The first time an owner announces an AI rollout to the shop floor, the room goes still. The senior tech who has run a service truck for twelve years crosses his arms. The Comfort Advisor who closed $1.8M last year — paid almost entirely on commission — does the math on what a Rilla scorecard means for his pipeline. The CSR who has answered the phones for nine years stares at the floor. Somebody says it out loud: "So AI is going to replace me." That moment is where most trades AI rollouts die. Not at procurement. Not at integration. At the all-hands. Shops that survive it and lift booking %, MPR, close rate, and dispatch yield over the next 12 months walk into the meeting with the conversation, the comp plan, the pilot, and the public win already written. Shops that wing it lose their best people inside 90 days — to the shop down the road still paying old-comp money and drowning in after-call work. This lesson is the conversation, the comp plan, the pilot, the public win, the Comfort Advisor variant, and the seven mistakes that get owners fired by their own staff before AI ever books a call.

The Conversation — What You Actually Say at the All-Hands

The conversation is not a slide deck. It is fifteen minutes, standing in the bay between the parts room and the dispatch board, with every CSR, dispatcher, tech, advisor, and apprentice in the room. Tuesday morning 7:00 a.m. before the trucks roll — Friday afternoon announcements give the floor seventy-two hours of group-text speculation that hardens positions you cannot un-harden by Monday. In person, not Slack, not email, not a recorded video. The question being asked is rarely "what does AI do." It is "do I still have a job in 90 days, and if I do, am I still paid what I was paid yesterday."

Structure of the fifteen minutes is fixed. Minute 1-2: name the tool and the problem. "We are piloting Avoca on the phones in 30 days because we are missing 22% of our calls — roughly $87K of margin walking out the door annually. Fix it or marketing spend keeps inflating cost per conversion and we run out of room for raises." Name a real tool — Avoca, Rilla, Dispatch Pro, ResponsiBid, Hatch — and a real number. Do not say "AI initiative." Nobody on the floor has ever fixed a problem called an initiative. Minute 3-5: name the fear directly. "Some of you wonder if this means replacing CSR seats, the dispatcher, the ride-along. I want to answer head-on before anybody walks out and texts their cousin." Then answer it.

Minute 5-10: lay out the role evolution in concrete terms. For CSR: "AI answers routine inbound and after-hours. You handle the 20-40% AI cannot close — price-shoppers, angry recalls, elderly homeowners who need slower pacing, complex routing. Volume goes down. Booking % on calls you handle goes up. Four-minute-ACW on every call stops." For techs and Comfort Advisors: "Rilla is virtual ride-along. Not to replace you — to give you coaching you have asked me for, for years, that I cannot give because I cannot ride along on 30-40 calls a day. Close rate up; commission up; scorecard shared, not punitive." For dispatchers: "Dispatch Pro is the suggestion layer. You still own the override. The 10-minute re-evaluation does the math you have been doing in your head — surfaces override choices instead of you tracking 19 calls and 7 techs by memory." Minute 10-13: lay out the comp plan changes you have already designed. Announce the tool and the comp tweak in the same meeting — the comp tweak is the only thing that makes the role evolution credible. Minute 13-15: name the timeline, the pilot scope, the success criteria, and the rollback path. "We pilot Avoca for 30 days on after-hours only. Booking % moves from 65% to 75%, missed-call % drops from 22% to under 10%. Hit it, expand to in-hours overflow. Miss it, kill the pilot and go back to the answering service. Dashboard read daily; you see the same numbers I see at the 4 p.m. huddle."

You do not take questions in the all-hands. You take them in 1:1s scheduled for that afternoon and the next morning. The loudest voice in the all-hands sets the floor's position before the quieter voices think. The 1:1 cadence — fifteen minutes per person, owner doing every one — is where the real conversation happens. Ten to fifteen hours of 1:1s for a 20-person shop. The most leveraged ten hours of an AI rollout.

The Comp Plan — The Only Thing That Makes the Role Evolution Credible

Announce a tool without announcing the comp plan, and the floor reads it as "I will be paid less for handling fewer calls." That reading is usually correct under old-comp rules and almost always wrong under new-comp rules — but the floor does not know that until you say it out loud. The comp plan is the only artifact in an AI rollout the floor reads literally and trusts.

CSR comp evolution. Pre-AI: $18-$26/hour with a small booking-% bonus that nobody hits because the floor is drowning. Post-AI: higher base ($22-$30/hour for the higher-judgment surface) plus booking-% bonus on calls personally handled (target 80-85%, $200-$500/month at target) plus show-rate bonus ($150-$300/month at 92%+) plus revenue-per-booked-call bonus above floor median ($100-$300/month). Total ceiling rises 8-15%; base rises 5-12%; variable rewards the calls AI cannot close. Show the math at the all-hands — "old top earner made $52K; new top earner makes $58K-$62K because bonuses move on metrics you control" — and you convert the skeptical CSR into the loudest in-room advocate in 90 seconds.

Comfort Advisor / Sales Advisor comp evolution. The hardest conversation. Comfort Advisors paid 80-95% on commission; Rilla scorecards trigger the deepest "AI is grading me" reaction. Commission percentage and ticket structure do not change. What changes: a Rilla coaching bonus — $300-$800/month at scorecard targets across intro, system condition, repair-vs-replace pivot, options presentation, financing pivot — plus a 30/60/90 close-rate ramp against documented benchmarks (45-55% close by month 3 on aged-equipment service, 28-40% financing close on $5K+ by month 6). Rilla scorecard explicitly coaching, not performance management, for the first 90 days. The 4-Rs commitment: Recorded, Reviewed, Returned within 24 hours, Rewarded quarterly. Service manager runs 30-40 virtual ride-alongs/day (vs. 2-3 in-person pre-AI). Coaching note flags three highest-impact moments per call. Advisors hitting scorecard targets earn close-rate-lift bonus, top-quartile training trip eligibility, senior advisor track. Run cleanly, the 4-Rs converts Rilla from threat into most-requested coaching tool in 60 days.

Dispatcher comp evolution. Quietest pushback because the role transforms most without comp loss. Pre-AI dispatchers $55K-$78K salaried. Post-AI: dispatch-yield bonus ($300-$800/month at top-quartile RPT) plus override-quality bonus (when dispatcher overrides outperform AI's recommendation in revenue terms). Role becomes exception handling and yield optimization; comp recognizes the strategic shift. Tech comp evolution. Smallest adjustment because billable hours, average ticket, MPR contribution math stays. Addition: Rilla-coaching-engagement bonus ($100-$300/month for techs consuming their weekly coaching note and showing scorecard movement) plus MPR-attach bonus aligned to the AI-coached pitch (target 22% to 35-50%). Bonus rewards engagement with the coaching, not adoption of the tool.

The Pilot Design — 30 Days, Narrow, Measured, Reversible

The pilot is the proof and the off-ramp. An owner who designs the pilot with a real off-ramp — written into the all-hands, posted on the wall, named in the comp doc — earns the floor's trust that AI is being run as a tool, not a religion. Shops lose pilots not because AI underperformed but because the owner could not credibly walk it back, and the floor read the inability to walk back as a sign that pushback would be punished.

Pilot scope. Narrow. Avoca on after-hours only for 30 days, not in-hours overflow and after-hours and warm-transfer all at once. Rilla on three Comfort Advisors for 60 days, not the full advisor row. Dispatch Pro in shadow mode for 30 days — surfacing suggestions to the dispatcher without auto-routing — before any production rollout. Hatch on one stale-lead segment (the 6-month-dormant GLSA pile) for 45 days. ResponsiBid on one product category (heat-pump replacement) for 60 days. Narrow scope produces three benefits: the team learns the tool deeply on contained surface, the owner reads a clean metric signal, and the off-ramp is genuinely available because killing the pilot does not require unwinding a shop-wide change.

Pilot success criteria. Written before the pilot starts, signed by owner and service manager, posted on the dispatch-room wall. Avoca: booking % from 65% to 75%, missed-call % from 22% to under 10%, after-hours capture from 12% to 60%+, zero customer complaints about misrouted bookings. Rilla: 30-40 virtual ride-alongs/manager/day, coaching notes returned within 24 hours, three Comfort Advisors showing 4+ point close-rate lift over 60 days, advisor NPS on the scorecard at 6+/10. Dispatch Pro shadow: 70%+ of AI suggestions match what the dispatcher would have chosen, no high-revenue calls dispatched to wrong tech, override rate trends down as dispatcher trust builds. Each pilot has 3-5 named numbers and one qualitative gate (no customer harm, no team revolt). Read at daily 4 p.m. huddle and weekly Friday recap.

Pilot off-ramp. Written and named at the all-hands. "Day 30, Avoca hits booking % 70%+ and missed-call under 12%, we expand. 65-70%, extend 30 days with named fixes. Under 65%, kill it and go back to the answering service while we evaluate Jobber AI Receptionist or HCP AI Agents as alternatives." The off-ramp signals AI is being evaluated, not adopted; commits the owner to honest numbers; removes the dynamic where pushback gets read as disloyalty. Shops that pre-commit run cleaner pilots. The off-ramp is the credibility layer.

Pilot governance. One pilot owner per tool. Service manager owns Avoca. Sales manager owns Rilla. Dispatcher owns Dispatch Pro shadow. Marketing manager owns Hatch and ResponsiBid. The owner owns the review cadence — daily 4 p.m. huddle, weekly Friday 3 p.m. recap, end-of-pilot decision meeting on the pre-committed date. Governance prevents the rollout from drifting into "we have AI now" abstraction and keeps it as "we are testing this specific tool against these specific numbers with this specific off-ramp."

The Public Win — Converting Skeptics into Advocates in 60 Days

The pilot is the math. The public win is the story the floor tells in the break room when nobody from management is in the room. Pilots producing metric movement without public wins still lose people — the metric is abstract, the story is not. The owner who designs the public win in advance — one or two specific moments inside the first 60 days the floor will tell as a story for the next two years — converts the most stubborn skeptic into the loudest internal advocate, and the conversion is durable.

The Avoca public win shows up in week three or four. The senior CSR — the nine-year veteran who crossed her arms at the all-hands — comes to the 4 p.m. huddle and says, "Avoca booked a $14,000 furnace replacement at 11:42 p.m. Saturday. The lady was crying — her mother was in the house and the heat had been out six hours. Avoca got the booking, set the slot for 7 a.m. Sunday, sent the on-my-way text with my photo. The customer told me when I called to confirm that 'that was the kindest voice I have ever talked to in an emergency.' I would not have answered. I was asleep." The senior CSR is now an advocate. The story circulates the bay for two months. The public win is the story the advocate tells about the metric.

The Rilla public win shows up in week four or five. A Comfort Advisor with a 36% close rate gets a Rilla note flagging that on the kitchen-table close at the Marin Park job — the one that died at "let me think about it" — he had not asked the spouse a single question in 47 minutes. He absorbs the note, runs the next two estimates with explicit spouse-questioning structure, closes both. At the next sales meeting he says, "I did not know I was doing that. Rilla pointed at the moment in 12 seconds. I want this on every call." The Comfort Advisor is now the loudest internal advocate. Two more advisors convert within three weeks.

The owner engineers the public win, does not wait for it. Pick pilot timing strategically — the after-hours emergency in week three is statistically inevitable at a 7-truck residential HVAC shop in January; launch the pilot in January for that reason. Pick pilot participants strategically — the Rilla pilot Comfort Advisor is the highest-curiosity, second-highest close rate, not the lowest performer who would resist coaching. Make the win visible: 4 p.m. huddle, Monday standup, Friday recap, Saturday WhatsApp group, all within 24 hours. Celebrate with the advocate's name attached — "Maria's $14K save Saturday." After two public wins in 60 days, the floor shifts from skeptical to actively requesting the next tool.

The Comfort Advisor Variant — Rilla and the Commission Truth

Every AI rollout has one role with the deepest pushback dynamic. In residential trades it is the Comfort Advisor on a Rilla pilot. Structural: Comfort Advisors paid 80-95% commission, tied to close rate, average ticket, financing attach. A Rilla scorecard reads as a graded performance review on every kitchen-table close — and they have spent careers learning that performance reviews precede PIPs precede terminations. Mishandle this and you lose your best $2M-producing advisor to the shop down the road inside 45 days.

The Comfort Advisor conversation has three honest pieces. One: name the commission truth. "Your commission structure is not changing. Top three earners last year cleared $185K, $172K, $158K. No adjustments to commission percentages, ticket bands, financing-attach payouts. Rilla is not a comp lever; it is a coaching lever." Say it out loud, in front of every advisor, in writing, signed. The signed document matters because Comfort Advisor pushback is grounded in a deep mistrust often earned by previous owners. Two: name the coaching value in their language. "You have asked me for more coaching. Honest answer has been I cannot ride along on 30-40 calls a day. Rilla does 30-40 virtual ride-alongs per service manager per day. Coaching feedback inside 24 hours on every kitchen-table close, flagging three highest-impact moments — intro, system condition, repair-vs-replace pivot, options presentation, financing pivot. Rilla's 18% close-rate lift across home-services deployments translates to ~$2,500-$4,500/month in commission lift per advisor at typical residential ticket bands." Three: name the 4-Rs commitment and the 90-day coaching-only window. "First 90 days, Rilla scorecard is coaching only. No performance management. No PIPs. No comp impact. The 4-Rs: every ride Recorded, Reviewed, Returned within 24 hours, Rewarded quarterly for scorecard movement. After 90 days we add a coaching-engagement bonus and an MPR-attach bonus. No downside added. If trust breaks in the 90 days, we kill the pilot and have the conversation about why."

The signed coaching-only commitment is the credibility moat. Rollouts that fail do so because the owner started using the scorecard for performance management inside 30 days, the floor read the change, and the best advisor left. Pre-commit to coaching-only, hold the line through quarter one, and Comfort Advisors become Rilla advocates by quarter two. The discipline is in the pre-commitment, not the post-hoc messaging.

The Seven Mistakes That Kill Trades AI Rollouts (and the Owner)

The pushback playbook works when it is run cleanly. The mistakes that kill rollouts in 2026 are documented, repeatable, and avoidable. Read this list once a quarter for the first two years and check yourself before every all-hands.

One: announcing the tool without announcing the comp plan in the same meeting. The floor reads silence on comp as a threat. Comp plan goes in the deck, on the wall, into the 1:1s. Two: skipping the 1:1s after the all-hands. All-hands names the change; 1:1s answer the personal question. Skip them and the 1:1 happens in the break room with the loudest skeptic running it. Three: no off-ramp. If the pilot cannot credibly be killed, it is not a pilot — it is an adoption. The floor reads the difference. Four: using Rilla scorecards for performance management inside 90 days. Comfort Advisor variant collapses immediately. Five: under-investing in the daily 10-minute huddle and the weekly 30-minute deep review. ServiceTitan's 2026 44% training-barrier number is the gap the huddle closes. Skip it and the rollout regresses to baseline inside 60 days. Six: not engineering the public win. Pilots that produce metrics without storytelling-grade moments do not shift the floor. Pick timing, participants, celebration mechanic. Seven: announcing AI without explaining what it is not. "AI does four mechanical things on a call: transcribe, classify, generate, execute. It does not decide whether the homeowner can afford a $14K replacement. That is still your call." Naming limits builds more trust than enumerating capabilities.

The owner who runs the checklist before every rollout gate (procurement, pilot launch, 30-day review, 60-day review, expansion decision) reduces failure rate by an order of magnitude. Shops that lose the rollout almost always lose it on mistake one or four; shops that hit top-quartile lift run three, five, six with discipline. Change management is the moat; the AI tool is fungible.

The 90-Day Rollout Cadence and the Owner's Weekly Discipline

The playbook is not a one-meeting event. It is a 90-day cadence the owner runs personally because the floor reads the owner's engagement as the signal of how serious the rollout is. Delegate to the service manager and the floor reads it as not-quite-real; run it yourself and the floor reads it as the operating priority for the quarter.

Days 1-7 (pre-launch). All-hands Tuesday 7 a.m. 1:1s Tuesday afternoon through Friday morning. Comp plan posted in the dispatch room. Off-ramp posted on the pilot-tracker wall. Daily 4 p.m. huddle on calendars. Days 8-14 (pilot launch). Tool goes live on the narrow scope. Daily 4 p.m. huddle runs without fail — missing the first week's huddles is the most common cause of pilot drift. Service manager publishes Friday 3 p.m. recap. Owner reads and reacts publicly. Days 15-30 (mid-pilot). First public win usually emerges day 18-25. Owner amplifies — names the advocate, tells the story at Monday standup, posts in WhatsApp. Second 1:1 cycle with the hardest pushback. Days 31-45 (pilot review). End-of-pilot meeting against pre-committed criteria. Public decision: expand, extend with named fixes, or kill. Owner writes the reasoning in one paragraph posted in the dispatch room.

Days 46-90 (expansion or recovery). If pilot expands, the next scope follows the same cadence (in-hours Avoca overflow, full advisor row Rilla, production Dispatch Pro). If killed, the recovery cadence is honest acknowledgment of what did not work, public conversation about the next tool to evaluate (Jobber AI Receptionist vs. HCP AI Agents vs. ServiceTitan Voice if Avoca died), and a 30-day pause before the next pilot launches. The 30-day pause signals decisions are deliberate, not reactive.

The owner's weekly discipline during all 90 days is fixed. Monday 7 a.m. standup with the rollout decision queued from Friday recap. Tuesday-Thursday daily 4 p.m. huddle. Friday 3 p.m. recap reading. Saturday morning text to the pilot owner with one specific observation. Six touchpoints per week, ~20 hours of owner time across 13 weeks. Shops that win top-quartile lift invest these 20 hours; shops that delegate them lose pilots and people inside 90 days. Change management is the owner's job — nobody else can do it credibly.

Key Takeaways

• The all-hands is fifteen minutes, in person, Tuesday 7 a.m. Name the tool, name the fear, lay out the role evolution, announce the comp plan in the same meeting, commit to the timeline and the off-ramp. Take questions in 1:1s scheduled for that afternoon, not in the all-hands.
• Announce the tool without announcing the comp plan and the floor reads the silence as a threat. CSR base rises 5-12% plus booking-%/show-rate/RPC variable bonuses. Comfort Advisor commission structure stays; Rilla scorecard adds coaching bonus only for first 90 days. Dispatcher gains yield bonus plus override-quality bonus. Tech gets coaching-engagement bonus plus MPR-attach bonus.
• The pilot is narrow, measured, and reversible. Avoca on after-hours only for 30 days, Rilla on three advisors for 60 days, Dispatch Pro shadow-mode for 30 days, Hatch on one segment for 45 days, ResponsiBid on one product category for 60 days. Each pilot has 3-5 named success metrics and a written off-ramp.
• Pilot governance: one pilot owner per tool, daily 4 p.m. huddle, weekly Friday recap, pre-committed end-of-pilot decision date. The owner does not own pilots; the owner owns the review cadence.
• Engineer the public win in the first 60 days. Pick the pilot timing and the pilot participants so the win is statistically likely. Amplify the advocate's story by name in the 4 p.m. huddle, Monday standup, Friday recap, WhatsApp group. Two public wins shift the floor from skeptical to advocating.
• The Comfort Advisor variant: signed coaching-only commitment on Rilla for the first 90 days, no PIPs, no comp downside. The 4-Rs commitment — Recorded, Reviewed, Returned within 24 hours, Rewarded quarterly. Break the commitment and lose the top of the advisor row inside 60 days.
• The seven mistakes that kill rollouts: silence on comp, skipped 1:1s, no off-ramp, Rilla used for performance management inside 90 days, under-invested huddle cadence, no engineered public win, no naming of what AI is not. Read the list once a quarter for the first two years.
• The owner runs the cadence personally for 90 days. ~20 hours of owner time across 13 weeks. Six touchpoints per week. Delegate the cadence and the floor reads the delegation as priority signal. Change management is the moat; the AI tool is fungible.