This is the companion resource to our webinar on how building services companies can use AI to increase revenue. Every prompt from the session is here in full, with fields you can edit so the output comes back written for your company, your market, and your buildings. Start with Assist. Build toward AI Native.
Know where you are on the curve
Most BSCs are Not Started or early Assist, and that is fine. Assist alone can move revenue if the prompts are good. The mistake is waiting for AI Native before doing anything: the Assist prompts below work the week you try them, and they are how you learn which stage of your funnel deserves a real build.
- No regular AI use in the sales process
- Maybe a personal ChatGPT account on the side
- No shared prompts or process
- A person drives the AI for each task
- Draft lists, emails, and bid language in minutes
- Prompting skill is the leverage
- The same prompt runs on a schedule or trigger
- Outputs land in CRM, email, or a call queue
- Still human-reviewed before anything goes out
- The system is designed around AI
- Call lists, drafts, and quotes arrive filled in
- Your team steers goals, not tasks
The funnel you already run
This is how every commercial cleaner already sells. AI does not invent a new funnel, it attaches to each of these stages at three depths: Assist, Automate, and AI Native. For most mid-market BSCs the revenue leaks at Identify and Outreach, waiting on RFPs and referrals. And Win is underrated: bad handoffs quietly cost margin and references.
Find target accounts by building type, size, location, and buying signals.
Research the targets and reach them by cold call, local visit, and email, tracking every touch.
Re-engage known targets at regular intervals with sales and relationship material.
The traditional motion: walk the account, present tailored information, price the proposal, get the signature.
Onboard the account and hand it to the service side of the business without dropping what sales promised.
The examples ship with a Phoenix medical-office operator filled in. Replace them with your market and your target buildings, and the highlighted parts of each prompt update as you type. Your edits stay in this browser, nothing is sent to us.
Identify
Find target accounts by building type, size, location, and signals.
Most salespeople start cold with Google Maps and hope. A ten-minute prompt can do what would take a junior rep a day of research: a ranked list of buildings that match your ideal customer, each with a reason to call.
Specificity is the whole game. “Office buildings in Phoenix” returns noise. Building type plus size plus a real signal returns a call list.
Ask for specific accounts to call by building type, size, geography, and signal, and dictate the output format so you get a table, not an essay.
Permits, RFPs, and ownership changes get pulled into your CRM or inbox every morning, deduplicated against accounts you already know. Reps only work net-new, verified rows.
A ranked daily dialer feed with building facts, the decision maker, the signal, and a suggested angle already on the card. No prompt engineering for the rep.
You are a sales researcher for a commercial janitorial company in Phoenix, AZ. Build a table of 25 target buildings matching this ICP: - Type: medical offices and outpatient clinics - Size: 15,000 to 80,000 rentable sq ft - Signal: a new certificate of occupancy in the last 18 months, OR a property manager change, OR a janitorial RFP in the last 12 months Columns: building | address | est. sq ft | type | signal | decision-maker name | decision-maker title | phone number | 1-line talk track | source URL Prioritize buildings where cleaning is likely still handled in-house.
- Re-run it weekly with “exclude these addresses” so the list stays fresh.
- The talk-track column is the part reps actually use on cold calls. Keep it.
- Verify before you dial. AI-sourced names and phone numbers are a starting point, not a fact.
Outreach
Research, the first touch, and tracking. This is where most pipelines go quiet.
The fix is not wittier copy, it is structure: a sequence where every first touch references a real signal. Generic “we clean buildings” emails die.
Run this right after Identify. Feed it one row from your list and a manager reviews five ready drafts a day instead of five blank pages.
Sequence design across call, email, and visit, personalization angles, and objection-handling scripts, all from one prompt.
Two or three subject lines rotate automatically, and call recordings summarize themselves into CRM fields. One caveat: A/B testing without volume is noise, so set a minimum touch count before declaring a winner.
The system picks who to contact today, drafts the message from live signals, and learns which messages win.
Write a 5-touch outreach sequence for a commercial janitorial BDR. Account: Paradise Valley Medical Plaza | a 42,000 sq ft medical office Signal: new property manager hired 6 weeks ago; the incumbent's contract ends in about 90 days Goal: book a 15-minute walkthrough with the property manager Tone: direct and local. An operator, not a national brand. Deliver: 1) Call opener (20 seconds) + 2 voicemail variants 2) Email #1 (under 90 words) + subject line A/B 3) Email #2 (offer a night QA checklist) 4) LinkedIn note (300 characters max) 5) Breakup email + 3 likely objections with one-line responses
- Always inject a real signal. If you cannot name one, the account belongs back in Nurture.
- Sequence structure matters more than witty copy.
- Record calls (with consent). Transcripts are the bridge to the Automate level.
Nurture
Most deals die in the quiet months. Nurture keeps you first when the contract opens.
An honest note: the Automate story is thinner here than at Identify or Outreach. Do not oversell bots emailing people forever. Thoughtful quarterly touches beat monthly spam, and the craft is the same as outreach at a slower cadence.
Good nurture assets: a night QA one-pager, a “what we find in the first 30 days” checklist, a local hiring and quality story.
Quarterly value emails, local market notes, inspection checklists, and short case snippets, drafted in minutes and reviewed by a person.
A 90-day dormant-account alert, and a reminder to the owner 120 days before every known contract end date.
Personalized microsites and quote pages per prospect, with variants that learn which content converts when volume justifies it.
Draft a re-engagement email for a property manager we walked a building with 4 months ago. Context: they liked our night QA process but stayed with the incumbent until year end; the contract window is now about 5 months out. Include: - Subject line A/B - 2 short paragraphs + a soft CTA (a 15-minute recap call) - A P.S. offering a free exit-inspection checklist they can use even if they don't switch Tone: helpful peer, not desperate vendor. 110 words max.
- Offer something useful even if they never buy. That is what real nurture is.
- Pair the email with a calendar alert 120 days before each known contract end date.
Bid
From site walk to signed scope: speed and consistency without losing judgment.
Your estimators already have templates. AI’s job is filling them from messy walkthrough notes, and a photo of a whiteboard or a notepad is a legitimate input.
Make the AI surface risk instead of hiding it. The clarifying questions at the end of this prompt are the most valuable deliverable.
Paste your standard template and the walkthrough notes or photos, and get a first-pass scope, exclusions, and cover letter to edit rather than write.
Account history, prior pricing, comps, and site data land in a ready draft folder. Less exciting, high value: no hunting for last year's rate card.
Measurements go into a labor model and come out as a branded proposal in a few taps, while the walkthrough is still fresh.
You are an estimator for a commercial janitorial company. Using the walkthrough notes below and our standard scope structure, draft: 1) Scope of work by area (lobbies, restrooms, clinical areas, exterior) 2) Frequency table (daily / 3x week / weekly) 3) Exclusions and assumptions 4) A one-page executive summary for the property manager 5) 5 clarifying questions to resolve before final pricing Constraints: medical outpatient, about 42,000 sq ft, 5 days per week, after-hours only, green chemicals in clinical areas. Walkthrough notes: [paste your bullets, or attach photos of your notes]
- Keep rates out of the first draft if pricing is sensitive. Use AI for scope language first.
- Photos of handwritten notes work. Do not retype what you can attach.
Win
The sale is not closed until ops can run the job.
Bad handoffs create stolen hours, failed QA, and sales wins that lose money. This stage is underrated, and it is where AI protects margin and reputation.
The move is simple: turn the signed scope into the artifacts ops needs in week one, using the same language sales sold.
Generate the client onboarding checklist, a room-by-room inspection sheet matching the sold frequencies, and the kickoff agenda.
A signed deal triggers the draft SOW, a staffing plan stub, the supply list, and the Day-1 inspection form.
The won bid becomes the inspection standard, the labor plan, and the client portal baseline automatically. Sales and service run on one system.
Create an ops handoff package from this sold scope of work. Deliverables: 1) Client onboarding checklist (sales → ops → site) with owners and due dates 2) Room-by-room inspection sheet matching the sold frequencies 3) Day-1 / Day-7 / Day-30 quality checkpoints 4) A 30-minute kickoff agenda for the property manager and site supervisor 5) Risk flags if staffing runs 10% short in month one Sold SOW summary: [paste the final scope, frequencies, and specials]
- Reuse the inspection language from the bid so what sales promised is exactly what ops gets measured on.
- This is the cheapest insurance policy on a new account.
Ground rules for good output
Specificity beats volume
A prompt without a building type, a size range, and a real signal returns noise. The bracketed fields in each prompt are the parts that make the output usable.
Dictate the output format
Ask for a table and name the columns, or ask for a numbered list of deliverables. If you do not specify the shape, you get an essay.
Verify before you act
AI-sourced names, phone numbers, and square footage are a starting point. Check them before you dial, and keep pricing out of drafts until a person has reviewed the scope.
Keep a person in the loop
At the Assist and Automate levels, every output gets reviewed before it reaches a prospect. The AI drafts, your team decides.
Build a shared prompt library
A prompt that works belongs in a shared doc, not one rep's chat history. That is how prompting skill compounds across the team.
Wondering which stage to automate first? The self-assessment scores your operation, and how we work explains what a discovery engagement covers.
This week: run the Identify prompt for your top ICP. When you are ready to weld the funnel together, that is what Colby builds.