A guaranteed maximum price is not just a number; it is a number with a story, and the story is the backup narrative that explains what the GMP includes, what it assumes, what it excludes, what it carries for contingency, and how it got from a rough conceptual estimate to a committed price. Owners do not approve a GMP on the number alone; they approve it on the narrative that makes the number credible and the assumptions visible. Writing that narrative is substantial work, and it is exactly the kind of structured, explanatory writing AI accelerates, with the discipline that every number and assumption in it must be true and the named delta thresholds that track the estimate's evolution must be real. This lesson shows you how to use AI to draft a GMP backup narrative fast, so the estimating judgment goes into the numbers and the assumptions while the AI handles turning them into the clear, credible story the owner approves.

What the GMP Narrative Is and Why It Matters

A GMP backup narrative is the document that accompanies a guaranteed maximum price and makes it understandable and defensible to the owner. It states the basis of the cost, the scope included, the assumptions the price rests on, the exclusions, the allowances, and the contingency, and it explains the estimate's progression through its stages, from conceptual to schematic to design-development to the final GMP. The narrative is what turns a bare number into a price the owner can evaluate and approve, because it surfaces the assumptions and the scope so the owner knows exactly what they are buying and what they are not.

This matters because a GMP without a clear narrative is a number the owner cannot trust, and a GMP with a narrative that hides or misstates the assumptions is worse, because it sets up a dispute later when an excluded item or an unstated assumption surfaces. The narrative is the instrument of transparency and trust between contractor and owner at the moment the price is committed, so it has to be both clear and accurate, clear so the owner understands it, accurate so it does not mislead. It is language-heavy, structured, and explanatory, the writing profile AI handles well, but every substantive claim in it, every number, every assumption, every exclusion, is an estimating judgment that must be true, so the AI drafts the narrative around the estimator's real numbers and real assumptions, never inventing them.

The Named Delta Thresholds and Why They Are Real

A GMP does not appear from nowhere; it evolves through estimate stages, and the named delta thresholds are the expected ranges of how much the estimate should change between stages, which give the owner a way to understand and trust the progression. The AACE classification, from a conceptual Class 5 estimate to a definitive Class 1, frames this, and in practice the thresholds are the named ranges the estimate is expected to tighten within: a conceptual-to-schematic delta of roughly plus or minus twenty percent, a schematic-to-design-development delta of roughly plus or minus ten percent, a design-development-to-GMP delta of roughly plus or minus five percent, the estimate converging as the design develops. These named thresholds are what let the owner see that the estimate is maturing in a normal, controlled way rather than swinging unpredictably.

The discipline here is that these thresholds and the deltas against them must be real, drawn from your actual estimate progression, not numbers the AI generated to fill the structure. If the AI drafts a narrative that says the schematic-to-DD delta was eight percent, that eight percent must be your actual delta, because an owner who later finds the stated progression did not match the real one loses trust in the whole GMP. So the estimator supplies the real stage estimates and the real deltas, and the AI drafts the narrative that presents them against the named thresholds, explaining where the estimate converged as expected and flagging any stage where it moved more than the threshold, which is itself useful information for the owner. The thresholds make the progression legible; the real deltas make it honest, and the AI renders the estimator's real progression into the clear narrative, never inventing the numbers that are the narrative's whole credibility.

The named delta thresholds make the GMP's progression legible: conceptual-to-schematic, schematic-to-DD, DD-to-GMP, each converging within an expected range. But the deltas against them must be your real numbers, because an owner who finds the stated progression did not match the real one loses trust in the whole GMP.

What AI Drafts and What the Estimator Owns

The division of labor is the level's standard one, applied to a high-stakes financial narrative. The AI drafts the prose: it turns the estimator's basis of cost, assumptions, exclusions, allowances, contingency, and stage deltas into a clear, well-organized, owner-readable narrative, fast, which is exactly the structured explanatory writing that consumes estimating hours and that AI does well. The estimator owns the substance: the numbers, the assumptions, the exclusions, the contingency logic, and the real stage progression, all of which are estimating judgments that must be true and that the AI receives as inputs, not generates.

The danger specific to this document is that the narrative explains money the contractor is committing to, so an error in it is a financial misstatement to the owner at the moment of the GMP commitment. If the AI invents an assumption the estimate does not actually rest on, or states an exclusion incorrectly, or misrepresents the contingency, the owner approves a GMP on a false basis, which is both a trust failure and a setup for a dispute when the truth surfaces. So the verification is that every substantive claim in the narrative is confirmed against the estimate: every number matches the estimate, every assumption is one the estimate actually makes, every exclusion is real, every delta is the real delta. The AI accelerated the writing; the estimator verified that the writing is true, because the GMP narrative going to the owner must say exactly what the estimate actually is, no more and no less, and the AI's fluency is no substitute for that accuracy.

The Assumptions Are the Document

The single most important content in a GMP narrative is the assumptions, because the assumptions are where the price is most exposed and where disputes most often originate, so they deserve particular care in both drafting and verification. A GMP rests on assumptions, about the design completeness, the site conditions, the schedule, the scope interpretation, and if those assumptions are wrong or unstated, the guaranteed price is not actually guaranteed against the conditions that turn out to differ. A well-drafted narrative makes the assumptions explicit and complete, so the owner knows the price is contingent on them and the contractor is protected if they do not hold.

AI helps draft the assumptions clearly, but the completeness and accuracy of the assumptions are entirely the estimator's, because the AI does not know what the estimate actually assumed; it only knows what the estimator tells it. The risk is twofold: the AI could state an assumption the estimate does not make, creating a false representation, or, more dangerously, the narrative could omit an assumption the estimate does rest on, leaving the contractor exposed on a condition they did not flag. So the estimator's work is to ensure every real assumption is in the narrative, stated accurately, and the AI's work is to render them clearly, and the verification specifically checks that the assumptions are both complete, all the real ones are present, and accurate, none are misstated or invented. The assumptions are where the GMP narrative most protects or exposes the contractor, which is why they get the most careful drafting and the most careful verification, with the estimator owning their completeness because only the estimator knows what the estimate truly rests on.

Drafting the Contingency and Exclusions Clearly

Two other elements of the narrative reward careful drafting because they are frequent sources of owner confusion and later dispute: the contingency and the exclusions. The contingency is the money carried for the unknown, and owners often misunderstand it, treating it as padding to negotiate away or as money they are owed back, so a clear narrative explains what the contingency covers, how it is governed, and what happens to it, in language the owner understands, which heads off the friction that a vague contingency line invites. AI helps here by drafting the contingency explanation clearly, turning the estimator's contingency logic into language that makes its purpose and governance legible to the owner.

The exclusions are equally important and equally prone to causing disputes, because an exclusion the owner did not notice becomes an argument when the excluded work is needed and the owner expected it included. A well-drafted narrative makes the exclusions prominent and specific, so the owner cannot later claim they did not know a scope was excluded, which protects the contractor. The AI drafts the exclusions clearly and prominently from the estimator's real exclusion list, and the estimator verifies the list is complete and accurate, the same completeness discipline as the assumptions, because an omitted exclusion, like an omitted assumption, leaves the contractor exposed on a scope they meant to exclude. The pattern across the contingency, the exclusions, and the assumptions is the same: these are the parts of the GMP narrative where clarity prevents disputes and completeness protects the contractor, so they get the careful AI drafting for clarity and the careful estimator verification for completeness and accuracy, because a GMP narrative clear and complete on these points is what makes the committed price hold rather than unravel into argument.

Where the Speed Actually Helps in Precon

It is worth being clear about why drafting speed matters specifically for the GMP narrative, because the benefit is not merely convenience. The GMP is often produced under a deadline, an owner needs the price to make a go decision, a target-value-design process is converging on a date, and the narrative is substantial writing that competes for the estimator's time with the estimating work itself. When the narrative drafting is slow, it either delays the GMP or, worse, gets rushed and produced thin, with assumptions and exclusions stated incompletely because there was not time to write them out fully, which is precisely the incompleteness that exposes the contractor.

AI's drafting speed addresses this by making a complete, well-written narrative affordable in the time available, so the estimator can spend their time on getting the numbers and assumptions right and let the AI produce the thorough narrative those numbers deserve, rather than choosing between a slow complete narrative and a fast thin one. This matters because the narrative's completeness is its protective value, and a thin narrative produced under deadline is where the contractor gets exposed, so the speed that lets the narrative be complete is directly protective, not just convenient. The estimator who uses AI to draft the GMP narrative gets to put a complete, clear, accurate story behind the price without the writing time forcing a trade-off against the estimating time or the deadline, which means the price the owner approves rests on a narrative thorough enough to actually protect the contractor, and that is the substantive value of the speed, not merely producing the document faster but producing the complete document that a rushed process would otherwise have made thin.

The Applied Problem: Draft a GMP Narrative With Real Deltas

Here is the exercise. Take a real or representative target-value-design project and produce a GMP basis-of-cost narrative, drafted with AI, that presents the assumptions, exclusions, allowances, contingency, and the named delta thresholds against the real stage progression, and validate it against your firm's CM-at-risk contract template. Run the workflow: you assemble the real basis of cost, the real assumptions and exclusions, the contingency logic, and the real stage estimates and their deltas; the AI drafts the narrative presenting these clearly, including the progression against the named thresholds; and you verify every substantive claim against the estimate and confirm the narrative is consistent with your contract template.

Produce two things. First, the GMP backup narrative, clear and owner-readable, with the basis of cost, the complete and accurate assumptions, the exclusions, the allowances, the contingency, and the stage progression against the named delta thresholds, every number and claim verified against the real estimate. Second, the verification note: confirmation that every assumption, exclusion, number, and delta in the narrative matches the actual estimate, with particular attention to the completeness of the assumptions, because that is where the contractor is most exposed. Confirm the narrative is consistent with the CM-at-risk contract template, since the GMP operates within that contract's terms.

The deliverable is the verified GMP narrative and the verification note, and the lasting product is a workflow that lets you produce the credible, accurate GMP story the owner approves, fast, so your estimating time goes into the numbers and the assumptions while the AI handles the substantial writing of turning them into the owner-readable narrative. This is the GMP core of the estimating chapter, and it follows the pattern: AI drafts the explanatory writing, the estimator owns and verifies the substance, and the deliverable, the thing the owner approves a multi-million-dollar price on, is both faster to produce and accurate, because the writing was AI's and the numbers and assumptions stayed the estimator's. The estimator who masters this produces GMP narratives that win owner approval on a true basis, fast, which protects both the speed of the precon process and the integrity of the price the contractor is committing to.

Key Takeaways

• A GMP is a number with a story: the backup narrative states the basis of cost, scope, assumptions, exclusions, allowances, and contingency, and explains the estimate's progression. Owners approve the GMP on the narrative that makes the number credible, not the number alone.
• The narrative is the instrument of transparency and trust at the moment the price is committed, so it must be both clear (so the owner understands it) and accurate (so it does not mislead and set up a dispute). It is language-heavy writing AI handles well, around substance that must be true.
• The named delta thresholds make the progression legible: conceptual-to-schematic (~±20%), schematic-to-DD (~±10%), DD-to-GMP (~±5%), per the AACE Class 5 to Class 1 progression. The deltas against them must be your real numbers, because an owner who finds the stated progression false loses trust in the whole GMP.
• AI drafts the prose; the estimator owns the substance: every number, assumption, exclusion, contingency, and real delta is an estimating judgment the AI receives, not generates. An error is a financial misstatement to the owner at the moment of commitment.
• The assumptions are the document: they are where the price is most exposed and disputes originate. The estimator owns their completeness (all the real ones present) and accuracy (none misstated or invented), because only the estimator knows what the estimate truly rests on, and an omitted assumption leaves the contractor exposed.
• The verification confirms every substantive claim against the estimate, with particular attention to the completeness of the assumptions, and confirms the narrative is consistent with the CM-at-risk contract template the GMP operates within.
• The artifact: an AI-drafted GMP basis-of-cost narrative for a target-value-design project presenting assumptions, exclusions, allowances, contingency, and the real stage progression against the named thresholds, with a verification note confirming every claim matches the estimate and the contract template.