Enterprise Adoption
Engineers loved the pilots. Then the security teams arrived, and deals that looked certain went quiet for months. This part follows the company's first year of selling to large enterprises: a funnel that loses most of its value in one stage, a ten-item questionnaire, and the discovery that the most persuasive sales material was the paperwork written for the EU.
deal_cycle.py and security_review.py were run (real output below). Conversion rates, durations, contract values and the questionnaire status are scenario numbers; the questionnaire items are the kind enterprise security teams commonly ask AI vendors. Ideas from the owner's startup notes are labelled. Disclosure: Claude, made by Anthropic, helps write this site.
1. Month 44: the deal that went quiet
Daniel Okafor had sold enterprise software for fifteen years, and he knew the sound of a deal going quiet. The pilot at the logistics company had been the best they had run: three teams, four weeks, sixty percent of Maestro's changes merged. The engineering director had sent a thank-you note.
Then the security questionnaire arrived. Two hundred and twelve questions. Daniel forwarded it to Arjun with a single line: We need this back in ten days or we lose them.
It took twenty-three. They did not lose the logistics company, but they lost three others that month in exactly the same place.
2. The map: from a first meeting to a signature
๐ธ๏ธ Where enterprise deals are won and lost
The funnel across the top, the questionnaire underneath, and trust as the thing every stage is really testing.
- The surprise.
- The product wins the pilot. The paperwork wins the contract.
3. The funnel
"""deal_cycle.py โ why enterprise revenue arrives late.
A funnel of enterprise opportunities moves through pilot, security review,
procurement and rollout. Each stage has a conversion rate and a duration.
Rates, durations and contract sizes are SCENARIO assumptions.
"""
STAGES = [("first meeting -> pilot", 0.40, 4), ("pilot -> security review", 0.65, 8), ("security review -> procurement", 0.70, 10), ("procurement -> signed", 0.80, 8)]
def main():
n, weeks = 100.0, 0
for name, r, w in STAGES:
n *= r; weeks += w; print(f"{name:32} {r:4.0%} {w:2} wk {n:5.1f} of 100 left, week {weeks}")
acv = 420_000
print(f"\n{n:.1f} signed per 100 opportunities after {weeks} weeks; ${n * acv / 1e6:.1f}M ACV at ${acv / 1e3:.0f}k each")
print(f"security review alone takes {STAGES[2][2]} of {weeks} weeks and loses {1 - STAGES[2][1]:.0%} of deals")
if __name__ == "__main__":
main()100 opportunities, 30 weeks
Real output. Security review takes 10 of the 30 weeks and loses 30% of the deals that reach it.
- Why security review is where deals die.
- By then the engineers are convinced, so a loss here is a loss of a deal the product already won.
- It is also the stage the vendor controls most: every answer can be prepared in advance.
4. The questionnaire
"""security_review.py โ what enterprise security teams ask, and the gaps.
Items common in enterprise security questionnaires for AI vendors, with the
status at month 42 and weeks to close each gap. Status and weeks are SCENARIO.
"""
ITEMS = [("SOC 2 Type II report", "in progress", 16), ("SSO / SCIM", "done", 0), ("audit logs of every agent action", "done", 0),
("data residency (EU region)", "gap", 10), ("no training on customer code by default", "done", 0), ("customer-managed keys", "gap", 8),
("model card + EU AI Act status", "done", 0), ("agent permission scopes per repo", "done", 0), ("pen-test report (last 12 months)", "gap", 6),
("incident response SLA", "done", 0)]
def main():
for n, s, w in ITEMS: print(f"{n:42} {s:12} {('+' + str(w) + ' wk') if w else ''}")
gaps = [i for i in ITEMS if i[1] != "done"]
print(f"\n{len(ITEMS) - len(gaps)} of {len(ITEMS)} ready; {len(gaps)} open; longest {max(g[2] for g in gaps)} weeks (run in parallel)")
if __name__ == "__main__":
main()Ten questions every security team asked
Real output. All four open items run in parallel; the longest is SOC 2 at 16 weeks.
Hana noticed it first. Half the questions that took Arjun longest were already answered, in documents nobody in sales had seen: the model documentation and evaluation report written for the AI Act, the incident process, the record of every change the gate had approved or refused.
Daniel turned them into a single trust pack: a forty-page PDF and a live page showing the gate's decisions. The next security review took six weeks instead of ten. We spent two million dollars on compliance, Gopala said at the board meeting, and it turned out to be our best sales asset.
- Owner's notes behind this section.
- Trust as the moat (Rauch): trust comes from quality, reliability and evidence, not from chasing it directly.
- Engineering for verification (Rauch): the bottleneck moves from writing code to proving it is correct. Enterprises buy the proof.
5. Sandbox: fix the security review
๐ฎ Sandbox: fix the security review
Same funnel as deal_cycle.py; other stages fixed. Scenario numbers.
- Try this.
- Cut the review from 10 weeks to 6 (the trust pack): same deals, four weeks sooner.
- Raise the pass rate from 70% to 90%: about four more signed deals per 100.
- Do both: roughly $1.8M more annual contract value per 100 opportunities, and a shorter cycle.
6. Five mistakes
- Treating the pilot as the sale.
- Answering each security questionnaire from scratch.
- Hiding compliance documents from the sales team.
- Promising data residency before it exists.
- Measuring sales by pilots started instead of contracts signed.
7. Delegation box
- The trust pack and what it promises
- Drafting questionnaire answers from existing documents
- Keeping the trust page current
- SOC 2 auditor
- External penetration testers
8. Budget and risk tracker
| Item | Status after Part 18 | Note |
|---|---|---|
| Enterprise funnel | 14.6% to signature | 30 weeks |
| Security review | 6 of 10 ready | trust pack cuts it to ~6 weeks |
| SOC 2 | In progress | 16 weeks |
| Trust | Evidence-led | gate record public to customers |
9. The series
The Maestro, Season 2: 23 parts in five phases (Season 1: 18 parts)
10. Reading the numbers
- Primary / standard
- Enterprise security questionnaire topics (common industry practice: SOC 2, SSO, data residency, encryption keys, penetration tests).
- Secondary โ ๏ธ
- None in this part.
- Our assumptions
- Rates, durations, contract values, questionnaire status and story events are the scenario. Owner's notes (Rauch) are a summary, not a transcript.