AI Agents for
IFS FP&A
A tool claiming to forecast project outcome is claiming to know physical progress, and physical progress is not in the finance system.
- Project FP&A turns on cost to complete, and that is a judgement owned by the project manager. An agent should assemble every input to it and never produce it.
- Estimate drift carries more information than the estimate. An unchanged figure across four periods usually means a copy forward rather than a stable project.
- Committed cost above the remaining estimate is arithmetic, not judgement. It means the reported margin is wrong today, and it is the urgent list.
- Report portfolio coverage alongside any rolled up figure. A portfolio margin covering 60 percent of projects invites a conclusion the data does not support.
- A profitable project can be cash negative for most of its life. Retention, certification lag and billing shape move cash without touching profit.
AI agents for IFS FP&A assemble the inputs to cost to complete, cost incurred by activity, committed cost on open orders, certified valuations and the run rate implied by actuals, then report where the project manager's estimate and those actuals disagree. ChatFin does not produce the estimate. Cost to complete depends on physical progress, and physical progress is not in the finance system.
What an agent can do instead is find the projects whose own numbers already contradict each other. Estimates that have not moved in four periods, committed cost that already exceeds the remaining estimate, and estimates that moved materially with no scope change recorded. None of that predicts an outcome. It produces a short list where the reported margin is wrong now rather than at some future review.
The number that matters is cost to complete, and it is a judgement
Project FP&A turns on one figure. What will it cost to finish. Everything else, including the margin you report, follows from it. An agent can assemble every input to that estimate and it should not produce the estimate itself.
| Input | Agent | Why |
|---|---|---|
| Cost incurred to date by activity | Strong | Retrieval, evidence in the ledger |
| Committed cost on open orders | Strong | Orders and receipts, calculable exactly |
| Certified subcontractor valuations | Strong | Documents it can read |
| Run rate by activity from actuals | Strong | Arithmetic over a real history |
| Physical progress against plan | Weak | Sits with the site, not in finance |
| Cost to complete | No | A judgement owned by the project manager |
The distinction is worth being blunt about. A tool claiming to forecast project outcome is claiming to know physical progress, and physical progress is not in the finance system.
ChatFin assembles the inputs and reports where the manager's estimate disagrees with the run rate implied by actuals. It does not overwrite the estimate.
Estimate drift is the signal, not the estimate
A single cost to complete figure tells you little. How it has moved, month by month, tells you most of what you need.
A worked portfolio review, 34 projects
The agent compares each project's reported cost to complete against the run rate implied by its own actuals, and ranks the gaps.
| Pattern | Projects | What it means |
|---|---|---|
| Estimate unchanged for four periods or more | 9 | Probably copied forward. Ask before believing the margin |
| Run rate implies a materially higher final cost | 6 | The estimate and the actuals disagree. One of them is wrong |
| Committed cost already above the remaining estimate | 4 | Arithmetically impossible. This is the urgent list |
| Estimate moved materially with no scope change recorded | 3 | Either a correction or a scope change nobody documented |
| Estimate tracking the run rate closely | 12 | Healthy. No action |
The four in the third row are the point of the exercise. Committed cost exceeding the remaining estimate is not a judgement call, it is arithmetic, and it means the reported margin is wrong today rather than at some future review.
Portfolio reporting that survives different project structures
The practical obstacle to portfolio FP&A on IFS is that business units structure projects differently, and a rolled up number quietly compares unlike things.
Report coverage before reporting the total
How many projects use activities, how many do not, and what share of value each represents. A portfolio margin over 60 percent coverage is a partial answer.
Map to a common reporting layer, versioned
Not by renaming things in the source. The mapping is the artefact and it needs an effective date.
Report structural changes before movements
A project restructured mid life shows a discontinuity that looks like performance.
Keep the reconciliation to the general ledger as a control
Project reporting that does not tie to the ledger will eventually be argued with, and the argument always arrives at the worst moment.
ChatFin reports portfolio coverage alongside every rolled up figure. A number presented without its coverage invites a conclusion the data does not support.
Cash on a project is not the same shape as profit
A profitable project can be cash negative for most of its life, and the two need reporting separately.
| Driver | Effect on profit | Effect on cash |
|---|---|---|
| Retention withheld | None | Delays cash until release, often past completion |
| Certification lag | None | Delays receipt by weeks per valuation |
| Payment terms out to the subcontractor | None | Improves cash, sometimes materially |
| Front loaded milestone billing | None | Improves cash early, reverses later |
| Cost incurred ahead of billing | Recognised as work in progress | Cash out with none in |
An agent can build the cash view from open payables, receivables, certification status and retention release dates. That is retrieval, and it is the part that never gets done because it spans four sources.
ChatFin builds the project cash view from open items and contractual dates, and prices receipts on observed payment behaviour rather than stated terms.
Commentary a project manager will not argue with
Drafted project commentary fails in a specific way. It describes the finance position using words the site does not recognise, and the review meeting is then spent translating rather than deciding.
Name the activity, not the account
A project manager thinks in activities and work packages. Reporting movement by nominal account guarantees a translation step.
Separate what finance did from what the project did
An accrual release, a reclass or a coding correction is not a project event, and presenting it as one is how commentary loses credibility on the first page.
Quote the documents
The valuation number, the order number, the variation reference. A manager can check those in seconds and will trust the rest of the page if they hold.
Say what changed since last period, not what the position is
The position is already in the report. The movement is the thing nobody has time to work out.
ChatFin drafts against the wording your reviewers approved in prior periods and reports by activity, with the source documents attached to each figure.
Where this underperforms on IFS
Three, and none of them is a modelling problem.
ChatFin ties out to a project report your team already trusts before it does anything else. A rebuilt figure that does not tie is a finding, and week one is the right time to find it.
Deeper in IFS FP&A
Three use cases sit under this hub, in build and shown rather than linked.
Worked on this page
Natural language querying of finance data · Financial forecasting · Financial modeling
Covered as sections here
Cost to complete · Estimate drift · Portfolio reporting · Project cash view · Variance analysis · Management commentary
Questions IFS finance teams ask first
Can an AI agent forecast a project's final cost?
No, and a vendor claiming it can is claiming to know physical progress. Cost to complete depends on how much work is genuinely done, which sits with the site rather than in the finance system. ChatFin assembles every input that is knowable, cost incurred by activity, committed cost on open orders, certified valuations and the run rate from actuals, and reports where the manager's estimate disagrees with them. The estimate stays with the project manager.
What is the most useful thing to report on a project portfolio?
Estimate drift and arithmetic contradictions. In a worked review of 34 projects, nine had an estimate unchanged for four periods or more, which usually means it was copied forward, and four had committed cost already above the remaining estimate, which is arithmetically impossible and means the reported margin is wrong today. That second group is a short, urgent and completely mechanical list.
Why does portfolio reporting on IFS go wrong?
Because business units structure projects differently and a rolled up number quietly compares unlike things. One unit uses activities and another does not, and a project restructured mid life shows a discontinuity that reads as performance. ChatFin reports coverage before it reports a total, maps to a common reporting layer with effective dates rather than renaming anything in the source, and reports structural changes before movements.
It finds the projects whose numbers already disagree.
ChatFin assembles cost incurred, committed cost, certified valuations and the run rate implied by actuals for every project, and reports where the reported estimate and the arithmetic disagree.
It builds the project cash view from open items, certification status and retention release dates, and reports portfolio coverage with every rolled up figure.
Bring one portfolio and its cost to complete estimates. The review takes an hour.
Book a demoAll images are served from the Pexels content network. No third party product screenshots, vendor logos or licensed media are used on this page.
Bring your cost to complete estimates.
We will compare each one to the run rate implied by its own actuals through ChatFin, and give you the list where committed cost already exceeds the remaining estimate.