OTA & CSO Proposal Pack

Real session, run 2026-10-04 with Claude Sonnet. The company and the area of interest are made up for the test. The DIU process details come from DIU's 2026 CSO guidebook.

Your 9-day DIU pitch plan (pitch day: 2026-10-13)

Do these three things today:

1. Tell your CEO the 28-slide investor deck is out. Cut it to about 8 slides and 20 minutes.
2. Start SAM.gov registration (UEI and CAGE code). It takes weeks, and you need it before any award.
3. Pick your pitch team: three people at most. The CEO opens and closes. Your best engineer answers technical questions and runs the demo.

What this pitch is

Format: 60 minutes, strictly timed. 20 minutes of presentation, 40 minutes of questions. The questions are most of the evaluation, so most of your rehearsal goes there.

Outcomes: (a) advance to a Request for Prototype Proposal, (b) portal hold (DIU may come back within 180 days), or (c) non-select.

They already checked you. DIU's business intelligence team looks at ownership, runway, leadership, and security before the pitch. Assume they know your cap table and burn.

Two lenses. Mission: relevance to the AOI, technical merit, innovation, schedule, data rights. Commercial: company viability, rough order of magnitude (ROM) cost, schedule, uniqueness.

Why the investor deck fails here

An investor deck sells market size and growth. This audience is a government end user and a commercial evaluator. They want to know whether you solve the problem as the AOI wrote it. Use the AOI's own words, not yours. Treat your own phrasing and the AOI's as different things until you've checked.

Open with the mission impact in numbers. For example, your product does X in minutes where today it takes Y. Use your real numbers from the insurance and ag deployments. If you don't have them, get them this week.

The 20-minute deck (about 8 slides)

Present for no more than a third of the session. Put the demo early, because evaluators remember what they saw.

1. The end user's problem, in the AOI's words (1 slide). Name who has this problem today and what it costs them.

The product working (1-2 slides, or live). Show the operator's workflow with before-and-after time. A live demo only if it's reliable. Always keep a recording as backup, with live narration.
3. How it works (1-2 slides). Modular architecture, how it integrates, and where it runs.
4.

Evidence (1 slide). Two insurance customers and one ag customer, with volumes, accuracy numbers, and any failure data you have. One step-change number (think 10x, not 10%) and one differentiator a competitor can't copy in six months (proprietary data, a feedback loop from real users).
5.

The prototype plan (1-2 slides). Milestones every 1-3 months, measurable success criteria, and what the government touches at day 30.
6. Path to production (1 slide). Who the government partner is and what funding is planned after the prototype.
7. Team and company (1 slide). Short. Runway and customers go in your Q&A answers, not on a slide.
8. The ask (1 slide).

The investor material is not wasted. Your customer traction, runway, and unit economics become your answers to the commercial-lens questions.

Your code stays yours: how to say it

DIU starts Phase 3 from your commercial terms, lightly tailored. So you can be firm, but state it cleanly and early, because the question is coming ("which code is the government's and which is yours?").

The usual position for a commercial company:

• Background IP and your core product stay yours. The government gets only what it needs to evaluate the prototype.
• Integration layers built for the government can be the government's.
• Commercial software goes under your commercial license. New code developed under the prototype gets government purpose rights, ideally time-limited.
• Offer source code escrow for sustainment concerns instead of delivering code.

Your one-sentence answer: "The core product stays commercial; integration layers built for you are yours; escrow covers sustainment." Do not say "we'd sort that out in negotiation." That's the weak answer.

Counsel review required before you commit to any IP, data-rights, or liability position. Book your lawyer for this week. Also have them check how Series A investor rights interact with these positions.

An Other Transaction (OT) isn't bound by the FAR or its standard data-rights clauses. Everything is negotiated, which is good for you, but you have to negotiate.

Gaps first-time companies get caught on

1. SAM.gov. Start now (see above).

Nontraditional status. Confirm you have had no contract subject to full Cost Accounting Standards (CAS) coverage in the past year. A 30-person startup with commercial customers almost certainly qualifies, but confirm it. Don't claim it until you have.
3.

Upstream licenses. If you use commercial satellite imagery, third-party data, open-source code, or third-party models, their licenses may limit what you can deliver to the government. Evaluators ask. Pull the licenses and know the answer by day 3.
4.

Defense data. Your results come from commercial data. Say plainly what hasn't been tested on defense data, propose the test, and explain how it deploys disconnected from the public cloud (if the AOI requires that).
5.

Ownership and foreign investment. Check your investor list and your hires for foreign ownership or export-control exposure. DIU will have looked. Write a one-paragraph compliance answer.
6.

Runway. Know your months of runway and what happens to the prototype if funding slips six months or you don't raise.

The 15 questions to rehearse

A strong answer has a number and a place it already happened. Keep each under 45 seconds, answer what was asked, then stop. If you don't know, say so and say how the prototype will find out.

Mission side

1. Which specific item in the AOI does each feature handle, and where does it already run at volume?
2. What happens if bandwidth or input quality drops sharply? (Have logs from real installs.)
3. Show us the operator's workflow. What step does it remove, and how much time does it save?
4. What does the government actually touch at day 30?
5. How is this different from what we already use or from known competitors?
6. What happens if you miss a milestone?
7. Can it run on government networks, or disconnected?
8. What hasn't been tested on defense data, and how would you test it?

Commercial side

9. Which code is the government's and which is yours if scope changes?
10. What data rights do you assert, and what does the government get?
11. What's your ROM, and what does production cost per unit and at what quantity?
12. What breaks if usage triples? (Numbers from a real load event.)
13. What gets cut if funding slips six months, and who still pays you? (Named customers, months of runway.)
14. Why can't a competitor copy this in six months?
15. How do you handle export control, supply chain, and foreign ownership?

Split the questions by category in advance so two people don't answer at once.

Day-by-day plan

What to ask in the session

Ask who the government partner is and what funding is planned after the prototype. A successful prototype can lead to a follow-on production award without a new competition (10 U.S.C. 4022(f)). Their answer tells you how real the opportunity is.

After the pitch

Advance: you draft the statement of work with the government team. Pricing starts from your commercial baseline; terms start from yours. Ask about transition funding at kickoff, and get "successful completion" written down in measurable terms because that's what unlocks the follow-on.

• Portal hold: keep your DIU contact updated with new customers and test results for the next 180 days.
• Non-select: ask for feedback.

What I need from you

• The AOI text and the invitation.
• Your pilot numbers.
• Your imagery and data sources.
• Whether you've held any CAS-covered contract in the past year.
• Who is on the cap table, and any foreign investors.

I'm not giving legal advice. Counsel needs to review your IP, data-rights, and liability positions before you commit to them. Keep export-controlled or controlled unclassified material out of any AI tool; work from public descriptions.