What I Do
The software your business is missing, chosen to fit the problem rather than a vendor's roadmap.
Six lines of work. Most clients arrive needing one, and they combine. Built on what you already pay for when that's the honest answer, and from scratch when it isn't.
What a unit of your output costs
Whatever you count (tons, units, installs, jobs, patients, billable hours), the question is the same: what did it cost to produce, and how does that compare to what you assumed when you priced it? I build the reporting that answers that while the work is still running, traceable down to the records behind every figure.
- Budget vs. actual by cost code, live rather than at close
- Fully burdened labor, equipment and material cost per unit
- Margin reporting, and estimate vs. actual reconciliation
- Invoicing, payments and expense tracking wired into the same data
Paperwork that routes itself
Most of the administrative day is moving information by hand: emailing a file to the person who needs it, re-typing a number into a second system, checking one report against another to see which is wrong. That work is mechanical, which means it can be done for you, the same way every time.
- Documents filed and routed to the right people automatically, with version and access under control
- Approvals routed by dollar amount, with a record of who approved what and when
- Receipts and invoices read, coded and recorded in your accounting system without retyping
- Less email, less data entry, less cross-referencing by hand
An agent that reads your real systems and stays yours
Most AI projects put a chat box in front of a document library. That finds a policy; it cannot tell you which machines are down. I build agents connected to your live operational data, that name the source of every figure, and that say they do not know rather than inventing a number.
- Your data is never used to train anyone else's model
- Runs in your own accounts, under an identity you control
- Grounded in your live records and your own document set, not a copy
- Daily production, cost, manpower and equipment updates, pushed or asked
- Changes only a short list of things you approved, and asks first
- Tested against documents and mail that try to give it orders
Purpose-built tools, not generic templates
When an off-the-shelf tool doesn't fit, or costs more in workarounds than it saves, I build the thing that does: client portals, field capture, internal dashboards, and progressive web apps that work on a phone in poor signal and install without an app store. This is where needs go that no platform, Microsoft included, will cover.
- Client and operations portals with real access control
- Progressive web apps for field use, offline-tolerant
- Public sites built to be found by search and by AI systems
- Hosted in your own cloud account, secured properly
Making your existing systems talk to each other
Most businesses don't have a software problem. They have a dozen systems that don't share data. I design the architecture and build the integrations that connect accounting, field operations, CRM, and reporting into one coherent picture, around whatever you already run rather than in place of it.
- System audits and integration roadmaps
- Connecting the tools you have through their official doors, so it keeps working after updates
- Data flow and single-source-of-truth design
- Accounting integration, including QuickBooks, tied to operational data
Getting more out of the Microsoft tools you already pay for
One option among the five above, and often the cheapest place to start. Most organizations already own Microsoft 365, SharePoint and Teams, and use a fraction of what they're capable of. Where that covers your problem, I say so and build there instead of selling you something new.
- SharePoint, Teams and Outlook doing routine work on their own
- Sign in with the Microsoft account your people already have
- Workflows and shared records built on what is already in your subscription
- Agents in Teams and Copilot, or custom ones where those can't reach
- The account itself checked against published standards and locked down, before anything is built on it
None of this is industry-specific. I've done it in construction, power delivery and healthcare, and those references count. But the requirements are only that you produce something countable, that the people doing the work know what happened, and that you can grant access to your own data. See each of these running, or read what an agent is, in plain words.
How I work
Every engagement runs through the same three tests before I call it done.
It has to remove steps
If the same information still gets entered twice, I haven't finished.
It has to be obvious
Tested with the person doing the work, not the person most excited about it.
It has to be built so AI can reason about it
Bad data should be structurally impossible to enter, not just caught later.
Have a specific problem in mind?
Tell me what's broken and I'll tell you honestly whether I'm the right fit.
Or take the first look: six taps, one page, no email needed.