Three serious conversation-intelligence tools compete in the trades in 2026 — Rilla for in-person sales coaching and the virtual ride-along, CallRail Conversation Intelligence for inbound-call sentiment and lead-source attribution, and ServiceTitan Conversational AI for in-platform CSR QA and call-summary scale. They are not competing for the same surface. Rilla owns the kitchen-table close and the in-truck advisor coaching conversation. CallRail owns the inbound call before and during the booking, the missed-opportunity flag, and the lead-source attribution chain. ServiceTitan Conversational AI owns the in-platform CSR QA, the post-call summary, and the consolidated reporting line that Titan-native shops want. The owner choosing among them is rarely choosing one or none — most shops above 6 trucks deploy two or all three, mapped to specific coaching, attribution, and QA workflows that the owner builds the rubric for. The wrong pick costs $40K-$120K of margin — a Rilla deployment without service-manager scorecard cadence wastes the seat licenses; a CallRail deployment without GLSA-bidding feedback corrupts the attribution; a ServiceTitan Conversational AI deployment without a CSR QA review routine produces summary noise. This lesson is the rubric the owner runs in the 30-day pilots, the 5 criteria that decide which surface gets which tool, the questions to ask each vendor on the demo, and the side-by-side scoring table that converts vendor pitch decks into deliberate decisions.

The Three Contenders and Where Each Actually Sits

Rilla is the virtual-ride-along category leader by published case studies and by feature surface for in-person sales coaching. Rilla's 2026 documented benchmarks: 30-40 virtual ride-alongs per service manager per day (vs. 2-3 in-person), 18% close-rate lift across home-services deployments, deployed across hundreds of multi-location operators in 2025-2026. The tool records the in-truck or in-home conversation between the Comfort Advisor and the customer, transcribes it, and surfaces the 5 highest-impact coaching moments (intro, system condition explanation, repair-vs-replace pivot, options presentation, financing pivot). Pricing in 2026: $200-$400 per seat per month, where seats are service managers and Comfort Advisors. Best fit: shops with 2+ Comfort Advisors, a service manager who has time for daily 15-30 minute transcript review, and a comp plan tied to close rate. Not built for inbound-call analysis or CSR QA.

CallRail Conversation Intelligence is the inbound-call analytics category leader in the trades, owning the surface between the call landing and the FSM record being written. CallRail's 2026 capabilities: every inbound call routes through a tracked number with source attribution; the AI transcribes, tags intent (service vs. sales vs. supplier vs. complaint), flags missed opportunities, scores sentiment, and writes back to the FSM platform (ServiceTitan, Sera, Housecall Pro, Jobber). Critical for marketing attribution — Ryze AI's GLSA bid optimization and Hatch's stale-lead reactivation both need CallRail's lead-source-to-revenue chain to function. Pricing in 2026: $145-$995/month depending on call volume tier and intelligence feature add-ons. Best fit: shops running GLSA, multi-source marketing, multiple tracked numbers; shops that need the missed-opportunity dashboard and the per-source ROAS. Not built for in-person sales coaching or in-truck ride-along.

ServiceTitan Conversational AI is ServiceTitan's in-platform conversation analytics, sitting under the Titan Intelligence umbrella. The 2026 release set includes call summaries on every CSR call, AI-tagged intent classification, sentiment scoring, CSR QA dashboards integrated with the CSR scorecard, and post-call action-item extraction (book the follow-up, send the financing pre-qual link, escalate to dispatch). Pricing: bundled into Titan Intelligence add-on at $300-$600/month per shop. Best fit: ServiceTitan-native shops that want consolidated reporting and CSR QA without adding a second vendor stack. Not built for in-person sales coaching (cannot record kitchen-table conversations) or for cross-platform lead-source attribution (ServiceTitan-only by design).

The owner choosing among them is choosing per-surface, not one-vs-the-others. A 6-truck residential HVAC shop on ServiceTitan with 2 Comfort Advisors typically deploys all three: Rilla for advisor coaching (3 seats), CallRail for inbound attribution, ServiceTitan Conversational AI for in-platform CSR QA. Total: $1,200-$2,500/month. The owner's question is not "which one" — it is "which workflow does each cover and what is the overlap risk."

The Five Criteria That Decide the Bake-Off

The bake-off is not won on a single metric like the voice-agent bake-off (where booking % is the obvious headline). Five criteria score the conversation-intelligence stack in 2026. Each is scored 1 to 5 during the 30-day pilot.

Criterion 1: Coaching depth — what the tool tells the manager. Does the tool surface the 5 highest-impact coaching moments per ride or call, with timestamps and verbatim quotes? Or does it produce a summary the manager has to re-listen to? Rilla wins coaching depth by a wide margin on the kitchen-table close — the tool was built for exactly this. ServiceTitan Conversational AI surfaces CSR coaching moments on inbound but cannot record kitchen-table conversations. CallRail surfaces missed-opportunity moments on inbound but does not coach on the close itself.

Criterion 2: Attribution chain — what the tool feeds into the marketing P&L. Does the tool write lead source, call outcome, and revenue back to the FSM platform with confidence, or does the marketing manager re-tag manually? CallRail owns attribution by design — every tracked number, every source, every outcome flowing to ServiceTitan / Sera / HCP / Jobber. ServiceTitan Conversational AI attributes within ServiceTitan only — perfect for ServiceTitan-only marketing but loses signal on cross-platform sources. Rilla is not built for attribution — the in-truck conversation is post-attribution.

Criterion 3: CSR QA cadence — what the floor coach uses daily. Does the tool produce a CSR-by-CSR scorecard with quote-level evidence the floor coach reviews daily in a 15-minute huddle? ServiceTitan Conversational AI excels here because the CSR scorecard ties to the FSM platform's call-handle metrics natively. CallRail produces CSR-level intelligence but the integration depth into ServiceTitan's CSR scorecard is shallower. Rilla does not do CSR QA — it is the wrong tool for the floor.

Criterion 4: Integration depth with FSM and marketing stacks. Does the tool's data flow into ServiceTitan / Sera / Housecall Pro / Jobber, CallRail / GLSA, Hatch, and the owner's dashboard without manual re-entry? CallRail's connector maturity across FSM platforms is the deepest in 2026 (sub-30-second sync, all lead-source fields, no breakage on platform updates). ServiceTitan Conversational AI is native by definition for ServiceTitan shops. Rilla integrates with ServiceTitan via case-study connectors but the depth on Sera or HCP is lighter.

Criterion 5: Cost per insight — the floor coach's economics. Compute the per-insight cost: monthly tool cost ÷ coaching or attribution insights surfaced and acted on. Rilla at $300/seat × 3 seats = $900/month ÷ 200-400 actionable coaching moments = $2-$5 per coaching moment. CallRail at $400/month ÷ 50-150 missed-opportunity flags = $3-$8 per flag. ServiceTitan Conversational AI at $500/month ÷ 300-600 CSR QA touches = $1-$2 per touch. Per-insight cost is the operating economics; the decision still depends on which surface each tool covers, not on per-insight cost alone.

The composite of the five criteria does not collapse to a single winner because the three tools serve different surfaces. The composite tells the owner whether each tool justifies its place in the stack at the shop's call volume, advisor count, and marketing complexity. The 30-day pilot scores each tool against the criterion most relevant to its surface; the owner commits or walks per tool.

The Questions to Ask Each Vendor on the Demo Call

Demo calls follow vendor patterns: Rilla plays a coaching highlight reel; CallRail walks through the missed-opportunity dashboard; ServiceTitan Conversational AI shows the call-summary integration. The owner's job is to break each pattern with surface-specific questions that expose the real depth.

For Rilla. (1) Show a 15-second clip from a real anonymized pilot where the AI surfaced the "I need to think about it" moment, the advisor's response, and the manager's next 1:1 coaching. (2) Median close-rate lift across bottom-quartile pilot shops, not the top. (3) Walk through the ServiceTitan integration — does close-rate-by-advisor scorecard auto-update or hand-enter? (4) Comp-plan tie in shops that landed the 18% close-rate lift — restructured before Rilla went live or after? (5) Two-party-consent disclosure language at the kitchen table.

For CallRail Conversation Intelligence. (1) A 30-second clip where the AI flagged a missed opportunity in real-time — the trigger, the CSR's action, what the marketing manager saw Friday. (2) Median GLSA cost-per-booked-call lift after CallRail-to-Ryze-AI bidding wiring — pilot distribution. (3) ServiceTitan / Sera / HCP / Jobber connector — sync latency, fields, breakage cadence on platform updates. (4) Sentiment-detection accuracy and false-positive handling. (5) AEO contribution — does CallRail's tracked number get picked up when ChatGPT or Perplexity surfaces the shop's contact?

For ServiceTitan Conversational AI. (1) CSR scorecard view with quote-level evidence on a real anonymized CSR — daily huddle pull. (2) Median CSR-booking-percent lift in 60 days, plus bottom-quartile band. (3) Call summary write-back — auto-write or manager approval? (4) Avoca cross-tool integration on Avoca-plus-Titan-Intelligence shops. (5) Cross-platform attribution road map — ever attribute calls from non-ServiceTitan sources?

Vendors who answer with audio, pilot distributions, and integration specifics are durable. Vendors who flip to slides, quote case-study tops, or hand-wave on integration are warning the owner about pilot conversion risk. Walk.

What to Test in the 30-Day Pilot per Tool

The 30-day pilot framework runs three parallel pilots — one per tool — with shared baseline metrics from week 0 and tool-specific measurement workstreams from week 1.

Week 0 baseline (shared across all three pilots). Booking %, missed-call %, CSR average-handle-time, close rate per Comfort Advisor (90-day rolling), average ticket per advisor, MPR per advisor, GLSA cost-per-booked-call, source attribution accuracy.

Rilla pilot workstreams. (1) Daily ride-along review count by service manager (target 25-35/week vs. 8 baseline). (2) 5-moment coaching extraction audit (sample 10 rides/week). (3) Close-rate-by-advisor delta (target 4-7 point lift by day 30; 8-14 by day 60). (4) Comp-plan and advisor reaction survey week 2 and week 4. (5) Two-party-consent compliance log — every recording disclosed.

CallRail pilot workstreams. (1) Source attribution accuracy (audit 50 calls/week, target 92%+). (2) Missed-opportunity flag review (every flag reviewed within 24 hours; outreach within 48; conversion-to-booking measured). (3) GLSA bid feedback (CallRail-to-Ryze-AI loop closes within 7 days; bid-optimization lift after week 3). (4) Sentiment-flag accuracy (sample 30/week, target <15% false positive). (5) Integration health log — daily ServiceTitan / Sera / HCP / Jobber write-back check.

ServiceTitan Conversational AI pilot workstreams. (1) CSR scorecard adoption (daily score with quote-level evidence; utilization above 80%). (2) Call summary write-back accuracy (sample 20/week; target 90%+ captures equipment, mood, next-step). (3) Action-item extraction (booked follow-ups, dispatched callbacks, escalations — sample 30/week). (4) CSR booking-percent lift on AI-tagged calls (target 3-7 point lift by day 30). (5) Floor coach time savings (target 50%+ reduction in CSR coaching prep time).

Day 30 commit-or-walk decision per tool. Rilla: keep if close-rate lift hit 4+ points and advisors aren't in revolt. CallRail: keep if attribution accuracy above 92% and missed-opportunity recovery converting at 25%+. ServiceTitan Conversational AI: keep if CSR scorecard adoption above 80% and floor-coach time savings above 40%. Walk on any tool that misses its core threshold. Expand on any tool that exceeds it.

The Side-by-Side Scoring Table

The decision artifact is a one-page scoring table with the five criteria on rows, the three tools on columns, and a 1-5 score in each cell. The table converts the three vendor pitch decks into a single deliberate decision.

Coaching depth row. Rilla: 5 (built for it). CallRail: 2 (not its surface). ServiceTitan Conversational AI: 3 (CSR coaching but not kitchen-table).

Attribution chain row. Rilla: 1 (post-attribution by design). CallRail: 5 (category leader). ServiceTitan Conversational AI: 3 (ServiceTitan-only attribution).

CSR QA cadence row. Rilla: 1 (wrong tool for the floor). CallRail: 3 (some CSR intelligence but shallow scorecard). ServiceTitan Conversational AI: 5 (native to FSM scorecard).

Integration depth row. Rilla: 3 (case-study connectors for ServiceTitan; lighter on Sera/HCP). CallRail: 5 (deepest cross-platform connectors). ServiceTitan Conversational AI: 5 (native).

Cost per insight row. Rilla: 4 ($2-$5 per coaching moment). CallRail: 4 ($3-$8 per flag). ServiceTitan Conversational AI: 5 ($1-$2 per touch).

Composite per tool: Rilla 14/25 (excels on coaching, weak on attribution and CSR QA), CallRail 19/25 (excels on attribution and integration, weak on coaching), ServiceTitan Conversational AI 21/25 (excels on CSR QA and integration, mid on coaching, weak on cross-platform attribution).

The composite numbers do not pick "the winner" — they pick the per-surface coverage. Rilla wins the coaching surface. CallRail wins the attribution surface. ServiceTitan Conversational AI wins the CSR QA surface. The 6-truck shop deploys all three at $1,200-$2,500/month combined. The decision the owner defends is not "which one" but "which surfaces, which tools, and which overlap risks am I accepting" — and the table is the artifact that answers it.

Decision Rules by Shop Size and FSM Platform

The default recommendations distilled from the 2026 pilot data and the five-criteria scoring.

1-3 trucks on any FSM. Skip Rilla — under 2 Comfort Advisors, the seat-license economics break and the service manager doesn't yet have a daily transcript review routine. Run CallRail at the entry tier ($145-$245/month) for source attribution and missed-opportunity flags. Add ServiceTitan Conversational AI (if on ServiceTitan) or HCP / Jobber native call-summary AI when bundled in the FSM tier. Total: $145-$500/month.

4-7 trucks on ServiceTitan or Sera. Rilla for the 2 Comfort Advisors + 1 service manager = 3 seats × $300/month = $900/month. CallRail Conversation Intelligence mid-tier = $290-$495/month. ServiceTitan Conversational AI = $300-$600/month. Total: $1,500-$2,000/month. The 5-criteria composite supports the full stack at this size; Rilla unlocks the close-rate lift, CallRail unlocks the marketing attribution chain, ServiceTitan Conversational AI unlocks the CSR QA.

8-15 trucks on ServiceTitan or Sera. Same full stack with more Rilla seats (4-6 seats = $1,200-$1,800/month), CallRail higher tier ($495-$795/month at higher volume), ServiceTitan Conversational AI at the same per-shop rate. Total: $2,000-$3,200/month. Per-truck cost drops 20-30% vs. the 4-7 truck case due to vendor pricing tier compression.

16-40 trucks on ServiceTitan or Sera. Rilla 8-15 seats ($2,400-$4,500/month). CallRail enterprise tier ($795-$995/month). ServiceTitan Conversational AI at per-shop rate. Total: $3,500-$6,000/month. Per-truck cost continues to compress.

40+ trucks or multi-shop / PE platform. Platform-mandated. Most PE platforms standardized Rilla cross-portfolio (2024-2026), CallRail cross-portfolio (2023-2026), and ServiceTitan Conversational AI or BuildOps Conversational equivalents in 2025-2026. Per-location cost band $4K-$8K/month combined.

Shops on Housecall Pro or Jobber substitute ServiceTitan Conversational AI with their native call-summary AI (HCP AI Agents summary, Jobber Copilot summary) and run CallRail + Rilla as the bolt-on stack. The composite shifts slightly — HCP and Jobber natives score lower on CSR QA depth — but the cost band stays similar.

Overlap and the Stack Discipline

The risk in running all three tools is overlap — call summaries from both CallRail and ServiceTitan Conversational AI on the same inbound call, coaching signal from both Rilla (kitchen-table) and ServiceTitan Conversational AI (CSR floor) that the service manager has to reconcile, attribution from CallRail and ServiceTitan that may disagree on lead source. The stack discipline names which tool owns which decision.

Call summary discipline. ServiceTitan Conversational AI owns the call summary for the CSR scorecard. CallRail's call transcript informs the marketing attribution and missed-opportunity flag but does not write to the customer record. The CSR's after-call work reads from ServiceTitan Conversational AI; the marketing manager's Friday review reads from CallRail. No overlap; clean ownership.

Coaching discipline. Rilla owns kitchen-table close coaching for Comfort Advisors. ServiceTitan Conversational AI owns CSR floor coaching on inbound booking. The service manager runs daily review on both — 15 minutes on Rilla transcripts for advisor 1:1 prep, 15 minutes on ServiceTitan Conversational AI for floor coach huddle. Different audiences, different surfaces. No conflict.

Attribution discipline. CallRail owns lead-source attribution from the inbound call to the closed revenue. ServiceTitan Conversational AI's attribution is internal-only and informs CSR scorecard but does not drive marketing decisions. Marketing manager reads CallRail; ServiceTitan attribution is the FSM-internal view. Conflicts resolved in favor of CallRail.

The stack discipline is the third artifact in the L4 capstone defense after the 5-criteria scoring table and the per-tool 30-day pilot result. The owner who walks into the peer-group or PE call with the scoring table plus the pilot results plus the stack discipline memo has documented the tool decisions across the conversation-intelligence surface. The owner who walks in with "we run all three" without the ownership rules has invited the question "where do they conflict" with no answer prepared.

Key Takeaways

• Three serious conversation-intelligence tools in 2026: Rilla (kitchen-table close coaching, $200-$400/seat/month, 18% documented close-rate lift), CallRail Conversation Intelligence (inbound-call attribution and missed-opportunity flagging, $145-$995/month by tier), ServiceTitan Conversational AI (in-platform CSR QA and call summaries, $300-$600/month in Titan Intelligence bundle).
• The tools serve different surfaces — Rilla owns kitchen-table coaching, CallRail owns inbound attribution, ServiceTitan Conversational AI owns CSR QA. Most shops above 6 trucks deploy two or all three; the owner's question is "which surfaces, which tools, which overlap" not "which one."
• Five criteria score the bake-off: (1) Coaching depth — what the tool tells the manager. (2) Attribution chain — what the tool feeds into the marketing P&L. (3) CSR QA cadence — what the floor coach uses daily. (4) Integration depth with FSM and marketing stacks. (5) Cost per insight.
• Composite scores expose per-surface coverage, not a single winner. Rilla 14/25 (coaching strong; attribution and CSR QA weak). CallRail 19/25 (attribution and integration strong; coaching weak). ServiceTitan Conversational AI 21/25 (CSR QA and integration strong; cross-platform attribution weak).
• Vendor demo questions break the pitch. For Rilla: real coaching moment audio with verbatim quotes, bottom-quartile pilot distribution, ServiceTitan integration depth, comp-plan tie precedence, two-party-consent disclosure language. For CallRail: missed-opportunity flag clip, GLSA-to-Ryze-AI bidding lift band, connector sync latency, sentiment-detection accuracy, AEO contribution. For ServiceTitan Conversational AI: CSR scorecard view with quote-level evidence, CSR booking-percent lift distribution, call summary write-back logic, Avoca cross-tool integration, cross-platform attribution road map.
• 30-day parallel pilots score each tool against its surface. Rilla: ride-along count, 5-moment extraction, close-rate-by-advisor delta, advisor reception, consent log. CallRail: attribution accuracy, missed-opportunity recovery, GLSA feedback loop, sentiment accuracy, integration health. ServiceTitan Conversational AI: scorecard adoption, write-back accuracy, action-item extraction, booking lift, floor-coach time savings.
• Day-30 commit-or-walk thresholds per tool. Rilla: 4+ point close-rate lift, advisor adoption. CallRail: 92%+ attribution accuracy, 25%+ recovery conversion. ServiceTitan Conversational AI: 80%+ scorecard adoption, 40%+ floor-coach time saved.
• Decision rules by shop size. 1-3 trucks: skip Rilla, run CallRail entry + native FSM summary. 4-7 trucks: full stack at $1,500-$2,000/month. 8-15 trucks: full stack at $2,000-$3,200/month. 16-40 trucks: scaled stack at $3,500-$6,000/month. 40+ trucks or PE platform: platform-mandated at $4K-$8K per location.
• Stack discipline names ownership. ServiceTitan Conversational AI owns CSR call summaries; CallRail owns lead-source attribution; Rilla owns kitchen-table coaching. The discipline memo is the third L4 capstone artifact after the scoring table and the pilot results.
• The scoring table is the decision artifact. Owners who walk in with the table, pilot results, and stack discipline get "what do you need to support it." Owners who walk in with "we run all three" get "where do they conflict, and how do you know."