How to Simplify Complex RFP Budgeting with Agentic AI

GAO's bid protest reports to Congress keep listing the same few reasons protests are sustained. In fiscal year 2025, unreasonable cost or price evaluation was again one of the top three, alongside unreasonable technical evaluation and unreasonable rejection of a proposal (GAO Bid Protest Annual Report, FY2025).

That tells you something about Volume II. A strong technical volume does not carry a proposal whose numbers are hard to follow, inconsistent, or unsupported. Cost evaluators read the price volume with the same care your team spent on the technical story.

The hard part is rarely the math. It is the volume of connected work: an agency template with its own tabs and formulas, labor rates that need support, travel priced to current per diem, indirect rates applied the right way, and an rfp budget narrative that explains every figure and still matches the spreadsheet after the last round of changes.

This guide covers what evaluators look for in a cost volume, where teams usually slip, and how an agentic AI budget assistant can take on the repetitive parts while your pricing lead keeps control of every number.

---

The Anatomy of a Federal Cost Volume (and Where Contractors Slip)

Under FAR Part 15, contracting officers must confirm that proposed prices are fair and reasonable, and on cost-reimbursement work they also perform a cost realism analysis (FAR 15.404-1). What you submit has to support both.

Most cost volumes contain two connected pieces:

The two must agree. When the spreadsheet says one thing and the narrative says another, evaluators notice, and competitors reading a debriefing may too.

Common Reasons Cost Volumes Lose Points

• Technical and cost volumes disagree. Volume I proposes a senior architect for a full year; Volume II prices a mid-level engineer for half that. Evaluators read this as a misunderstanding of the work.
• Unallowable or unexplained costs. Items that are unallowable under FAR 31.205 (entertainment, for example) or costs with no stated basis weaken the whole volume.
• Indirect rates with no support. A G&A rate stated without reference to your accounting data, a provisional billing rate, or an approved agreement invites questions.
• Numbers that drift between files. A late spreadsheet change that never reaches the narrative leaves two different totals in the same submission.

---

What an RFP Budget Narrative Must Cover

An rfp budget narrative (also called a cost narrative, budget justification, or basis of estimate) answers the questions every evaluator asks: how did you arrive at these numbers, are the rates realistic, and are the costs allowable, allocable, and reasonable?

A complete narrative usually walks through:

• Direct labor. Each labor category, the qualifications the solicitation requires, the hours proposed, and the basis for each rate (payroll data, salary surveys, or published benchmarks such as GSA CALC+).
• Indirect rates. What your fringe, overhead, and G&A pools include and how they are applied, including any handling on subcontracts.
• Escalation. The annual rate you apply to later years and its basis, for example the Employment Cost Index published by the Bureau of Labor Statistics.
• Other direct costs and travel. Equipment, licenses and services, and travel priced to the Federal Travel Regulation and current GSA per diem rates.

The Cost Realism Trap

On cost-reimbursement contracts, evaluators estimate a probable cost of performance. If your hours or rates look too low for the work described in your technical volume, they can adjust your evaluated cost upward, which erases the advantage of a low bid and signals that you may have misread the scope.

A clear, well-supported narrative does not take that judgment away from evaluators. It gives them fewer reasons to make an adjustment, because each rate and quantity comes with its basis.

---

The Manual Bottleneck: Excel and Word Hand-Offs

Most teams still build Volume II through a chain of hand-offs:

1. The pricing lead builds or adapts the pricing spreadsheet.
2. Versions move by email while hours and rates change.
3. The proposal manager copies figures into the narrative by hand.
4. Reviewers mark up both files.
5. A late change lands in the spreadsheet and has to be carried into the narrative again.

Every step is a chance for the two files to drift apart. Three problems show up again and again.

1. Late Changes That Never Reach the Narrative

In the last days before submission, a revised subcontractor quote or a new fringe rate changes numbers across several tabs. Someone then has to find every place those numbers appear in a long narrative. Missing one leaves a contradiction in the package.

2. Senior People Doing Data Entry

Pricing leads and capture managers end up checking whether a table in the narrative matches a cell in the model, instead of working on price-to-win and trade-offs.

3. Rate Research by Hand

Supporting labor rates means searching benchmark data such as GSA CALC+, and pricing travel means looking up per diem for every destination and year. It is repetitive work where small copy errors are easy to make.

---

How Agentic AI Changes Budget Work

A chat assistant writes text. An agentic assistant works through a task in steps and uses tools along the way: it reads your documents, edits a real spreadsheet, looks up rates, and asks you when something is missing.

That is how Budgeting in Craxy AI works. An AI budget assistant builds grant budgets and contract cost volumes with you, in a chat next to a live spreadsheet.

What it does:

• Works in your agency's template. You upload the solicitation, the agency's Excel budget template, and supporting files. The assistant fills a working copy and leaves your original untouched.
• Keeps formulas live and visible. Totals are formulas you can inspect, not pasted values.
• Shows where every number came from. Each typed number carries a source: a line in your documents, a rate lookup, or an answer you gave.
• Looks up current rates. It pulls GSA per diem rates for travel and GSA CALC+ labor-rate benchmarks directly, and uses web search for other published figures such as foreign per diem.
• Asks before it guesses. When salaries, fringe or indirect rates, or quantities are missing, it asks you. It only uses company names and figures it can ground in your files or answers.
• Drafts the budget narrative. It writes the narrative as an editable document alongside the spreadsheet, from the same figures.

---

Step by Step: Building a Budget and RFP Budget Narrative with Craxy AI

Step 1: Start the Budget

Create a budget and upload the solicitation, the agency template, and any supporting documents such as quotes, salary data, or your rate letter. If you are already writing the proposal in Craxy AI, you can start the budget from the proposal so it works from the same solicitation.

Step 2: Let the Assistant Map the Requirements

The assistant reads the solicitation and template and lays out a plan: the budget type, periods, cost categories, and what inputs it needs. It lists what it found and what is missing, then asks you for the missing pieces in one batch of questions.

Step 3: Fill the Template

The assistant writes labor, fringe, indirect costs, travel, other direct costs, and totals into the working copy using formulas. Your rates stay visible as inputs, so a change to one rate flows through the formulas.

Step 4: Benchmark Labor and Price Travel

For labor categories, it can pull GSA CALC+ benchmarks and show where your proposed rate falls. For travel, it pulls the current GSA per diem for each destination.

The table below is an illustrative example of how a benchmark comparison reads (made-up round numbers, not real CALC+ data):

A rate near the middle of the range is easy to defend. A rate far below it is the kind of number cost realism reviews question.

Step 5: Draft the RFP Budget Narrative

With the numbers in place, the assistant drafts the narrative section by section: the basis for each rate, how indirect rates are applied, escalation, travel, and other direct costs. You can edit it directly.

If you change numbers later, ask the assistant to update the narrative. It works from the spreadsheet, so the narrative is rebuilt from the current figures rather than retyped by hand.

Step 6: Finish and Export

Before the assistant marks the budget finished, it checks that the templates are complete, formulas calculate without errors, every number is traceable, and totals match across files. You then review, download the workbook as Excel and the narrative as Word, and attach both to your proposal.

---

Manual Process vs. an AI Budget Assistant

---

Why Documentation Matters After Submission

A well-documented cost volume pays off beyond evaluation.

During a Bid Protest

When a protest challenges a cost or price evaluation, GAO reviews the agency's record. It generally defers to the agency's judgment when the evaluation was reasonable, consistent with the solicitation, and documented. A cost volume where every rate has a stated basis makes that record easier to build.

During an Audit

When certified cost or pricing data is required (FAR 15.403-4), auditors look for estimates supported by historical data, quotes, or published indices. Keeping the source of each number with the budget, rather than in someone's inbox, makes those questions faster to answer.

---

Frequently Asked Questions

What is an RFP budget narrative?

An RFP budget narrative is the written part of a cost or price volume. It explains how each number in the pricing spreadsheet was derived: labor categories and rates, indirect rates, escalation, travel, and other direct costs.

How is agentic AI different from a chat assistant for cost proposals?

A chat assistant only produces text. An agentic assistant works in steps and uses tools: it reads your solicitation and templates, edits a real spreadsheet with formulas, looks up rates such as GSA per diem and CALC+ benchmarks, and asks you for inputs it cannot find.

Does the assistant use my own fringe, overhead, and G&A rates?

Yes. You provide your indirect rates, from your accounting data or a rate agreement, and the assistant applies them in visible formulas and explains them in the narrative. It asks for them rather than assuming.

Is my pricing data used to train AI models?

No. Your data is never used to train AI models. It is encrypted with AES-256 at rest and in transit, kept isolated per company, and hosted on SOC 2 Type II compliant US infrastructure.

Does using AI guarantee a compliant cost volume?

No tool can guarantee that. An AI budget assistant reduces copy errors and missing sources, but your team still confirms rates, allowability, and the final price before submission.

---

Bring Your Cost Volume Up to the Standard of Your Technical Volume

Evaluators score Volume II as seriously as Volume I. Teams that treat the budget narrative as a last-minute formality give away points they could have kept.

If you want to see how the approach fits your process, our guide to the Shipley proposal process shows where pricing fits in the overall schedule, and our compliance matrix template helps you track Section L cost instructions.

---

Build Your Next Budget with Craxy AI