Builder AI Pilot RFP Template (18 Questions) | Jome

A builder VP sent us her draft RFP for an AI sales vendor evaluation last quarter. It was 11 pages of generic enterprise software questions — security frameworks, SLAs, uptime, data residency. It would have screened for a CISO's approval and missed every question that actually predicts whether the AI works in a builder sales org.

This post is the replacement. Eighteen questions, built around the work an OSC actually does, organized so the answers tell you whether the vendor understands new home sales or just pitched their roadmap.

Drop it into your next vendor email as-is. Score the responses. The shortlist will reveal itself in 48 hours.

TL;DR — how to use this template

• Send all 18 questions in one email. Vendors who can't answer all 18 in writing inside a week aren't ready.
• Score each answer on a 1-3 scale: 1 = vague or wrong, 2 = acceptable, 3 = specific and credible.
• Total possible: 54. A vendor under 36 is probably not your fit.
• Time-box the response window to 7 business days. Builder pilots that drag through long evaluations rarely launch.

The 18 questions, organized in six sections

Section 1 — Data fit (3 questions)

These tell you whether the AI actually speaks new home sales.

Q1. How many real new-construction sales conversations is your model trained on? (Looking for: a number with a source, not "we use a great LLM.")

Q2. Show me a transcript of a sample call your AI ran with a builder lead. Include the buyer's first three turns and your AI's responses. (Looking for: industry-correct vocabulary — OSC, community, division, spec, TBB, co-op — and a conversation that sounds like sales, not customer service.)

Q3. How does your AI handle community-specific facts — lot premiums, included features, design center pricing, current incentives? (Looking for: a real loading process, ideally automated from the builder's existing systems, not "we'll have your team write a script.")

Section 2 — Channel coverage (2 questions)

These tell you whether the AI can run a real cadence.

Q4. Voice, SMS, email, voicemail handling, inbound, outbound — which of these does your platform handle natively, and which require a separate tool? (Looking for: all six, on a coordinated cadence. Anything less and you'll be stitching together orchestration yourself.)

Q5. Walk me through how your platform decides when to call vs text vs email a specific lead. (Looking for: a real cadence engine that adapts by buyer behavior — opens, callbacks, no-shows, time-of-day — not a fixed sequence.)

Section 3 — Integration (3 questions)

These kill more vendor projects than any other category.

Q6. Do you have a named integration with [your CRM — Lasso, BuilderTrend, Salesforce, Pipedrive, custom]? (Looking for: a yes with a screenshot or a customer reference, not "we have an API.")

Q7. What's the typical integration timeline from contract signature to live calls? (Looking for: 24-72 hours for standard CRMs. Anything over two weeks is a flag.)

Q8. What does your AI write back to the CRM after every interaction? Include a sample of an actual lead-record update. (Looking for: structured intent tags, stage updates, objection categories, and conversation summaries — not "call completed.")

The good news for builders running Lasso, BuilderTrend, Pipedrive, or Salesforce: AI sales extensions like Jome integrate with pretty much every CRM your team already uses without requiring a migration. The lift is configuration, not stack replacement. (For deeper context on how AI tools coexist with Lasso specifically, see our Lasso CRM alternatives and complements analysis.)

Section 4 — Control and guardrails (3 questions)

These tell you whether the AI is yours to operate or theirs to manage.

Q9. Can my team change scripts and cadences without engineering involvement? Show me the interface. (Looking for: a real configuration UI, not "email us and we'll update it.")

Q10. What topics will the AI refuse to discuss, and how is that configured? Pricing, financing terms, delivery dates, warranty — what's locked, what's editable? (Looking for: granular per-topic controls plus per-community/division/brand variation.)

Q11. Where do conversation logs live, and who has access? Walk me through the audit trail. (Looking for: searchable transcripts, timestamps, every send and every consent basis logged. (FCC TCPA guidance is the floor here.))

Section 5 — The human handoff (3 questions)

This is the architectural decision most vendors gloss over.

Q12. What triggers a handoff from AI to my OSC? Walk me through the exact decision logic. (Looking for: clear triggers — buyer requests, intent thresholds, complexity rules — not "the AI just knows.")

Q13. When the handoff happens, what does my OSC see? (Looking for: a full conversation summary, intent flags, suggested next action, and a transcript link — not just "lead is hot.")

Q14. What happens if my OSC isn't available when the buyer is ready to talk? Can the AI re-engage to schedule, or does the conversation die? (Looking for: an in-flight reschedule capability, with the AI handling the calendar booking.)

The right model is: AI does the grind, your team closes. Where the handoff happens defines the whole program. We've covered the OSC-to-OSM handoff playbook in more detail for the human-to-human side of the same problem.

Section 6 — Outcomes (4 questions)

These tell you whether the vendor knows what success looks like in builder sales.

Q15. What does success look like at 30, 60, 90 days for a builder pilot? Give me specific metrics, not "improved engagement." (Looking for: contact rate, response rate, appointments booked, tours, contracts. With ranges, not single numbers.)

Q16. Show me a builder pilot you've run. Numbers. (Looking for: at least one case study with specific outcomes — leads worked, contact rate, contracts. If they can only share aggregated benchmarks, that's a yellow flag.)

Q17. What pricing model do you use for a single-division pilot? (Looking for: month-to-month or short-term commitment with a clean out-clause. If the vendor wants seven figures up front, they're not pricing for a builder pilot — they're pricing for an enterprise software cycle.)

Q18. What does cancellation look like 60 days in if the pilot doesn't work? What data do we keep, what do you keep? (Looking for: clean exit terms, full data export, no penalties. Vendors who can't answer this cleanly often have lock-in baked into the contract.)

How to score the responses

For each question, give the vendor 1, 2, or 3 points.

• 1 — Vague, generic, or wrong. ("Our AI uses advanced technology to deliver results.")
• 2 — Acceptable but not specific. ("We integrate with most major CRMs and have customers on Lasso.")
• 3 — Specific and credible. ("Lasso is a named integration; here's a screenshot of the data flow and a reference customer running it for 14 months.")

Total possible: 54.

• 45-54 = strong shortlist candidate
• 36-44 = run a pilot if data and price look right
• Under 36 = not your fit, regardless of how the demo went

This scoring is rough on purpose. The point isn't precision — it's separating vendors who actually understand builder sales from vendors who pitched their generic AI deck with a builder-flavored cover slide. (Bokka Group's analysis of how AI is changing builder marketing is a useful adjacent reference for category context.)

Common RFP mistakes

Mistake 1 — adding 40 enterprise security questions and burying the sales-ops ones.Security matters. But if the security questions outnumber the ops questions, the RFP will screen for the wrong shortlist.

Mistake 2 — letting the vendor schedule a demo before the RFP is back.Demos are the vendor's home court. The RFP is yours. Stay on yours until the responses are in.

Mistake 3 — skipping the cancellation question.A 60-day exit clause is the most important contract term in this category. If you forget to ask, the vendor won't volunteer.

Mistake 4 — running parallel RFPs to two vendors.Cuts the depth in half. Pick the strongest candidate from the field, run one real pilot, decide.

Mistake 5 — accepting "we'll send pricing after the demo."Get a price band before the demo. The demo will move the price either way; you need a starting number.

What this means for your team

A good RFP doesn't replace a real pilot. It accelerates getting to one. The 18 questions above filter out the 60-70% of vendors who aren't ready for a builder sales op, leaving you with a shortlist of two or three to put in front of your sales VP for a 30-minute demo.

That's two weeks saved per vendor, on a category where most builder evaluations drag for two quarters. Send the email Monday.

Next reads

• Builder Sales AI Myth-Busting — the buying-motion primer
• OSC-to-OSM Handoff Playbook — what the AI-to-human handoff should look like
• What Is Voice AI Latency? — the latency test to add to your RFP
• Why Builders Miss 15-30% of Inbound Calls — the problem the coverage layer solves

What to do Monday morning

Copy the 18 questions into an email. Send to every AI sales vendor that's been in your inbox this quarter. Set a 7-day deadline. Score the responses on a 54-point scale. Schedule demos with the top two.