Bid like a firm ten times your size.

AI-assisted estimating for federal IT services. Send us the solicitation you're bidding and get back a defensible hours-and-price basis — built line by line, traceable to the page it came from.

One pursuit, no commitment. You'll know inside a week whether it changes how you bid.

The estimate is the bid. It's also the part nobody has time to do properly.

It takes days you don't have

A PWS lands. Someone senior reads all of it and builds a spreadsheet. That person is usually the one who should be selling.

Two estimators, two answers

The same solicitation priced by two people differs widely, and neither can fully explain the gap. The number reflects who was free that week.

It doesn't survive scrutiny

Asked in a debrief where a number came from, the honest answer is often "experience". That is not a basis of estimate.

The cost isn't the losing bid. It's the one you win at a price that can't be delivered.

How it works

Four steps, and you can audit every one.

  1. 01

    Shred

    The solicitation is read end to end and split into individual requirements — every shall, every task area, every deliverable, in document order.

  2. 02

    Map

    Each requirement is matched to a standard activity in the catalog and given a risk posture. Two decisions per line; nothing else is left to judgment.

  3. 03

    Compute

    Hours fall out of the catalog's published figures and the cadence the document states. Arithmetic, not opinion.

  4. 04

    Price

    Labor cost is burdened through your own indirect rates — fringe, overhead, G&A, fee — into a bid-ready price, with escalation across option years.

The output is a basis of estimate, not just a total. Open any line and you see the requirement it came from, the activity it mapped to, the hours and why, the role performing it, and the arithmetic that produced the number.

Where the AI is, and where it isn't

AI reads the solicitation. It does not invent the hours.

What it does

Reads hundreds of pages, splits them into individual requirements, and matches each one to a standard activity in the catalog. The work a person does slowly, at 2am, a week before the deadline.

What it is not allowed to do

Invent an hour figure. Hours come from the catalog's published bases and the cadence the document states. Where the catalog has no activity for something, the model cannot approximate — the gap is recorded and filled deliberately, with a source.

Why the distinction matters

An estimate you cannot explain is worse than no estimate. Every number traces to a published figure and a cited framework, so it survives the only question that counts: where did that come from?

The reading is automated. The arithmetic is deterministic. Neither one is a guess.

The catalog

The part that took years, and the part you can't build on a deadline.

4,137distinct activities, each with an hour basis at three risk postures
1,117capability groups across build, transition and operations
75workstreams — infrastructure, security, data, applications, service management
566labor categories with salary benchmarks and sources

Every activity cites the framework it comes from — ITIL, NIST, PMI, DAMA, BICSI, TOGAF, RSMeans and others. When a reviewer asks why an activity takes the hours it takes, the answer is a citation, not a conversation.

It grows. When a solicitation contains work the catalog has never seen, that gap is recorded and the catalog extended — deliberately, once, with a source. It does not get approximated into the nearest existing line.

What you receive

Deliverables, not a dashboard.

The labor basis

Every requirement in document order, with activity, hours, risk posture, role, level, location and the narrative justifying each.

The price build-up

Direct labor through fringe, overhead, G&A and fee to a bid-ready price — your rate structure, adjusted per pursuit.

The basis of estimate

A written BOE you can hand to a contracting officer, a teaming partner or your own delivery lead without translation.

Materials and ODCs

Bill of materials and other direct costs carried alongside labor, with the same traceability.

Assumptions and gaps

What the solicitation didn't say, what we assumed, and what to clarify before the question deadline.

A second opinion

Every estimate is independently re-derived and challenged before it reaches you — not spot-checked.

Ways to work together

Start where the risk is lowest.

Bid support — per pursuit

Send one solicitation, get the full basis back. No commitment beyond the single pursuit, and the most honest way to find out whether this is for you.

Retained capture support

A standing arrangement across your pipeline, with your rate structure and award history loaded in, so estimates track how you actually deliver.

Calibration engagement

We take contracts you've won and delivered and tune the model against your real performance. The result is a basis calibrated to your firm, not the market.

Design partners: a small number of firms get the service at cost in exchange for candour — use it on live bids and tell us where it breaks. That feedback is worth more to us right now than the fee.

Where we are

Honest about what's proven and what isn't.

The catalog and the method are built and in use, and the engine has been run against a corpus of real federal solicitations. The structure it produces — the shred, the mapping, the traceability — holds up.

Magnitude is calibrated per client, not claimed universally. A generic model cannot know how your firm staffs a help desk. That is what a calibration engagement fixes, and it is why we ask for your delivered history rather than quoting an accuracy figure at you.

Anyone who quotes you an accuracy percentage for an estimating tool, without first seeing how your firm delivers, is selling you a number they made up.

Send us the next one you're bidding.

One solicitation, one basis of estimate, no commitment.

Email the solicitation

hello@andazza.com