Pitch Box: Grounded AI for Agency RFP Responses Built on Your Own Evidence

Your best new-business lead gives notice on a Tuesday. The RFP that lands the following week needs a case study only she could pull up quickly, the one with the exact attendance figure from an activation three years back. Nobody else on the team knows which folder it's in. This is not a hypothetical for a 50 to 150 person experiential agency running a full pursuit calendar. It's the moment institutional pitch knowledge, built over years, becomes untraceable within days. The problem was never the writing. It was where the proof lived, and who could find it.

The Post-Mortem Lie: Why the Deck Review Never Finds the Real Failure

After a loss, the instinct is to schedule a deck review: sharper narrative, cleaner layout, a stronger closer on the final slide. In practice, the structural failure rarely lives in the presentation layer. It lives three weeks earlier, when the team couldn't surface a verified attendance figure from an analogous activation, reached for a number that felt close enough, and submitted it anyway. An evaluator asked about methodology in the Q&A. Nobody in the room had a real answer.

Win rate is an evidence problem before it's a writing problem. A 100-person experiential agency running 40 qualified pursuits a year, with senior business-development labor billed internally around $175 an hour, spends between $210,000 and $420,000 annually on pursuit assembly before a single strategic decision gets made. That range depends on how well the case-study library is maintained, and for most agencies it isn't maintained at all. It's three people's memory and a Dropbox folder nobody owns.

What Evaluators Are Actually Scoring: Proof Density Over Prose Polish

Procurement and marketing evaluators tend to reward claim-to-evidence density, specific outcomes tied to analogous work, over well-crafted sentences about capability. A mediocre paragraph carrying three verified proof points routinely outscores a polished paragraph carrying zero. Proof density means every claim in a response carries a source traceable to a real program record: an attendance figure, a media value, an activation footprint, a client-reported result. When an evaluator can follow a claim back to a methodology, the response earns credibility that prose quality alone cannot manufacture.

Your best case studies are trapped in old decks and in people's heads. That's the real bottleneck, and it's the one most pursuit tools never touch, because they're built to help you write faster, not to help you find what's true.

Why Do Fabricated Proof Points Lose Pitches at the Table, Not in the Scoring Room?

Approximated metrics tend to survive written scoring, where a confident sentence reads fine on the page. They collapse during oral presentations and follow-up Q&A, when a committee member asks for the methodology behind a specific number and the team has no real answer. The scoring sheet rarely records the failure; it just shows a close loss.

The mechanism is what makes this invisible in a post-mortem. The sheet doesn't log 'unverifiable metric,' it logs a drop in confidence about the team's execution capability. The bid team sees a tight score and diagnoses a narrative problem. The actual failure was evidentiary, and it happened live at the table.

"An AI that invents your proof points will lose you the pitch. Grounding isn't a feature, it's the whole game," said Brian Morgan, Founder at Sandbox Group LLC (2025). That's the design principle behind Pitch Box: every claim it drafts carries a source tag back to the agency's own program record, so the follow-up question gets a traceable answer instead of a blank stare.

Grounded Generation: How Pitch Box Drafts From Evidence, Not Assumptions

Grounded generation is a drafting method where every claim in a proposal is retrieved from an agency's own verified records rather than generated from a language model's general training data. It works by pairing RFP requirements to matching proof in a case-study library before a single sentence gets written, so the draft is built on what actually happened, not what sounds plausible.

Pitch Box runs that method in five steps:

1. Ingest the RFP and extract every requirement, evaluation criterion, and buying-committee detail, parsing a full RFP of roughly 26 sections in about 60 seconds.
2. Retrieve matching proof from the agency's own case-study library and self-building knowledge base, which scrapes the agency's site and past work so the library compounds with every pursuit.
3. Draft each section with 100% of claims traceable to that evidence and 0 invented facts, bracketing anything the knowledge base can't yet support.
4. Run the Bid Qualifier for a go or no-go call before senior hours go into a long shot.
5. Hand a sharpened draft to the pursuit team, who own the final response.

The system connects into tools agencies already run, including Google Drive, Box, Slack, HubSpot, Salesforce, Asana, and Canva, with a built-in workflow layer for task and reviewer management. None of that replaces the pursuit team's judgment. It removes the blank page and the scramble to find the right proof point, so senior hours go to the work that actually needs them.

Built by Someone Who Lived the RFP Grind From the Inside

Pitch Box comes out of Sandbox Group LLC, founded by Brian Morgan, a designer, coder, and operator with 15 years in brand experience and experiential marketing. He has been on the subcontracting side of the pitch grind himself, working with experiential agencies and delivering programs for clients like Intel and Alphabet. That's not a resume line dropped in for credibility. It's the reason the product is built around retrieval and proof instead of prose generation: he watched the evidence problem happen from inside the pursuit team, not from a whiteboard.

The economics hold up under simple arithmetic, not a vendor-supplied efficiency claim. If a 100-person agency running 40 qualified pursuits a year cuts research time by 8 hours per bid at a $175 blended hourly rate, that's $56,000 recovered across the pursuit team annually, money that was previously spent hunting for a case study that already existed somewhere in the building.

"The RFP grind is a capacity problem, not a talent problem. Your best people aren't slow, the evidence is scattered," said Brian Morgan, Founder at Sandbox Group LLC (2025). Pitch Box's own knowledge base is the direct answer to that scattering: proof that used to live in one person's head now lives in a system the whole shop can query.

Run the Pre-Bid-Season Audit Before the Next RFP Lands

You don't need a faster blank page. You need your own wins, retrievable at pursuit speed. Before the next contested RFP hits, run three questions against your current setup:

1. Locate where your proof actually lives today, not where it's supposed to live.
2. Identify who can retrieve it inside a 48-hour deadline, and what happens if that person is out.
3. Test what happens when the person who ran your last analogous program is unavailable to answer a follow-up question.

If those answers make you uncomfortable, the gap isn't a writing gap. It's an evidence architecture gap, and it's the one Pitch Box was built to close: ingesting the RFP, drafting from your own case studies with zero invented facts, running a go or no-go before senior hours are spent, and handing your team a draft they sharpen and own. It's priced for the engine, not the seats, so the whole pursuit team works from the same evidence without a per-user conversation standing in the way.