2026 Concierge Playbook: Selling Agent Systems for Brokerages Before Midnight

Everyone online shows you another tool, almost nobody shows you how that tool turns into money. The opening thesis of the host is exactly this: most AI lessons show you the AI but not the business . The result is predictable, shiny demos on screen and zero clients in the pipeline. The model Pink proposes is disarmingly simple: find one expensive, constantly repeating problem inside a company and build a system that fixes it automatically. A broker does not care about artificial intelligence; he cares how fast his leads get answered, how many opportunities get rescued, and what the month-end report proves. This article walks through that model step by step in the property niche: the three agents, the pricing design, and the sales road to a first client.

The first decision in agency work is not what to sell but whom to sell it to, and here the host walks against the crowd. The most fashionable niche is rarely the most profitable one; the field you already know closes fastest. In a sector where the language, the tools and the daily workflow feel familiar, you build systems that slot into existing routines without forcing retraining. The host weighs three candidates: property teams drowning in viewing and follow-up traffic; law offices that could reclaim hours from document review, intake and client updates; financial firms where large tickets support high care fees but sales cycles run long and compliance bites hard. The message is blunt: never enter an unfamiliar sector to show off, bring relief to a pain you understand.

The Niche Call and the Property Bet

The rest of the lesson runs on real estate, for three solid reasons. First, the return case fits in one sentence: slow follow-up equals lost commission. Second, the tools property teams already use connect naturally to the Base44 Super Agent world: Gmail, calendars, tracking systems and shared sheets. This information comes from the base44 source and confirms the platform claim of tool connections, web browsing and background work. Third, carrying one example through the whole system makes it teachable: the same three agents , the same data flow, the same reporting language. Add the universal scene: you message an agent about a home, hear nothing for hours, and drift to whoever replies faster. On the office side that silence means losing a serious buyer before anyone notices the lead arrived.

The first demo agent attacks the painful truth the trade calls speed to lead . Picture a buyer submitting a form at 9:48 in the evening; you spot it at 8:30 next morning, by which time two rival agents have replied and one is already in conversation. The loss comes from slowness, not from poor advice. This information comes from the sierrainteractive source, and its research gives a striking ratio: answering within five minutes instead of thirty makes you roughly 21 times more likely to qualify the lead. Online buyers write to several names and usually stay with whoever responds first, so the opening minutes decide who wins the client. That is why the first ring of the system is a follow-up agent watching fresh inquiries and drafting personal replies within minutes.

Setup starts by opening a fresh Super Agent inside Base44, and the first job is the model lock: both the chat model and the background-task model get pinned to Opus 4.8. This information comes from the anthropic source, which presents the Opus 4 family as the leader for long-running work and agent workflows. Keeping both fields on one model preserves consistency while the agent studies long messages and makes routing calls in the background. Next, Gmail gets connected through plugins so the agent can read incoming property inquiries. The first prompt defines the capture task: pull name, address, location preference, property type and budget into structured fields, then draft a personal first reply. The critical detail is that the agent does more than write: it hands over the extracted facts in usable shape.

Capture alone never suffices; a 48-hour and 96-hour follow-up sequence arrives with the second prompt to win back quiet leads. The rule is elegant: no reply means trying again from a fresh angle, any reply stops the sequence at once and alerts the responsible adviser. Parroting the same message is banned, and continuing to mail someone who already answered is banned too. Testing never gets skipped: a sample inquiry from Sarah Jenkins gets simulated, a three-bedroom hunt on Oakridge Estate with a $650,000 budget. The agent must extract her name, address, location, property type and budget, draft a personal reply mentioning Oakridge, and flag the two follow-up checkpoints. This rehearsal also produces the strongest sales demo imaginable: an inquiry landing at 11 at night, answered personally by 11:01. One delayed reply can cost a commission worth thousands, and no broker needs a technical lecture to feel that.

Triage and Reporting Agents

The second agent sorts the crowd at the door with an intent score matrix. Every incoming inquiry earns between 1 and 10 on three measures: how precisely the person describes the property, how financially ready they look, and how near their moving date sits. Someone naming an exact address with pre-approval ready and a move next month outscores a casual market question by miles. The bands stay crisp: 8 to 10 means high priority, 5 to 7 medium, 1 to 4 low. Setup repeats the model lock, then Gmail plus a sheet plugin get connected and authorization completes. The sheet becomes the waiting room where low-priority leads rest. The value of sharp rules shows here: the agent cannot produce vague or wobbly judgments, it measures every inquiry with one ruler.

Scoring is half the workflow; the other half is the routing decision that follows the number. A high-priority lead fires an instant alert carrying contact facts, the assigned score and a short rationale straight to the owner. Low-priority names slide quietly into the log while nobody wakes up to a needless midnight ping. The test runs two extreme profiles and watches the agent separate the serious from the strolling. This information comes from the followupace source, confirming that smart sequences branch on lead behavior rather than fixed calendars; it recommends capping early contact at two or three touches a week, writing through the lead's preferred channel, and screening every template for housing-rule compliance before launch. For the office the picture is clean: serious buyers reach a human at once, everyone else enters orderly nurture.

The third agent slices the pie nobody watches: the weekly performance report. Brokers rarely know how their lead flow truly performs; they guess, sense, and argue about it in meetings. This agent gathers the numbers into one summary, pins it to a fixed calendar, and hands the owner a clear view of system health. Setup repeats the familiar lock, then Gmail and sheet links go in; sheets serve both as the tracking source and the archive of finished reports. The first prompt commissions the agent as an operations analyst and freezes the skeleton: total inbound count, high-priority share, average speed-to-lead time , and inquiry-to-viewing conversion. Those four figures melt into a tight executive narrative. Then a Sunday-evening scheduled run gets defined, so the report lands weekly in identical shape.

The test asks the agent to compress a week of data into an executive summary , and the result becomes the strongest document defending the monthly fee. Each month the client sees on a single page what the money bought: how many leads came, how many were serious, how much faster replies grew, how many viewings got booked. The platform risk behind that choice deserves outside confirmation, which research supplies. This information comes from the techcrunch source, confirming the mid-2025 cash purchase of the Base44 startup by Wix for $80,000,000; coverage notes 250,000 users within six months, 10,000 inside the first three weeks, and a $189,000 May profit despite heavy language-model bills. The record also keeps the sober footnote: behind the solo-founder legend stood an eight-person team with a $25,000,000 retention package.

The Pricing Design

With three agents ready the question shifts: how should this be charged? The host answers with a two-layer setup fee plus monthly care structure. Hourly billing collapses on this work because speed turns into punishment: the faster and sharper you get, the less you earn. This information comes from the growthlogics source, which narrates the hourly-model breakdown vividly: once an agent finishes in twenty minutes what a junior did in two days, the hourly invoice shrinks by roughly ninety percent. The setup payment covers everything that makes the agents client-ready: learning the current workflow, adapting prompts, wiring Gmail and sheet or tracking systems, and end-to-end testing before launch. Monthly care covers everything after deployment: connections expire, models change, teams adjust how they handle leads. Monitoring, prompt refreshes, performance checks, connection repairs, template updates and small additions like fresh lead tags all bill to this line.

Numbers for first clients stay grounded: $1,000 to $3,000 for setup. That band keeps the offer reachable while the portfolio grows, and buys the chance to collect a powerful case study in return. Once real performance data, client words and working proof accumulate, setup can climb toward $3,500 to $5,000 and beyond. For property teams monthly care usually rests between $500 and $1,500 per office or team, moving with agent count, monitoring load and connected systems. The first version of the offer stays deliberately narrow: promising ten automations to sound impressive invites delays, scope creep and more parts that can break. Two or three crisp agents keep delivery manageable and measurement easy.

Once the package stands, it must meet real owners, and the good news is that early clients need no large ad budget or tangled funnel. While the offer is still being tested, direct outreach beats paid promotion; advertising gets expensive fast before anyone knows what sells. LinkedIn, plain mail and targeted cold contact reach managing brokers and owners with little upfront spend. Feedback also arrives faster: ignored messages trigger offer tweaks, and when several prospects name one problem the agency leans into it. The message itself must stay specific; lines that fit any company persuade nobody and prove no property knowledge. A strong note flows in three moves: name a problem they recognize, describe the built system in one line, tie it to a measurable result. The sample arc runs like this: warm leads lost because follow-up passes thirty minutes on busy viewing days; an automatic arrangement answering within 90 seconds and pre-screening budgets; steady speed to lead plus field teams freed from 10-plus hours of weekly admin.

Cold contact opens its door with a free fifteen-minute operations audit plus a short screen-video demo. The sales call starts not with slides but with a discovery call built from questions: what happens minute by minute when a lead arrives from a portal or the site at 8 in the evening, how many hours weekly the team burns sorting unqualified inquiries, and how management currently tracks follow-up performance across the desk? Such questions make the talk specific to that brokerage and let delays, missed handoffs and repetitive chores surface in the client's own words. Once problems arrive through their answers, the offered system explains itself with ease. Then the three agents go on the table as one starter pack: follow-up agent, triage agent, reporting agent. Handing over a long automation menu for the prospect to choose from means endless meetings and foggy requests; one system, one legible result.

Sales and Scale

After agreement, scope gets documented before any client-specific work begins, and a 48-hour standard onboarding starts. The build may mirror the demo, yet prompts, accounts, routing rules and brand voice now belong to that client's real operation. The welcome list runs mechanically: company name, service area, office hours, brand tone and operating rules get confirmed; chat and background models stay verified on the Opus 4.8 lock. The final gate is a pre-launch validation checklist: several simulated leads of varied intent go in, while score logic, log formatting, correct account delivery and sub-120-second mail arrival each get tested. That discipline lifts delivery off the founder's personal hours and turns it into a repeatable arrangement.

Landing the first client excites, but the larger prize opens after the system runs long enough to yield real numbers. No more demos get sold and no trust in a fresh idea gets requested; proof from inside a working brokerage does the talking. The system runs about 30 days and results compress into a short case study : lead reply time fell from 4 hours to 85 seconds, the team reclaimed 18 hours of weekly manual sorting, and 3 extra viewings got booked in month one from leads that might have died silent. Those figures beat long capability essays by miles. Cold outreach eases too, because verified outcomes replace untested promises. Clients two and three in the same niche close faster: workflows are known, objections sound familiar, broker vocabulary is decoded. Setup fees walk from the $1,500 band toward $3,500, sales cycles shrink from weeks to days, and the three agents harden into reusable assets.

The host closes by shrinking the picture to three decisions feeding a 90-day execution plan . Decision one is single-niche focus: it sets how credible outreach sounds, how fast client workflows get understood, and how much past work gets reused. Specialization strengthens case studies and improves delivery margins over time, while the generalist writing fresh offers per prospect stays trapped in price competition. Decision two is the charging shape: upfront setup plus monthly care, full stop. Hourly billing rewards longer labor over better outcomes and dresses the service as basic tech support, while value-shaped pricing leaves room to build efficiently. Decision three is documentation: hiring a technical operator, delegating upkeep, writing standard procedures for environment cloning and validation. Together the three turn the agency from a heap of custom projects into an arrangement that grows without depending entirely on founder hours.