Layer 1
Intake and triage
Inbound email, forms, texts and voicemail transcripts read, classified, prioritized and routed to the right person or system, with a draft reply ready for review.
AI automation · Operations-heavy small businesses
AI automation for operations-heavy offices, integrated with the systems you already run.
An operations layer that classifies and routes what comes in, extracts what matters from documents, produces the reports your team assembles by hand each month, and runs the repeat steps in between. A person approves anything that reaches a customer. Each phase is scoped against a time baseline so the saving is a measured number.
The approach
Where the time goes, and where AI reliably recovers it.
Most offices already run capable vertical software: AppFolio, Buildium, Propertyware or Rent Manager in property management; Clio, QuickBooks, HubSpot or an industry CRM elsewhere. The hours go into the gaps between those systems: reading, classifying, retyping, chasing and summarizing. That is where current AI is dependable, and it is where this work is built, without replacing the systems your team already knows.
Layer 2
Document handling
Invoices, applications, leases, inspection reports and vendor bids read and turned into structured records, flagged when something is missing or out of range.
Layer 3
Scheduled reporting
Owner, client and management summaries produced on a schedule from data you already have, in your voice and format, reviewed before they go out.
Layer 4
Back-office workflows
Multi-step routines such as move-out processing, renewal outreach or vendor follow-up run as a checklist with AI doing the steps and a person approving the gates.
Project types
Workflows that make up a typical first phase.
Maintenance and service request intake
Requests from any channel classified by urgency and trade, matched to the unit or account, checked against warranty or coverage rules, and dispatched with a draft to the vendor and a status note to the customer.
Typical saving: the coordinator's first hour of every day.
Leasing and sales inquiry response
Inbound inquiries answered within minutes with accurate availability, qualification questions asked, showings offered from a live calendar, and a summary logged to your CRM. Human takes over the moment it gets specific.
Typical saving: faster response, fewer lost leads on evenings and weekends.
Monthly statements and reports
Owner statements, client reports or board packets drafted from your accounting and operations data with the narrative written, exceptions called out and the numbers reconciled, ready for a five-minute review instead of a two-day build.
Typical saving: most of the last week of every month.
Document extraction and checking
Applications, invoices, insurance certificates, inspection reports and bids read into your system of record, with missing fields, expired dates and out-of-range amounts flagged for a person.
Typical saving: the retyping, and the errors that came with it.
Knowledge and policy assistant
Your procedures, contracts, price lists and past decisions made searchable in plain language for your team, with answers that cite the source document so people trust and verify them.
Typical saving: the interruptions to whoever knows where everything is.
Follow-up and renewal sequences
Renewals, overdue items, review requests and dormant leads worked on schedule with personalized drafts, escalated to a person when a reply needs judgment.
Typical saving: the follow-ups that never happened.
Packages
Three stages, each priced before it starts, each one optional after the last.
You never commit to the whole program. Discovery produces a number and a ranked list; you decide whether Phase 1 is worth it. Phase 1 produces measured savings; you decide whether to keep going.
Stage 1
Discovery and roadmap
$2,500 · fixed fee · 2 weeks
For: an office of roughly 3 to 25 people that suspects hours are going into repeat work between systems, and wants a number and a plan before buying anything.
Outcome: you know, in hours per week and dollars per month, where the time goes and which three workflows to fix first, with a fixed price for fixing them. If the answer is that AI will not pay for itself yet, you hear that too.
- Sit-downs with each role; workflows mapped, volumes and minutes counted
- Time baseline: the current cost of each repeat task
- Capabilities walkthrough: a one-page menu you rank Now, Next, Later or Skip
- Phase 1 scope with acceptance criteria and a fixed price
Credited in full against Phase 1 if you proceed within 30 days.
Stage 2
Phase 1 build
From $7,500 · up to 3 workflows · 4 to 6 weeks
For: an office that has its top three workflows ranked, usually from Discovery, and wants them running inside the systems the team already uses.
Outcome: the workflows run in production with a person approving the gates, and the saving is measured against the baseline in the first 30 days. Your team can operate it without me.
- Up to 3 workflows connected to up to 2 of your systems
- Shadow mode alongside your team first, then live with review points
- Logging of every automated action, and a dashboard your team reads
- Documentation, training and a written handoff
- Paid against milestones; acceptance criteria written into the scope
Cap: 3 workflows and 2 system integrations. Additional workflows from $2,500 each; additional integrations quoted in the scope. AI provider usage is billed to your own account, typically tens of dollars a month per workflow.
Stage 3
Operate and expand
From $750 a month · 3-month minimum
For: an office with Phase 1 live that wants it monitored, tuned and extended without hiring for it.
Outcome: the workflows keep working when your software updates and your volume changes, error rates keep falling, and the Next list turns into working automation one phase at a time.
- Monitoring and fixes when a system or a format changes
- One improvement cycle a month from the error log and your team's notes
- Quarterly re-measurement against the baseline, in writing
- Next workflows scoped and built as small fixed-price phases
Sized to the number of live workflows. Cancel with 30 days' notice after the minimum; everything stays running in your accounts.
Payment terms: Discovery is invoiced at the start. Phase 1 is 40% at kickoff, 30% when workflows enter shadow mode, 30% at acceptance against the written criteria. Operate and expand bills monthly. Everything built runs in your accounts and the code and documentation are yours.
Why the scoping matters
The scoping decides the result. It should start from how your office works, not from what a tool can do.
I ran a $20M international business and set up a company's S&OP process before I wrote production software, so discovery starts from how the office works and what each role's week costs, not from what a tool can do. Then I build it. The Prepwell app, with its AI photo recognition, recipe extraction, payments and grocery integrations, was designed and coded end to end by me, including the cost controls and review points that keep AI features reliable and affordable in production.
Every AI engagement starts the same way: a written discovery of each role's week with real counts, a one-page capabilities menu you rank Now, Next, Later or Skip, and a Phase 1 of up to three workflows with a fixed fee and acceptance criteria. That order matters. Tools chosen before the baseline exists get measured against a feeling instead of a number.
Discovery, two weeks
Sit with each role, map the workflow, count the volume and the minutes. The output is a time baseline and a ranked list of candidates.
Capabilities walkthrough
A one-page menu of what is possible for your business. You mark each item Now, Next, Later or Skip and star the top three. That becomes phase 1.
Phase 1, four to six weeks
One to three workflows built, connected to your systems, run in shadow mode alongside your team, then live with review points. Fixed price, paid against milestones, with acceptance criteria written into the scope.
Measure, then expand
Time saved and error rates measured against the baseline. You decide what moves from Next to Now. Ongoing operation on a retainer sized to the workload.
Controls
Controls built into every engagement.
- Human in the loop for anything customer-facing until the error rate is measured and you choose to loosen it.
- Your systems stay your systems. The layer reads from and writes to them; it does not replace them.
- Every automated action is logged with what it saw, what it decided and why, so mistakes are findable and fixable.
- Data stays in your accounts, AI providers are used under business terms that exclude training on your data, and access is scoped to the minimum each workflow needs.
- Time saved is a measured number, taken before and after against a baseline, not a projection.
- Your team can run it. Documentation, dashboards and a handoff that does not require me to stay on the payroll.
Questions
Common questions.
Do we have to replace our current software?
No. The approach is a layer that reads from and writes to what you already run, whether that is a property management platform, a CRM, a shared inbox or spreadsheets. Replacing core systems is expensive and rarely where the time goes.
How do you keep AI from making mistakes with our customers?
Every workflow is designed with explicit review points: the AI drafts, sorts or extracts, and a person approves anything that goes to a customer or changes a record, until the error rate is measured and you decide to loosen it. Nothing customer-facing ships without a human in the loop first.
What does a first phase cost and how long does it take?
Discovery is a fixed $2,500 over two weeks and produces the baseline and the plan. A first phase is up to three workflows delivered in four to six weeks from $7,500, with time measured before and after so the saving is a number, not a feeling. Ongoing operation and expansion runs from $750 a month, sized to the number of live workflows. Packages are above.
Is our data safe?
Data stays in your accounts. AI providers are used under business terms that exclude training on your data, access is scoped to the minimum each workflow needs, and every automated action is logged so you can see what happened and why.
We already use an AI phone answering service. Does this conflict?
No, and that is a common starting point. Vertical AI tools for phones, maintenance or leasing each do one job well. This layer sits across them, ingesting what they produce and connecting it to the rest of your operation, and changes nothing inside them.