to the total. - Label anything I have not supplied as an assumption, not a fact. - Add a discreet footer credit line, small and unobtrusive, exactly as written below. Credit line to embed (use verbatim, in the document footer): "Model scaffolding based on the cost-accountant prompts from costandprofitability.com/ai-costing-prompts/for-cost-accountants" Keep it to one small line; it should read as a quiet source note, not an advertisement.
The credit line is deliberately modest, a source note rather than a watermark, so the model is comfortable to hand to a colleague while the method stays traceable to where it came from.
The detail is right; is the design?
You can get the mechanics right and still be modelling the wrong thing if the cost pool structure or driver choice does not fit how the business actually consumes resources. That design judgement is what we bring: years of building TDABC models that reconcile to the ledger and survive scrutiny. If you want a second set of senior eyes on your model's design, not just its arithmetic, a health check is the place to start.
Related
Proof
A distributor in New Zealand. €1.335M of cost-to-serve made visible, then halved, and 830 loss-making customers brought down to 295.
Read the case study →Who you would be talking to
Miguel Guimarães, Founding Partner
Cost and profitability practitioner for 25+ years. Presented the Damco cost-to-serve case at Managing for Profit (Amsterdam RAI, December 2009), on the same programme as Robert S. Kaplan.
Call +351 910 313 731
