What VADD asked for, and how we would deliver it.
VADD's posting for a Software Platform Integration Lead describes about fifteen AI-built tools, half already in production, and a clear set of first-year goals. Each goal is listed here. Our markups, in the clouds, show how we would meet it.
| Goal | What it means day to day | Markup: how we deliver |
|---|---|---|
| One loginand a user directory | Everyone signs in once. Their role decides what they can see, change and approve, in every tool. | 1One sign-in in front of every tool, with roles and a second approver for money moves. We run this setup today. |
| Shared recordsprojects and jobs | A project or job is created once, and every tool uses that same record instead of its own copy. | 2A shared project and job service that each tool reads from, so names, numbers and people never drift apart. |
| Connected toolsschedule, estimate, drawings | A schedule change shows up in the estimate and the cash flow without anyone retyping it. | 3Schedule exports from Primavera P6 or MS Project linked to estimate lines, so dates and spend move together. We have already built an estimating system where estimate, contract, payments and job costs share one record. |
| Feedbackin every tool | People report problems and ideas from the screen they are on, and see what happened to them. | 4A feedback button in every tool, feeding one weekly review, with each request tracked through to a release. |
| CM platformbuilt from these pieces | The new construction management platform assembles the shared parts instead of starting over. | 5Built on the shared login, records and document store, so it is one more room in the building, not a new building. |
| Productionhosting and access | Cloud servers, on-site computers, domains, certificates, zero-trust access, backups and monitoring, with a status log for every tool. | 6We already run servers, secure tunnels, zero-trust access, encrypted nightly backups and health alerts. |
| AI key controlscost and access | AI access is managed in one place, not handed out key by key. | 7Keys held on the server, usage tracked per person, and local models for private documents. |
Field hardware
Scanners, tablets, cameras and per-job servers stay with VADD's embedded-systems group. We connect them to the shared login and records.
License review
Every package we use allows free commercial use, and each alpha ships with a license review, as the posting asks.
Guides with every release
Each new tool ships with a short white paper, a user guide and its feedback channel, so it gets used after launch day.
What VADD's platform could do, playing live.
Four short samples of the kind of tools we would build on a shared platform. Each plays on its own; tap one to jump. All figures are made up.
- ✓Budget lines add up to the contract value
- ✓No line is over budget
- ✓Retainage matches the contract terms
- Electrical, Job 2: moved 3 weeks. $2.1M of spend shifts from June to July.
- Field reports: 14 filed this week; 2 flagged a safety follow-up.
- Contingency: $4.02M left of $4.80M (84%). Two change orders are waiting for approval.
- Project manager submits a $180,000 change order for electrical on Job 2
- The approver signs off from their phone
- Budget and history updated; every total still ties
What we would build next for a data center builder.
Once the tools share one login and one set of project records, these become straightforward. Ideas to discuss with VADD's group leaders.
Schedule linked to estimate
The connection the posting names first: a schedule change updates the estimate and the monthly spend forecast, with no retyping.
Field capture straight into reports
The second one it names: photos, notes and headcount captured on site and rolled into the weekly report automatically.
Badges tied to the user directory
One person, one record: their login, their badge and the jobs they are assigned to, kept in step.
Contracts and POs read on upload
Values land on the right budget lines, checked for anything that does not add up.
Pay applications checked before approval
Each pay app compared with its commitment and the budget, so overbilling is caught before money goes out.
Off the low-code module
Move the low-code pieces onto the shared stack, one at a time, without breaking what people already use.
One document store
Contracts, drawings, reports and invoices stored once, found by project and job, backed up every night.
A status page for every tool
What is live, what is in testing, who uses it and what feedback is open: the status log the role calls for.
We have already done this job, in three industries.
Each of these started as a prototype and is now software people use every day, built with AI coding agents and run by us. One of them is a contractor's estimating system.
| Project | What it does | Scale |
|---|---|---|
| Operations hubHospitality group, about $100M | Custom software on top of Restaurant365: bank feeds and reconciliation, daily sales, labor, invoices, P&L packages, recipe costing, and investor management and relations. We also hold the fractional CFO role and set up daily automation across the company. | 12restaurants on one hub, 20+ jobs run on their own every day |
| Tax preparation pipelineAccounting firm, our own | Blockchain Crypto Tax Prep runs on a pipeline we built: quotes, a secure client portal, AI agents that prepare the work, and a preparer who reviews and signs off. An ownership hub tracks who owns what across entities. | 1,530documents taken in through the portal, 264 client files |
| Estimating and job costing systemRenovation contractor, live in production | From site visit to signed contract to final payment. The owner films a walkthrough, narrating the scope room by room, and AI drafts the estimate from it, starting from the most similar past job. Prices the owner corrects are saved for every future estimate. Estimates export to Excel and import back, and print as PDF or Word on the company letterhead. Job costing tracks spend, hours and margin against the estimate, and the payment schedule flags each draw when it comes due. | 190line items in a single estimate |
| Our own operations platform1st Class, internal | Many tools run from one central server: one login with separate roles, secure access from anywhere without a VPN, nightly encrypted off-site backups, health checks on every service, and phone alerts when something needs a person. | 1login across every tool we run |
Custom web apps
Hubs, portals, dashboards and approval screens, built for desktop and phone.
AI agents that do real work
Agents that prepare files, draft emails and answer from live data. A person approves before anything goes out.
Document reading
Contracts, invoices, statements, PDFs and spreadsheets read and checked against the numbers they should match.
Scheduled automation
Jobs that pull data, post entries, build reports and send emails on their own, with a watchdog that alerts a phone when one fails.
Connecting systems
Accounting software, bank feeds, card processors, point of sale, email and calendars. With no API, we drive the website the way a person would.
Reporting and finance packages
P&L, balance sheet, cash flow, projections, scorecards and investor reports, with Excel and PDF out.
Secure hosting
Servers, domains, zero-trust access, logins and roles, encrypted off-site backups and health monitoring.
Client and team portals
Secure document upload, status tracking and automatic reminders, so nobody chases files by email.
Alerts and messaging
Phone alerts, daily recap emails, WhatsApp and chat bots that tell the right person what needs them.
Accounting and CFO work
Fractional CFO, bookkeeping, bank reconciliations, job costing, ownership records and tax preparation.
Estimating and job costing
Estimates drafted by AI from a site walkthrough, a price book that learns, and job costs tracked against every estimate.
Guides and rollout
User guides, step-by-step procedures and in-person training, so a tool gets used after launch day.
The Integration Lead role, line by line.
VADD's posting for a Software Platform Integration Lead describes five jobs. Here is where we have already done each one.
| VADD needs | What it means | Where we have done it |
|---|---|---|
| Own production | Hosting, domains, secure access, backups, monitoring, users and roles, AI key controls | We host and run every system on these sheets: our own server, Cloudflare access, nightly encrypted backups, alerts, scoped logins, and AI keys held on the server instead of in anyone's browser, and local models on our own hardware for private work. |
| Integrate | One login, shared project and job records, tools that feed each other | Our tax pipeline, client portal, assistant and reporting share one set of client records. The restaurant hub joins bank, sales, labor and invoices into one daily picture. |
| Build with AI agents | Turn a short outline from an engineer into a working alpha, with a user guide | This is how we build everything. We are accountants and developers who direct Claude Code, have Codex review the result, and read, test and correct it before it reaches a user. |
| Drive adoption | Win users one at a time until the tools replace the spreadsheets | Our users are accountants, a chef, a CFO and a renovation contractor, none of them developers. We measure what they actually use and change the tools to fit. |
| Report | Regular progress with the CEO, COO and VPs | We already prepare reporting for an ownership group and its investors. Plain numbers, what changed, what needs a decision. |
Our AI bench
We pick the model for the job instead of tying VADD to one company, and we have one check another's work before anything ships.
Claude Code
Our main coding agent. It builds the tools, and we read, test and correct what it writes.
Codex
A second coding agent from OpenAI. We use it to review plans and code written by the first, so mistakes get caught by a different set of eyes.
Claude and GPT
The models inside the tools themselves: reading contracts, drafting messages, answering questions about the project.
Local models on our own Mac mini
AI that runs on hardware we own, so private documents never leave the building and routine work costs nothing per question.
Accountants and developers. That is the point.
Project controls is accounting with hard hats on. Budgets, commitments, change orders and job costs are ledgers. We reconcile money for a living, so the numbers in your platform will add up the way a CFO expects.
- ✓We know a budget has to tie back to the estimate plus every approved change, and what to do when it does not.
- ✓Job costing, bookkeeping and finance are our day job, one of the hands-on backgrounds the posting asks for.
- ✓Estimating is too: we built a contractor's estimating system, from AI-drafted estimates to job costing against them.
- ✓Every figure we show carries its source and the time it was read. A confident wrong number is the worst thing a finance tool can do.
- ✓Two-person approval, audit history and roll back are how we already run our clients' books.
- ✓The posting says about 95% of the code will be written with AI agents. That is how we already work: we direct the agents, then read, test and correct what they write.
- ✓We write and read Python, run our own Linux servers, and handle networking, security and APIs ourselves.
VADD's own plan, run by a team that has done it before.
We follow the plan in VADD's posting. A team starts faster than one new hire learning on the job.
Close by
We are in Maryland, a short drive from Leesburg, so site visits and in-person sessions are easy to schedule.
Devices
We'd partner with VADD's embedded-systems group on devices and connect them to the platform.
Scheduling
We read Primavera P6 and MS Project exports. Your schedulers stay the experts on the schedule; we make it flow into budget and cash flow.
Next step: a 30-minute demo of the systems on these sheets, running live.
We will also sketch how VADD's tools could come together behind one login, starting with the ones in daily use.