> ## Documentation Index
> Fetch the complete documentation index at: https://control-dev.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.
# Adjust a budget with an AI client
> Watch an AI client reforecast a budget as a new version: rebase on actuals, layer a hiring plan and a weighted pipeline, and find the costs the adjustments missed.
This example shows a mid-year reforecast built up in reviewable layers: the client rebases the remaining months on
actuals as a new budget version, folds in a hiring plan and a probability-weighted sales pipeline from attached files,
and then answers the question every reforecast should end with — what did we miss?
The recording has narration, and the figures, entities, and account names in it are illustrative. Nothing in the
video is a target or a benchmark.
## What happens in the recording
1. The user asks a connected AI client for a new version of the annual P\&L budget: keep the elapsed months as
budgeted, and rebase the remaining months on accounts where first-half actuals show the baseline has clearly
shifted.
2. The client compares plan with actuals account by account, creates the revision as a copy of the current budget, and
writes the rebased months — reporting the profit impact of the rebase with a chart and a reason for every reset.
3. The user attaches a hiring-plan workbook. The client layers the new hires into the revision and re-forecasts the
accounts that scale with headcount, stating its per-account assumptions.
4. The user attaches a CRM pipeline export. The client folds it into the revenue forecast probability-weighted and
recognized from each deal's close month, and charts organic run-rate against the weighted pipeline.
5. The user asks what other costs the adjustments imply. The client finds and sizes the gaps — a stale tax line,
delivery costs on the new revenue, commissions, and recruiting — applies the mechanical ones on confirmation, and
holds the ones that depend on decisions the user has not made yet.
## Try it in your workspace
1. [Connect an AI client with MCP](/product-docs/ai/mcp-server) and confirm the session has write access.
2. Ask for the reforecast as a **new version** of the current budget, and state which months stay frozen as originally
planned. The original budget remains untouched for comparison.
3. Rebase first, then add one layer at a time — a hiring plan, a pipeline, a price change — confirming each write
separately so every movement in the plan traces to one named adjustment.
4. Ask for the profit impact after each layer, and for the reasoning behind every changed account.
5. Finish by asking what the adjustments missed. Knock-on costs — taxes, commissions, delivery costs, recruiting —
sit in accounts the source files never mention.
## Adjust in layers, not in one pass
A reforecast that arrives as one bulk write cannot be reviewed: if the total moves, there is no way to say which
assumption moved it. Each layer in the recording is a separate confirmed change set with its own stated assumptions
and its own before-and-after comparison, so a wrong assumption can be revised without unwinding the rest.
The judgment calls belong to you, not the client. Which accounts have genuinely shifted baseline versus a one-off
half, what probability weighting and recognition timing the pipeline deserves, and which knock-on costs to book now
versus hold — the client should propose each with its reasoning and wait, exactly as it does in the recording when it
holds commission and recruiting costs pending a compensation plan.
Example prompts: "Create the reforecast as a new version and keep January through July exactly as budgeted." "List
every account you rebased, the old and new monthly level, and why." "Weight the pipeline by stage probability and
recognize from close month — show me the weighting before you write." "What costs should this plan now carry that it
does not?"
## Troubleshooting
### The reforecast changed months it should not have
State the frozen range explicitly and ask the client to confirm it read the range back before writing. If elapsed
months were rebased anyway, the budget is a version — compare it with the original in Control, and ask the client to
restate the affected months from the original values.
### The pipeline landed on the wrong months or at face value
Weighting and timing are inputs, not inferences. Give the probability basis and the recognition rule, and ask the
client to show the weighted, month-by-month revenue table before anything is written. A pipeline folded in unweighted
or recognized on booking date instead of close month overstates the near months.
### Headcount-linked costs did not move with the hiring plan
Ask which accounts the client treated as scaling with headcount and at what rate per head. Costs like software seats,
fringe, and equipment follow hires with different lags; an account it left flat needs an explicit rule, not silence.
### The plan total moved but you cannot say why
That is the failure the layered approach prevents. Re-run the adjustment as separate confirmed layers, and require a
per-layer impact summary. If the layers are already written, ask the client to reconcile the revision against the
original account by account and attribute each difference to one of the adjustments.
## Next steps
Seed the initial plan from actuals before adjusting it.
Bring in a plan that already exists in a workbook.
Compare any version with actuals on a reconciled basis.
Review versions, refine allocations, and lock the plan in Control.