What happens when you stop doing it manually.

Custom platforms, AI agents that take real bookings, and systems that answer the phone. Client builds with real numbers, plus the exact plan we would ship for your industry.

Real builds. Real systems. Real numbers.

Seven delivered client engagements: custom platforms, AI agents that take real bookings, and phone systems that answer themselves. Three of the builds below were created for our partner, Sandbox VR Alpharetta. All custom code against real APIs, not templates or workflow tools wired together. Every number below is real, and where a figure is an estimate of displaced manual work rather than a measured result, it says so. The sample build plan at the end is exactly that: a plan, clearly labeled, not a delivered project.

Sandbox VR Aplos AI
Partner Build · VR Entertainment Venue

One connected system for the entire guest journey

Client Sandbox VR Alpharetta, partner
Tools Cloudflare Workers · Workers AI · KV · Pages Functions · Twilio · Apps Script
Type AI guest service + event sales + revenue operations

Questions arrived through the website, phone, text and inbox. Bookings and payments lived in separate systems. Event leads needed fast, persistent follow-up, while the venue's operating picture was scattered across twelve data sources. Every handoff was another chance to lose context or lose the guest.

A connected production platform spanning the AI website agent, AI phone attendant, automated email and SMS follow-up, a custom customer and event-sales CRM, an event-lead power dialer, and a role-gated dashboard for revenue, bookings, calls, reviews, marketing performance, capacity and forecasts.

The same customer can now be recognized across bookings, inquiries, calls, texts and website conversations. Staff can see the full history, work prioritized event leads, compare quoted and paid revenue, reconcile Square with Checkfront, and ask plain-English questions of current operating data.

12

Live sources in one payload

payments, bookings, 3 ad platforms, calls, reviews, analytics, weather

Ask it

Answers from live data, not training data

plus an ops brief rewritten on every refresh

One guest

One connected history

from first question through booking, visit and follow-up

Read the full case study →

Sandbox VR
Partner Build · VR Entertainment Venue

An AI agent that answers, quotes and actually books

Client Sandbox VR Alpharetta, partner
Tools Cloudflare Workers · Durable Objects · Checkfront API · Twilio
Type Guest-facing AI agent

A venue sells sessions on evenings and weekends, which is exactly when nobody is free to answer the website, the phone or the inbox. The honest fallback for any chat widget is “give us a call”, which on a venue already missing calls is not a handoff but a hope. Event enquiries, the highest-value booking type, died at a five-field form.

An agent that reads live availability at the moment of asking, takes a real booking, puts a group on a path where everyone pays their own share, runs event intake as a conversation rather than a form, works the store's text line, and knows which booking a guest is standing on. It texts a human when a conversation is worth one, with restraint designed against alert fatigue rather than against missed alerts.

The model never performs the money action. It resolves what the guest wants; deterministic code writes the booking, takes the payment and spends the voucher code. Every answer is filtered against the venue's own facts before it leaves, and a drift checker reports anywhere else in the system that contradicts them.

It books

Not a lead form that types back

live availability, real bookings, group payment, vouchers

Out of hours

Answered, not queued

the busiest hours are the ones nobody can answer

0

Money actions taken by the model

it decides intent; code performs every write

Read the full case study →

Sandbox VR
Partner Build · VR Entertainment Venue

They built their own prospect list, then worked it

Client Sandbox VR Alpharetta, partner
Tools Python · Twilio Serverless · Apps Script · Google Ads · Meta · TikTok APIs
Type Prospecting + outbound + ad ops

Corporate events, school trips and group bookings do not walk in, and the corporate lane was a few dozen companies somebody had thought of, with no way to know which had a decision-maker anyone could reach. Call outcomes lived in whoever made the call. Outreach ran from the venue's real mailbox, where one run of bounces on a cold domain costs not a campaign but the channel.

A list builder that assembles local companies by real driving distance and refuses any row without a reachable decision-maker, with headcount taken free from the SBA's public PPP release rather than bought, chamber directories ingested whole, and every address deliverability-checked before a send. Then seven role-assigned outreach lanes, a power dialer with continuous dialing and a safety buffer, and a contact ledger the phone system writes itself.

A dialed lead that goes to voicemail gets a text automatically, once ever, and it takes two independent signals: a machine verdict from answering-machine detection, plus a call that ran under a minute. The duration is a veto and never a reason to send, because a screened handset put a machine on the line for a call the owner answered in person.

Built, not bought

Their own prospect database

by real radius, gated on a reachable decision-maker

~7.5 hrs

Recovered every week

dialing, logging, research and ad verification

1 per lead

Automatic voicemail texts, ever

two independent signals required to send one

Read the full case study →

The Cube Signs
Client Build · Sign Fabrication

One system from first inquiry to installed sign

Client The Cube Signs · Atlanta, GA
Tools Next.js · Supabase · Resend · WhatsApp Cloud API · Meta Graph API
Type Custom software build + full handoff

The shop ran on a template website, text messages, and memory. A high-value storefront job and a tire-kicker banner inquiry arrived through the same generic contact form and got the same follow-up. Client updates during design, fabrication, permitting, and install were typed by hand whenever the owner remembered. No pipeline view, no lead queue, no dollar total on active work.

One application, live on the shop's own domain: a new marketing site built on 29 photos of their real work, an intake form that scores every lead HOT, WARM, or COLD automatically, a leads board and drag-and-drop project Kanban with a live pipeline dollar total, and automatic client updates on every status change by email and WhatsApp (six Meta-approved templates, including a Google review request at completion). Plus a Facebook and Instagram composer for posting finished signs from one screen. Branding, stages, templates, scoring, and intake questions are all owner-editable in Settings.

The client owns the entire stack: code repository, database, hosting, domain, and every connected account. Fixed project price, no required monthly fees to Aplos, no dependence on us to run it. A real qualified lead arrived, was scored, and was queued automatically within the first two weeks live.

24/7

Leads captured & scored

every inquiry auto-scored HOT / WARM / COLD

0

Hand-typed status updates

moving a job on the board is the update

100%

Client-owned stack

repo, database, hosting, accounts all handed over

Read the full case study →

Client Build · Entertainment Franchise

AI platform for franchise site evaluation

Client National entertainment franchise (50+ US locations)
Tools Python · Census APIs · Custom ML engine
Type Custom AI tool

Each new franchise location costs $500K–$2M to open. The client was evaluating sites through gut feel and broker recommendations: no repeatable data-driven process. Bad site decisions were expensive and hard to defend to stakeholders.

We built a two-module AI platform. Module 1: enter any US street address and get a real-time demographic score against the brand's actual location fingerprint: HIGH / MEDIUM / LOW classification with projected revenue range and confidence score. Module 2: 94 US metros scored and ranked by expansion potential, giving leadership a prioritized pipeline of where to expand next. Runs on public Census data: no paid subscriptions required.

Before: weeks of manual research and broker gut-feel per site, on decisions costing $500K–$2M each, with no way to defend a pick to stakeholders. After: candidate addresses scored through one interface against the brand's real location fingerprint, and a ranked list of all 94 metros for the expansion pipeline.

Scored

Decision support

planning aid; outcome validation not published

94

Markets ranked

prioritized expansion pipeline

On demand

Per site evaluation

address evaluation through one interface

Read the full case study →

Client Build · Architecture Firm

Custom project management system built for the architecture deal lifecycle

Client Mid-size architecture firm
Tools Next.js · TypeScript · Prisma · PostgreSQL
Type Custom software build

The firm was managing a full project pipeline: proposals, active builds, invoicing, expenses, milestones, client documents: across spreadsheets and email threads. Revenue was slipping through the cracks: unbilled milestones sitting unnoticed, overdue invoices going unfollowed, required documents missing at closing. No single place to see the health of every project at once.

A fully custom project management system built specifically for the architecture lifecycle: from deal creation to final close. Dashboard surfaces critical action items the moment you log in: upfront invoices not sent, overdue invoices by dollar amount and days past due, unbilled milestones ready to bill, missing required documents, and open checklist items. Includes kanban board, milestone calendar, invoice generation, expense tracking with markup, and a client-facing portal.

Before: spreadsheets and email threads, where unbilled milestones sat unnoticed and overdue invoices went unfollowed until someone happened to check. After: every dollar at risk surfaces on one dashboard the moment anyone logs in, and the required-document checklist makes a missed form at closing structurally impossible.

Flagged

Billable milestones

every billable milestone flagged automatically

1 screen

Full pipeline view

replaced spreadsheets + email threads

0 forms

Missed at closing

required doc checklist enforced per project

Read the full case study →

Client Build · Auto Repair · Anonymized

The same three questions, answered without picking up

Client Independent auto repair shop, metro Atlanta
Tools Twilio Functions · Twilio Assets · Synthesized voice prompts
Type Phone system build + full handoff

A one-bay shop cannot answer a phone and work at the same time. Most calls were not work: are you open, what time do you close, where exactly are you. Each cost a stop, a wipe of the hands, and whatever was half done. Worse, the business number was the owner's personal mobile, published on the Google listing and every invoice, so it could not be handed to anyone and never stopped ringing.

A real business line with a three-option recorded menu: hours, address, or put me through. The informational branches answer completely and offer the choice again, so a caller who then wants a person is one keypress away rather than redialing. Choosing a person forwards to the owner, who now gets interrupted for a booking rather than for closing time.

The questions were fixed, few, and identical every time, which makes a recorded menu the right tool rather than the cheap one: it cannot mishear, invent an answer, or be talked into a quote, and it runs for a few dollars a month with no subscription. For a one-bay shop, a system that survives a slow month beats a cleverer one that gets cancelled in January.

~3 hrs

Recovered every week

owner estimate, itemized on the case study page

Private

Owner’s own number, back

personal mobile no longer published anywhere

~$4

Monthly running cost

shop’s own account, no fee to Aplos

Read the full case study →

A blueprint, not a case study.

Sample build plan · Dental

If a 4-chair dental practice hired us tomorrow.

Type Sample plan, not a delivered project
Sources ADA, Dental Economics, Dental Intelligence
Quoted price range $8K to $15K depending on scope

We have not shipped this build for a dental practice yet. This page is the proposal we'd hand a practice on day one of an engagement. The architecture mirrors patterns we ship for other service businesses, scoped here for dental. We're publishing it so dental practices can see what we'd actually build before booking a call, and because the first practice that wants this gets a discounted build in exchange for logo and testimonial.

Insurance verification automation, appointment reminder + reschedule sequence, recall call automation, review request automation. A baseline and pilot measurement plan for each. Indicative quote range. Clearly labeled as a proposed build, with no measured dental-client outcome claimed.

Insurance verification Appointment reminders Recall outreach Review requests

Read the full blueprint →

We run on our own automations.

Every lead source on this site, four forms and the AI chat, feeds one automated pipeline: a CRM contact and a Slack alert in under five seconds, with zero manual entry. The chat in the bottom-right is one of our builds. It answers questions, qualifies leads and hands the useful conversations to a person.

Explore the free n8n templates →

Three steps. Then it’s yours.

01

Free Audit Call

In 30 minutes, we map the current process, find the costly leaks and tell you what is worth automating.

02

Scoped & Built

One flat price, locked in writing. We connect your tools and build in your environment, typically in 1–2 weeks.

03

You Own It

You get full access, a walkthrough and written handoff. Your automation, your accounts, no Aplos maintenance fee.

Book a free audit.
We'll find your leaks.

Every business has 2–3 processes worth automating immediately. Book a free 30-minute call and we'll find yours.

Book a Free Build Call Call →