Record in-home and in-office sales conversations so one manager can coach 25 appointments a week instead of six.
Every business that closes in person has the same structural problem: the conversation that determines revenue happens where the manager is not. Call-based teams solved this a decade ago with conversation intelligence. In-person teams — home services, remodeling, dental case presentation, senior living, multifamily leasing — never did, because the only coaching tool was physically riding along, which caps a manager at six or seven appointments a week and biases toward whoever is geographically convenient. Recording the in-person conversation on the rep's phone and transcribing it (Rilla is the category-defining product here) removes the travel constraint entirely. The manager reviews 25-30 appointments a week, coaching is based on what was actually said rather than on the rep's recollection, and the gap between top and bottom performers becomes visible and specific.
This is step one and it is not negotiable. Recording law varies by state: one-party consent states require only the rep's consent; two-party (all-party) consent states require the customer's. If you operate in California, Florida, Pennsylvania, Illinois, Washington, Massachusetts, or any other all-party state, you need a disclosure and an affirmative customer agreement at the start of every conversation. Get this reviewed by counsel, write the script, and train it before the first recording.
Write your actual sales process as observable behaviors: did the rep confirm the problem before quoting, present three options, address the objection rather than discount, ask for the close. If you cannot write it down, you cannot coach against it and the AI has no rubric to score.
Record your two best reps for two weeks before rolling out to anyone else. This gives you the scoring standard, and — more usefully — it usually reveals that your top performer does something nobody documented, which becomes the highest-value coaching point for everyone else.
How you introduce this determines adoption. Reps hear "we are recording you" as surveillance unless you lead with the trade: they get specific coaching instead of vague feedback, and the manager sees the appointments where they did everything right and still lost. Show the top-performer findings from step 3 first — it reframes the tool as a way to learn rather than a way to be caught.
Block the time. The single reason this workflow fails is that managers do not do the reviews. A realistic cadence is 30-45 minutes daily, covering 5-6 appointments, with one written coaching note per rep per week. Capacity without cadence produces nothing.
The transcript will surface eight things a rep could do better. Pick one. Coach it for two weeks. Measure it. Then pick the next. Multi-point feedback from a recording is overwhelming in a way that in-person feedback never was, because there is written evidence for every point.
The two numbers that move. Rilla publishes aggregate improvement claims that are vendor-reported and should be treated as such — measure your own baseline before rollout so you can evaluate honestly rather than accept a case study.
Use these templates as-is or customize for your business.
"Before we get started — I record my appointments so my manager can review them and help me do a better job for customers like you. It stays internal to our company. Are you comfortable with that?" IF CUSTOMER DECLINES: "No problem at all." — Stop the recording immediately and note the decline in the CRM. Do not argue, do not re-ask, and do not record anyway. One recorded objection is a lawsuit; one declined recording is a Tuesday. TRAINING NOTE: This must be delivered naturally at the start, not mumbled. A rep who sounds uncomfortable asking makes the customer uncomfortable agreeing.
Score each appointment 0/1 on behaviors, not impressions: [ ] Confirmed the customer's stated problem in their own words before presenting [ ] Walked the job / examined the issue before quoting [ ] Presented three options (good / better / best), not one price [ ] Anchored on value before stating price [ ] Addressed the first objection without discounting [ ] Asked for the business explicitly (not "let me know") [ ] Set a specific next step with a date if not closed [ ] Mentioned financing / membership where applicable Never score more than 8 items. A long rubric produces no coaching.
REP: [name] WEEK OF: [date] APPOINTMENTS REVIEWED: [n] WHAT WENT WELL (be specific, quote the transcript): - At the Henderson appointment you said "..." — that's exactly the right move because... ONE THING TO WORK ON THIS WEEK: - Behavior: [single item from the rubric] - What I heard: [quote] - What to try instead: [specific alternative wording] - How we'll measure it: [n appointments reviewed next week for this behavior] NOTHING ELSE. One item. Next week we pick the next one.
Get a new AI workflow every week. Prompts, tool stacks, and ROI math included.
Single agent with function-calling: one LLM with a defined toolbox (CRM, calendar, knowledge base) decides which tool to invoke at each turn. Easiest to debug; appropriate for most well-scoped business workflows.
Learn the agentic glossary →Where this workflow tends to break in production — and what to put in place before you ship it.
Recording without valid consent in an all-party state
Mitigation: Scripted disclosure trained and role-played before rollout; immediate stop on decline; CRM field logging consent per appointment.
Reps perceive the tool as surveillance and disengage
Mitigation: Lead the rollout with top-performer findings; coach one behavior at a time; never use recordings in a disciplinary action you did not disclose up front.
Managers stop doing reviews after week three
Mitigation: Calendared review block with a weekly written coaching note as the visible deliverable; track reviews-completed as a manager metric.
Rubric drifts into subjective impressions
Mitigation: Keep the rubric to 8 binary observable behaviors; re-baseline against top performers each quarter.
Skip this entirely if you cannot or will not implement a compliant consent process — recording a customer without consent in an all-party state is a criminal statute in several jurisdictions, not a policy risk. Skip it if your sales are transactional and short; there is nothing to coach in a four-minute interaction. Skip it if your managers are already at capacity and nobody will run the review cadence, because the tool produces recordings, not improvement. And skip it if your close rate problem is actually a lead quality problem — coaching cannot fix unqualified appointments.
A phased approach to get this workflow running and delivering ROI.
Days 1–30
Foundation
Days 31–60
Optimization
Days 61–90
Scale
Four tools, one genuinely important design difference, and a free tier good enough that most small teams should not be paying at all.
The general-purpose voice agent and the trades-specific one cost about the same. One of them books more jobs. The difference is not model quality — it is everything around the model.
Four vendors, four incompatible pricing models, and one arithmetic trap that doubles your bill without anyone telling you.
One practical AI workflow per week. No fluff.
Get the full guide with step-by-step setup, workflow templates, and copy-paste assets.