Every estimate that goes out with a bid is also a set of design decisions, whether or not anyone wrote them down. By the time the delivery team arrives, the number is mentioned in the contract, and the design it quietly assumed comes with it. That’s why I think the estimate is the first architecture decision on a project, and usually the least reviewed one.
Every estimate line is a design in disguise
An estimate is built from a solution sketch, and every line carries a design choice inside it:
“Integration with the finance system: 15 days” is true for one kind of integration, probably a nightly file in one direction.
“Data migration: 20 days” assumes the source data is reasonably clean and loads once.
“Reporting: 10 days” assumes standard reports with a few filters.
None of these numbers reach the team actually delivering the solution with their design assumptions attached, unless presales writes those assumptions down next to each line. Otherwise they live in the head of whoever made the sketch, usually under deadline and often across three bids at once.
This is what I call estimate-led architecture: the presales number, set by people who won’t deliver it, quietly deciding the design.
Once the contract is signed, the number becomes the constraint and the sketch disappears. The delivery architect arrives, does the analysis that presales had no time for, and finds that the honest design costs more. The real choice is between three options: raise a variation in the first month (spending relationship credit before there’s any), absorb the cost in margin, or design to the number.
The third option wins most often, because it’s the only one that needs no difficult conversation.
The estimation challenges are real
Presales architects work from a request for proposal, a handful of clarification answers and a price the client can approve. They’re measured on whether the bid wins, and by the time delivery finds out what the number decided, they’re three bids further on. This is a description of the role, and I don’t mean it as a criticism: the information the honest design needs is rarely available at bid time. The budget is a legitimate requirement, and designing within a constraint is what architects have to do.
What’s important, though, is to make these trade-offs explicit. A trade-off made openly (”we’ll integrate nightly rather than in real time, because that’s what this budget allows”) is architecture. The same trade-off made by a spreadsheet, without anyone telling the client, is a surprise waiting for the project team.
AI-assisted estimating makes this even harder to see. A drafting agent can produce a detailed, line-by-line estimate from past proposals in an hour, and the detail reads as rigour. The lines inherit the designs of the projects they were copied from, and their assumptions stay as invisible as before. The estimate gains precision without gaining honesty.
Read the estimate as an architecture document
What solution architects should do at project mobilisation is review every estimate line and identify the design assumptions behind it, both explicit and implicit. Read the estimate as if it were the first design document, because it is. List the design decisions it made, explicitly or by omission, and take that list to the client in the first month, as part of requirements discovery, with the cost of changing each one. Some the client accepts as the right trade-off; others become a change, raised early, when it’s cheapest and least personal: “The bid assumed a nightly file. Your analysis shows you need it in real time. Here’s what that costs.”
TL;DR
Write the design assumption next to every estimate line, before the number is submitted.
Get one hour of the person who will deliver the work before the price goes out.
In week one, list the design decisions the estimate made and walk the client through them in the first month.

