Your SaaS, usable by AI agents in two weeks.
Soxoa builds the production MCP server, sign-in for agents, and directory listings that let ChatGPT and Claude users work in your product. Built by John Arndt, who already runs them on Soxoa's own products. Fixed scope, and the code is yours.
AI consulting for other teams: AI consulting in Sacramento · AI consulting in the Bay Area
Already running on Soxoa's own products.
Each of these is a live MCP server you can connect from ChatGPT or Claude today. Yours is built the same way, in your repository, in ten working days.
Not a SaaS? Four engagements for working AI.
Start wherever your organization actually is. Every engagement stands alone and ends with a written record and a named owner on your side for the next decision: a working build when the evaluation clears the bar, an honest stop when it does not.
AI Readiness Assessment
“Where are we, honestly?”
Take the assessmentA self-scored survey of ownership, workflows, data and tools, people, and governance. It names your weakest dimension and the engagement that fits, including the case where the right move is to buy nothing yet.
See which readiness gap to investigate before committing budget.
About five minutes. Free. Runs in your browser; answers are not stored.
Live Build Workshop
“Can my team actually do this?”
Plan a workshopJohn and the people who do the work build one automation together, on approved examples of their real work, in one day. The team sees every decision, including what gets rejected and why.
Leave with one tested automation and a team that knows how to run and review it.
One working day on site or remote, plus a short preparation call.
AI Build Sprint
“Get the important one into production.”
Scope a sprintA fixed-scope sprint that takes one to three workflows from mapped to operating in your environment. Options are tested on real examples before anything is installed, a named person reviews what needs judgment, and everything is documented for handoff.
If the evaluation passes, put the workflow into operation with an owner and human review path.
Days to a few weeks, depending on the workflows. Scope and price agreed before work begins.
Fractional Chief AI Officer
“Who owns AI here, month after month?”
Discuss a fractional roleJohn joins your leadership rhythm as the accountable AI lead: setting priorities, deciding on tools and vendors, governing use, overseeing delivery, developing the team, and reporting what changed each month.
Give leadership a current record of what runs, what was declined, and who owns the next decision.
Monthly engagement. Scope and price agreed before work begins.
You work with the person holding the pencil.
John Arndt runs every engagement himself. He spent more than a decade in AI, cybersecurity, fraud and risk, privacy, payments and enterprise software, where a wrong answer costs real money. Soxoa also builds and runs AI software of its own, in public.
| Product | What it does |
|---|---|
| Annotated | Keeps a passage, video moment, or podcast clip together with your commentary and its original source. |
| MentionedOn | Tracks how AI assistants describe and recommend local businesses. |
| Intakra | Turns public buying signals into a reviewed why-now brief for sales teams. |
| Dokyumi | Extracts structured data from business documents, with review before it lands. |
| Portravo | Turns construction plans into reviewable quantities and estimates priced against dated local benchmarks. |
| Postedly | Lets an AI assistant prepare physical mail and faxes, with the sender reviewing every detail before payment. |
General notes. They apply to every sheet.
Build it on the real work
A demo on clean sample data proves nothing. Every option is tested on approved examples of your actual work before anything is installed.
Ask to see the tests
Ask what the system was tested on, where it failed, and what changed. The people who own the workflow should understand the evidence behind the decision to put it into operation.
Some work should stay manual
Rare cases, judgment calls, and anything where a mistake costs more than the time saved. Marking those early is part of the job, not a failure to sell.
A named person signs off
Where output reaches a customer, a record, or a decision, someone accountable reviews it. The review path is designed first, not bolted on after.
Your team can run it without us
Operating notes, access, and an internal owner are deliverables. If the system only works while we are in the room, it is not finished.
Reference sheets.
Agentic AI
Decide where a fixed workflow is enough, where an AI agent helps, and which actions need a human owner.
AI implementation
How a single workflow goes from mapped to operating, with the evaluation and handoff detail behind the Build Sprint.
AI training
Role-specific, hands-on programs for leaders and teams, beyond the one-day workshop.
Government & public sector
A distinct practice, scoped around agency authority, procurement, and public accountability.
Private & local AI
Evaluate local, self-hosted, and open-source options around your data boundaries, real work, and ability to maintain the system.
Bring the workflow that slows you down.
Thirty minutes with John. You describe the bottleneck; he tells you what he would mark up, what he would leave alone, and whether Soxoa is the right fit.