iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
To stop an AI tool from hardcoding values, tell it to place every changeable assumption in a labelled input area and have each formula refer to those cells. Then check the workbook yourself: look for numbers typed inside formulas, confirm that formulas stay consistent across forecast periods, and make sure the model’s internal checks pass in every period. Treat the generated file as a draft, not a finished model.
What counts as hardcoding
Hardcoding means a fixed value embedded directly inside a formula. ICAEW’s Financial Modelling Code (2024 edition, marked 08/24) gives a tax rate typed straight into a calculation as the standard example. The problem is not the number itself. It is that the number may need to change, and a user looking at the output has no obvious place to update it.
A manually entered value sitting in a clearly labelled input cell is not hardcoding in this sense. It is an assumption, and it is usually exactly where it should be. The distinction matters because AI tools often produce both kinds of number, and only one of them is a defect.
Free tools Windows power users keep installed
One-click scans. No signup required.
| Pattern in the generated file | Why it is a problem | Better construction |
|---|---|---|
Revenue formula ends in *1.08 |
The 8% growth rate cannot be changed without editing every formula that uses it | Growth rate in a labelled input cell; formula points to that cell |
Tax line uses *0.25 in each forecast column |
Changing the rate means editing every column, and a missed column produces inconsistent results | One tax rate input with source and date; all columns reference it |
A start date typed inside a formula with DATE(2027,1,1) |
The timeline cannot be moved without rewriting formulas | Model start date in the assumptions sheet; timeline formulas reference it |
Formula contains *12 to turn a monthly figure into an annual one |
Obvious and unlikely to change, so usually acceptable | Leave in place if the meaning is plain, or label it in a reference area if not |
Step 1: Define the model before you ask the AI to build it
An AI tool given a vague request will invent its own structure, and that structure is harder to check. Before prompting, write down the outputs you need, the forecast periods, the operating drivers, and how assumptions feed schedules and then the three statements. A structured model plan gives you a standard to compare the output against.
#1 Best Overall
The UK government’s Financial Model Essentials guidance, aimed at founders, CFOs, and leadership teams preparing models for investor scrutiny, recommends a bottom-up, driver-based forecast. It also recommends grouping assumptions, keeping an assumptions log, and running sensitivity analysis. Those practices are easier to request when the model plan already exists.
Step 2: Write a prompt that forces separation
Ask for an assumptions sheet, not just “a financial model.” The prompt below turns the standard guidance into instructions an AI tool can follow. It is a practical synthesis of the cited guidance, not a guarantee of compliance, so the audit in the later sections is still required.
Build the model with a clearly labelled assumptions sheet. Put every value that could change during the forecast in a documented input cell, including its unit, source, and rationale. Reference those inputs in formulas; do not embed changeable assumptions as numbers inside formulas. Keep genuinely fixed constants only when their meaning is obvious, and label any less obvious constant. Make assumptions, calculations, and outputs easy to distinguish. After building, list the checks performed and flag formula inconsistencies, embedded numbers, hidden sheets, external links, and any check that failed. I will review the workbook independently.
Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.
Two details make this prompt work better than a generic instruction. First, it asks for a source and rationale for each input, which turns a bare number into a documented assumption. Second, it asks the tool to report its own checks, so you have a list to verify rather than a general claim of accuracy.
Step 3: Decide which constants can stay in formulas
Blanket removal of every number is a mistake. ICAEW’s guidance is judgement-based: values that could change over the life of the model should be inputs, but a genuinely unchanging constant with an obvious meaning can stay where it is. Unfamiliar constants should be separated and labelled. Removing values such as 0 or 1 usually makes a formula harder to read, not easier to audit.
| Constant | Keep in formula? | Reason |
|---|---|---|
| 0 or 1 in a logical or switching formula | Yes | Meaning is plain and moving it would obscure the logic |
| Hours in a day (24) or months in a year (12) | Usually yes | Fixed by definition and immediately understood |
| A unit conversion factor, such as one that converts thousands to millions | Label it in a reference area | Its meaning may not be obvious to the next user |
| Tax rate, growth rate, churn rate, headcount, start date | No | These can change and belong in documented inputs |
Separate scaling calculations from base calculations as well. A formula that both converts units and applies a business driver is hard to inspect. Two simpler formulas are easier to audit than one long one.
Rank #2
Step 4: Audit the generated workbook
Do not accept the tool’s own statement that the model is clean. ICAEW’s guidance on reviewing AI-generated models states that asking the AI to confirm these defects is not a substitute for checking them. Work through the file in this order.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match-
Find numbers typed into formulas. Press
Ctrl+`or go to Formulas > Show Formulas to display every formula instead of its result. Scan for digits other than the obvious constants you approved in Step 3. -
Find hardcoded values in calculation areas. Select the forecast range, then go to Home > Find & Select > Go To Special > Constants. Typed numbers in cells that should contain formulas will be highlighted. Repeat with Formulas to see which cells do contain formulas, and compare the two selections.
-
Compare formulas across periods. Copy the formula from one forecast column across the row and look for any cell whose structure differs from its neighbours. Excel’s background error checking can flag some inconsistencies. Enable it under File > Options > Formulas and check that the Inconsistent formula rule is ticked. It catches only some patterns, so treat it as a helper, not a full test.
-
Trace where each input goes. Click an input cell and use Formulas > Trace Dependents. If a changeable assumption feeds no formula, or feeds only a few, the model may have a second hidden copy of it somewhere.
Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy. -
Check for hidden sheets. Right-click any sheet tab and choose Unhide. A generated model can contain sheets with calculations or old assumptions that the visible structure does not show.
Rank #3
-
Check for external links. Look under the Data tab for Edit Links. If it is available and lists files, the model depends on another workbook, and its values may not be the ones you expect.
ICAEW’s list of common review targets for AI models also includes missing sections and incomplete debt schedules. Check the outline against the plan from Step 1, not only the formulas you can see.
Step 5: Test behaviour, not just appearance
A model can look tidy and still behave incorrectly. Once the formulas pass inspection, test the logic directly.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →-
Change one input at a time. Adjust a growth rate or tax rate and confirm that every affected statement and schedule moves, and that nothing else does. If a changed input leaves a line unchanged, that line is probably hardcoded or disconnected.
-
Check the balance sheet for plugs. A balance sheet that balances only because a cash or equity line absorbs the difference is not evidence of correctness. Confirm that the balancing item is derived from real movements in cash and other accounts.
-
Check debt schedules and capacity constraints. Confirm that repayments do not exceed balances, that interest follows the stated rate, and that capacity limits in operating schedules actually bind when they should.
-
Check internal checks in every forecast period. A check cell that shows “OK” in the first year may show an error in a later one. Scroll to the last period and confirm the check still passes.
Recommended Free Tools
Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.
ICAEW’s AI review guidance specifically calls for these tests, and it notes that a model’s apparent balance should not be accepted when a plug forced it.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Where prompts end and review begins
A well-written prompt reduces the number of defects an AI tool produces. It does not remove the need to check the output. ICAEW’s guidance puts the principle directly: “The most effective way to review an AI-generated model is to treat it as a draft that must be checked.” That sentence comes from ICAEW’s article “How to identify AI errors in financial models,” published in June 2026.
Keep the human check independent. The person who writes the prompt should not be the only reviewer. A second reader using the audit steps above will catch problems that the author, who already knows what the model is supposed to do, may overlook.
If you want broader grounding in modelling practice, Danielle Stein Fairhurst’s chapter “Best-Practice Principles of Modelling” in Using Excel for Business and Financial Modelling (Wiley; chapter first published 25 March 2019) covers documenting assumptions and linking calculations. Check the current edition and availability before buying.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
The workflow works because each step narrows the gap between what the AI produced and what you can verify. Define the model, separate inputs, instruct the tool clearly, then inspect and test the result yourself.
Note on sources: this guidance is based on ICAEW’s Financial Modelling Code (2024 edition, marked 08/24), ICAEW’s June 2026 article on AI errors, UK government Financial Model Essentials guidance (undated in the reviewed copy), and CFA Institute and Financial Modeling Institute materials. The Financial Modelling Code is a general code, so its hardcoding principle depends on context, as described above.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
