Budget Reconciliation
Close the gap between the programme and the budget, and price what goes
Once installed, Claude loads it on its own when your conversation matches. You can also call it directly with /budget-reconciliation.
Install just this one
npx archtmpl@latest --skill budget-reconciliation --globalFirst time? The whole install, step by step
- Open Claude Code — the terminal version or the desktop app, either one.
- In a terminal, paste the line above and press Enter. In the app, paste it into the chat and ask Claude to run it.
- Restart Claude Code. That's the whole install.
Set up plugins for me: run `claude plugin marketplace add https://archaiflow.com/plugins/marketplace.json` and then `claude plugin install budget-reconciliation@archaiflow`Paste into the Code tab (not Chat or Cowork) and approve when Claude asks. The third-party marketplace it mentions is this site. Windows may ask to install Git once.
What this skill does
Budget Reconciliation
The estimate comes back high and the meeting that follows decides the project. Handled as a list of things to delete, it produces a worse building at a price nobody predicted and a client who has lost confidence. Handled as four different kinds of move, each priced in what it costs the design, it is a decision the client can actually make.
Workflow
Step 1. Establish what each number is, before comparing them
The step that is skipped, and it often closes a third of the gap on its own.
| Budget | Estimate | |
|---|---|---|
| What it covers | ||
| What it excludes | ||
| When it was set / made | ||
| Contingency in it | ||
| Escalation in it | ||
| Fees, permits, FF&E, IT, moving | ||
| Land, financing |
Then ask the questions the table raises:
- Does the client's budget include soft costs the estimate excludes, or the reverse? A client saying "my budget is X" usually means everything, and an estimate usually means construction only.
- Are the two on the same date? A budget set eighteen months ago against a current estimate is partly an escalation conversation, not a design one.
- Is contingency counted once, twice, or not at all?
- Are they on the same area basis? Point at
area-measurement.
Report the gap after this reconciliation, not before. Say what the original gap was, what this step accounted for, and what remains. The remainder is the real problem, and it is usually smaller than the number in the room.
Step 2. Sort the moves into four kinds, because they behave differently
| Kind | What it means | What it costs |
|---|---|---|
| Less area | Build less of it | Programme. Usually the largest single lever and the least popular |
| Less quality | Same building, cheaper | Durability, maintenance, and how it reads. Often reappears as an operating cost |
| Less scope | Defer or remove part | Phasing risk, and it is rarely as cheap later |
| More money | Raise the budget | The client's problem, and sometimes the right answer |
Every proposal goes in exactly one. The four are worth separating because a client who will not lose area may readily raise the budget, and a list that mixes them makes that invisible.
Name "more money" as an option every run, even when it is unwelcome. A reconciliation that only offers cuts has decided something the client should decide.
Step 3. Price each move in both directions
| # | Move | Kind | Saves | What the design gives up | Reversible later? | Source of the figure |
|---|
The fifth and sixth columns are the ones that make this different from a cut list:
- What the design gives up in a sentence, concretely. Not "reduced quality". What a person in the building would notice.
- Reversible later matters more than clients expect. Leaving out a fit-out is reversible; a lower floor-to-floor, a smaller structural grid, or a thinner wall is not, and neither is anything buried.
Anything with no figure in Saves is reported as unpriced. Never estimate to
fill the column.
Step 4. Find the moves that are cheap now and impossible later
Sort the list by when the decision has to be made rather than by what it saves:
- Moves that must happen now or never: anything structural, anything about levels, anything below grade, anything in the envelope's build-up.
- Moves that can wait: finishes, fit-out, landscape, FF&E, some equipment.
- Moves that get more expensive the longer they wait: anything requiring redesign of a set already coordinated.
Say plainly which decisions the meeting actually has to reach today. Most cut lists treat every item as equally urgent, and the ones that matter get lost.
Step 5. Say what the reconciliation costs to produce
Two things, honestly:
- Redesign. Which moves require redrawing, how much, and whether that is
additional services under the agreement. Point at
draft-additional-services. A practice that absorbs three rounds of redesign silently has paid for the client's budget problem. - Programme. What each move does to the date, including a re-estimate.
Step 6. Report
- The reconciliation from Step 1, first, with the gap before and after.
- The remaining gap, and the four-kind table with what each kind could contribute.
- The priced moves, with
what the design gives upandreversible. - What has to be decided today versus what can wait.
- Everything unpriced, listed.
- The redesign and programme cost of proceeding.
Close by saying nothing was estimated here, and that the moves are the client's to choose.
Step 7. Save
Ask where. This becomes the record of what was decided and what it cost, and it is read back when somebody asks in a year why the building does not have the thing that was in the early drawings.
Rules
- Never supply a cost, a rate, or a benchmark.
- Reconcile what the two numbers are before reporting a gap.
- Sort every move into exactly one of the four kinds.
- Offer "more money" as an option, every run.
- Every move states what the design gives up, concretely.
- Mark every move reversible or not.
- An unpriced move stays unpriced. Never estimate to fill a column.
- Say which decisions have to be made today.
Anti-patterns
- Producing a list of cuts.
- Comparing a client's all-in budget against a construction-only estimate.
- Ignoring escalation between the date of the budget and the date of the estimate.
- Mixing the four kinds in one list, which hides that the client would rather raise the budget.
- Cutting the envelope, which is buried, before cutting the fit-out, which is not.
- Absorbing the redesign silently.
- Estimating a saving to make the list add up.
- Treating every item as equally urgent.
Resources
None. This skill is one file. Output is written directly at the path you choose.
What it does not check
What this does. Establishes what the two numbers are before comparing them, sorts the ways of closing the gap into four kinds, and prices each move in both directions: what it saves, and what the design gives up.
What this does not do.
- It carries no costs, rates or benchmarks. Every figure comes from the
estimator, the contractor, or a source you name.
/lookup-costquotes a benchmark with its source. - It does not estimate. Where a saving has no number against it, it stays without one and is reported as unpriced rather than guessed.
- It is not construction-stage value engineering.
value-engineering-advisorsorts a contractor's list into accept, alternate and block. This is the earlier conversation, where the moves available are larger and cheaper. - It does not decide. The client decides, and the point of the exercise is giving them something decidable.
- It does not manage the fee consequence. Redesign costs the practice time;
draft-additional-servicesandutilization-trackerhandle that side. - It does not replace the Architect of Record.
What you need before starting. The estimate, with its breakdown and its exclusions. The budget, and what it was meant to cover. The programme. What the client has said is untouchable.
Files it puts on your disk
.claude/skills/budget-reconciliation/1 file · 8.3 KBSKILL.md8.3 KB