You Don't Need to Become
an AI Expert.
We build AI systems that handle your administrative work — deployed on your infrastructure, with your data, under your control. You get the benefits. We handle the technology. Then we hand you the keys.
There's a skill for that.
Serving non-profits, schools, churches, unions, and startups — nationwide.
Most people feel behind on AI — and don't have time to catch up.
You want the advantages. You don't want the homework. That's exactly where we come in.
Missed calls & slow follow-up
Every unanswered inquiry is a person who needed you and moved on.
Paperwork that never ends
Invoices, contracts, grant reports, intake forms — hours a week, every week.
Scheduling chaos
Coordinating people, rooms, and time is a full-time job on top of your real job.
Nothing gets remembered
Client and member history lives in someone's head — until they leave.
Fear of getting it wrong
Data privacy, security, vendor lock-in — the risks feel bigger than the reward.
No time to learn a new tool
You're already stretched. The last thing you need is another platform to master.
You already trust this model. It used to just be called IT.
Before managed IT providers existed, small organizations either couldn't afford an in-house tech department or muddled through alone. MSPs showed up with a simple pitch: we'll manage your servers, backups, security, and helpdesk — you don't need to understand any of it, you just need to trust us and pay a predictable fee. That industry is now worth well over $150 billion, built almost entirely on trust rather than technical evaluation.
Think about the last time your organization set up a new office suite, installed the apps everyone actually needed, configured the network, or loaded years of records into new accounting software. Chances are you didn't do that alone — you brought in a specialist, they handled it, and you got back to your actual work. AI workflows are no different. You could spend weeks figuring out which tools fit your organization and stitching them together yourself, or you could bring in a specialist to get you up and running.
That's the model we're bringing to AI. Call it AI IT.
| Traditional Managed IT | Infosoul "AI IT" |
|---|---|
| Servers & networks managed for you | AI systems managed for you |
| Helpdesk when something breaks | On-call consultant when something breaks |
| Data backups & security monitoring | Data privacy & security-first builds |
| Predictable monthly fee | Predictable setup fee, retainer optional |
| You don't need to understand networking | You don't need to understand AI |
"You don't hire a managed IT provider because you understand networking. You hire them because you trust they'll handle it. That's exactly how we want you to think about AI."
Built for organizations that run lean and mean well.
Every sector has its own administrative burden. We build AI appliances for each one.
"We help non-profits reclaim time for their mission, not their inbox."
- Donor stewardship and giving history live in one place — nothing falls through the cracks.
- Grant reporting pulls from past submissions instead of starting from a blank page.
- Volunteer scheduling runs itself, with reminders sent automatically.
- Your donor and constituent data never leaves your control.
- Donor Agent — tracks giving, drafts thank-yous3–6 hrs/wk
- Grant Agent — drafts reports, tracks deadlines4–8 hrs/wk
- Volunteer Agent — schedules, reminds, tracks hours2–4 hrs/wk
- Inbox Agent — triages and drafts replies3–5 hrs/wk
"We give your principals and teachers back 50+ hours a semester."
- Master scheduling handles teacher certifications, class sizes, and constraints automatically.
- Parent requests get routed to the right person without staff chasing emails.
- Compliance paperwork gets tracked, with gaps flagged before they become problems.
- Student data stays on infrastructure you control — never sent to a third party.
- Scheduling Agent — builds and adjusts master schedules40–50 hrs/term
- Parent Agent — routes and answers common requests3–6 hrs/wk
- Compliance Agent — tracks paperwork and deadlines2–5 hrs/wk
"We handle the admin so you can focus on ministry."
- Attendance and giving trends surface automatically — no more manual spreadsheets.
- Visitors get a warm, timely follow-up every time, without staff having to remember.
- Volunteer schedules build and adjust themselves.
- Congregational data stays private — never used to train someone else's model.
- Attendance Agent — tracks trends, flags changes2–4 hrs/wk
- Follow-Up Agent — welcomes and re-engages visitors2–5 hrs/wk
- Volunteer Agent — schedules and reminds2–4 hrs/wk
"Members get answers 24/7 — not a hold queue."
- Common questions about policies, rosters, and leave rules get answered instantly.
- Grievances and cases get logged and tracked with full transparency.
- Staff only step in when a case actually needs a human — not for every routine question.
- Member data and case files stay on infrastructure the union controls.
- Member Agent — answers policy & contract questions10–15 hrs/wk
- Case Agent — logs, tracks, and escalates grievances4–8 hrs/wk
"You had the idea. We handle the rest."
- Your website, marketing content, and lead follow-up run without you touching them daily.
- Leads get scored and prioritized so you spend time on the ones ready to buy.
- Routine operations — invoicing, scheduling, reporting — happen in the background.
- You own the system outright. No monthly platform tax as you grow.
- Launch Agent — website, content, payments setupOne-time build
- Lead Agent — scores and prioritizes prospects5–10 hrs/wk
- Operations Agent — invoicing, scheduling, reporting4–8 hrs/wk
The appliances are the product. This is the practice behind them.
Most clients arrive wanting one thing fixed. Some need a packaged appliance; some need software written, a pipeline built, or an honest opinion about whether AI is the answer at all. We do the whole range, and we say which one you actually need.
AI consulting
What to automate, what to leave alone, what it will cost and what it will not do. Often the deliverable is a decision, not a system.
Research & development
Prototypes against your own data and your own questions, built to be evaluated rather than demonstrated.
Full-stack development
Front to back — interface, service, database, deployment. Built to be handed over and maintained by somebody else.
Data analytics
Getting a straight answer out of the records you already keep. Cleaning, joining and reporting on data that was never designed to be analysed.
Cybersecurity
Assessment and hardening for organisations that hold sensitive records and cannot afford a breach or a headline.
Custom software
When the off-the-shelf tool nearly fits and the gap is the whole job.
AI integrations
Wiring models into the systems you already run — your CRM, your inbox, your case management — instead of adding one more place to log in.
Workflows & automation
The repeated, boring, error-prone sequence, done the same way every time without somebody remembering to do it.
Animation, video & audio
Explainers, welcome films, narration and score, produced in-house with open tools. See Overground.
Archive digitisation
Putting a record online that exists only on paper, film or a dead hard drive — in its existing format. We index it; we do not edit it. Built for community newspapers, congregations and historical societies whose own archive is unreachable. Nothing is reformatted without being asked.
Every one of these runs on the same rule as the appliances: deployed on your infrastructure, with your data, under your control. We set you up and you keep the keys.
Think of AI like a washing machine for your admin work.
You don't need to know how it works. You load it, press the button, and it handles the task. There's a skill for that — here's the current menu.
Document Skill
Scans, extracts, and routes contracts, invoices, and forms.
See it in action →Client Memory Skill
Remembers every client or member detail and surfaces it instantly.
See it in action →Security Skill
Monitors systems and flags vulnerabilities before they're exploited.
See it in action →Avatar Skill
A branded AI voice and face for welcome videos, training, and multilingual announcements.
See it in action →Give your organization a digital spokesperson.
A branded AI voice and face — used for welcome videos, training content, and multilingual announcements — built once, reused whenever you need it.
| Frontier (ElevenLabs + Nano Banana) | Open-Source (Self-Hosted) |
|---|---|
| Highest realism, fastest to stand up | Slower to set up, no per-use vendor fees |
| Voice and likeness processed by a third-party vendor | Voice and likeness stay on infrastructure you control |
| Best for one-off or occasional content | Best for organizations generating content regularly |
This isn't a mockup. The face was generated by Segmind Tiny-SD (Apache 2.0), the voice by Piper (MIT), and the mouth movement by Wav2Lip — all open-source, self-hosted, no cloud API involved. The face is entirely synthetic and fictional; no real person's likeness was used. Hover over it, or tap it — every frame of lip motion was generated from the audio, not a canned animation or a volume pulse.
Generated using open-source components we'd deploy the same way for you: Piper for voice (MIT), Wav2Lip for lip-sync (CC BY-NC 4.0 — non-commercial; a real deployment would use a commercially-licensed lip-sync engine instead). We couldn't run any of this inside our own demo sandbox — it has no GPU and its network is deliberately locked down — so we ran it on ordinary consumer hardware (a 4GB GPU) instead and pulled the result in. That's the honest version of what a real deployment looks like: matching the model to the hardware you actually have, which is exactly what our Hardware Fit tool above helps you plan for.
Before we build one: we get written consent from anyone whose voice or likeness is used, and every generated video is clearly labeled as AI-generated. That's non-negotiable — trust is the whole point.
We also make the pictures and the sound.
Animation, video and audio production, done the way we do everything else: with open tools, on hardware you control, and handed over with the keys. Below is an example of what that looks like — Overground, a teaching tool for exploring the Milky Way, running live in your browser.
Overground · a tour of the Milky Way
The planets where they actually are tonight. Five thousand real stars in their true directions, the constellations in three dimensions, Voyager and the other probes at today’s distances, and a flight out to the centre of the galaxy. Every number on every card is the real one; only the distances are squeezed so the whole sky fits in one room, and it says so. Made for classrooms, libraries, congregations and anyone who wants to look up. Click a world to fly to it.
Animation
Stop-motion in code: puppets, sets and film grain generated rather than downloaded, so there is nothing to license and nothing to lose. Character work, motion capture, and a director’s camera, rendered in a browser or to a video file.
Video
Welcome films, explainers, training pieces and teaching tools like the one above. Title cards, a film look, and highlight reels cut automatically from a recording — without a production company’s invoice, and without your footage leaving your building.
Audio
Narration from open-source speech models, in any language your community speaks. Sound effects and score synthesised in code for the piece, not licensed for it. The same voices power our digital spokesperson.
Overground is a desktop-browser piece and takes a moment to build itself on first load. It works best with a mouse: click a world or a name to fly there, drag to look round, roll the wheel to close in.
Frontier or open-source. See both, side by side.
Pick a task. Choose which kind of AI answers it. Every response below is real, generated output — not scripted copy — from an actual frontier model and an actual self-hosted open-source model.
The frontier response was generated by Claude (Anthropic). The open-source response was generated by Qwen3 8B (Apache 2.0), running locally via Ollama on ordinary consumer hardware — no cloud calls, and the timing shown is the model's actual measured generation time. Both were generated once and verified, not simulated on each page load. In a live engagement, responses run against your real data, inside a sandboxed environment tailored to your systems.
We're not tied to one vendor. We're tied to what works for you.
We help you choose — and we'll tell you honestly when one is a better fit than the other.
| Factor | Frontier (OpenAI / Claude / Gemini) | Open-Source (Self-Hosted) |
|---|---|---|
| Best for | Complex reasoning, fastest setup | High-volume, routine, data-sensitive work |
| Where your data goes | Vendor's servers | Stays on your infrastructure |
| Cost as you grow | Scales with usage / seats | Predictable, fixed infrastructure cost |
| Vendor lock-in | Higher | None |
| Setup effort | Low | Higher — we handle it for you |
What can actually run on your setup?
Think of this like a device-agnostic repair shop — we don't care what hardware you're running or which vendor it came from. Tell us what you've got, and we'll tell you honestly what fits locally, and what's better left to a frontier API.
A directional estimate, not a benchmark. Our diagnostic session includes a full technical assessment of your actual environment.
Why not just use ChatGPT or Claude yourself?
ChatGPT (OpenAI) and Claude (Anthropic) are genuinely impressive general-purpose AI assistants. We often build on the same underlying models. The difference is what happens after you open the tab.
| ChatGPT / Claude, on their own | Infosoul "AI IT" |
|---|---|
| One conversation at a time | Runs continuously in the background |
| You do the typing, prompting, and copy-pasting | Built once so no one has to prompt anything |
| Doesn't know your donor list, roster, or invoices | Wired directly into your actual data and workflows |
| You manage the account, the risk, the setup | We manage security, privacy, and uptime for you |
| A great starting point | A finished system that just works |
We're not affiliated with OpenAI or Anthropic. We simply build, secure, and maintain the systems on top of their models — or open-source alternatives — so your team never has to.
We set you up. You keep the keys.
No jargon, no lock-in, no learning curve required.
Tell us what's broken
A short conversation about what's wasting your time or costing you money.
We build your system
We design and deploy the AI appliance that solves it — on your infrastructure, with your data.
We hand you the keys
You own the code, the data, and the infrastructure. Nothing is held hostage.
You choose the relationship
Forget us. Keep us on call. Or bring us on as an ongoing partner. Your call, always.
Forget us, keep us on call, or partner with us.
It's your system. We don't believe in locking you in.
Setup Only
- We build, deploy, and train you on your system
- You own the code, data, and infrastructure
- No ongoing relationship required
- You're free — walk away anytime
On-Call Consultant
- Everything in Setup Only
- Troubleshooting when something breaks
- New features built as you need them
- No retainer — pay only for what you use
Ongoing Partner
- Everything in On-Call
- Priority access and proactive check-ins
- Quarterly strategic reviews
- Continuous system optimization
We think about how systems fail — before we build them.
If we can crack it, we can patch it.
Our background is in systems security: understanding how things break, where the weak points are, and how to close them. We bring that same mindset to every AI system we build — so you get something that's not just functional, but defensible.
We're independent. We're not tied to a single model vendor, and we don't profit from locking you into one.
- ✓ Your data stays on infrastructure you control
- ✓ Security reviewed before deployment, not after
- ✓ No hidden costs — everything disclosed up front
- ✓ Open-source options mean no vendor lock-in
- ✓ You own everything we build for you
We work with clients nationwide.
Most of our work happens remotely — video calls, shared documentation, ongoing support. In-person consultations are available on request for discovery sessions, training, or handoff (a travel fee may apply).
Location isn't a barrier. It's just a detail.
The AI Age, For the Layman.
Before an organization buys anything, somebody has to explain the thing. This is that explanation, given in your room, to your people — and it is the way most clients meet us. No background assumed. The heart of it is not the technology: it is the history, the open questions these machines put to us about thought, creativity, consciousness and life, and what they are doing to work, power and the truth. The room asks the questions, and watches the machine get one wrong.
The rooms it is built for. Photographs from open-licence galleries, credited in media/CREDITS.md; they are not our events.
The AI Age, For the Layman50 min class or lunch slot · 75 min evening edition, Superintelligence for the People
Seventy years of history in twenty minutes, so that “it is all new” stops working on you. Then the questions the room actually came with: does it think, is it creative, is anything home, is it alive, and what would follow if it were. They are open, they cannot be avoided, and they are asked here properly rather than on the companies’ terms. Then three discoveries, one of them this September, and what they do to your medicine, your patents, your bank and your university, which almost nobody has put to the people they affect. The technology only as far as those questions need it. The evening edition, Superintelligence for the People, takes the word apart: what is actually being built, who profits, what the people who own it believe in their own words, why the warehouse may not be the only road there, and what you keep.
- Room — any size, seminar room to hall.
- Leaves with — a one-page plain-language guide for every seat.
Every Field, For the Layman50 min, one field at a time
The same talk, asked of one field. The second book in the series puts six questions to every discipline in the catalogue: what the machine already does here, what it cannot do here and why, what it changes about what the field is, who owns it here, what happens to the people, and what the field’s people should do, ending in a line for Monday and a line for the argument. Fifty-one fields, from mathematics and medicine to law, history, music and the trades. A campus can book it for the history department, the nursing school and the law faculty in one week, and each room hears its own. Ask for the entry as the listing.
- Room — a department, a school, a programme’s students and faculty together.
- Leaves with — the entry for the field, and the one-page guide.
Ask It Yourself90 min, hands-on
The room puts its own questions to a model, watches it get one wrong, and learns how to check. Laptops or phones, or two to a screen.
- Room — up to 30 with laptops, 60 in pairs. Internet needed.
- Leaves with — a checking habit, and the guide.
Make It Yourself90 min, the generative one
Everyone makes something real for their organisation: the newsletter, the flyer, the grant paragraph, the announcement in two languages and in a spoken voice. Built in advance from three of your own documents, so the exercises are yours. Models come from the open-source galleries, Hugging Face and the Ollama library, free to download and keep.
- Room — up to 25 with laptops. Internet needed.
- Leaves with — a finished piece, and the method.
Run Your OwnHalf day
Install an open-source model from Hugging Face or the Ollama library on your organisation’s own laptop and leave with it working. No subscription, and nothing leaves the building.
- Room — up to 12, each with a laptop. Power strips and internet needed.
- Leaves with — a running machine, and a written setup sheet.
Give It HandsHalf day, for staff and committees
Agents, the socket and the foreman, in plain words: a script running in the machine’s window, the standard plug (MCP) that lets it reach your files, and the software that runs several of them on a job. Then the practical part: connect the model you run to one of your own folders, read-only, and decide as a group what it may touch, who checks, and what it may never do. Uses the free agents people are actually installing this year: Pi, OpenClaw and Hermes Agent, and what tells them apart.
- Room — up to 12, each with a laptop. Best after 301.
- Leaves with — a working read-only agent on your own machine, and a written policy: what it can touch, who checks, whose machine.
The talk, the workshops, the buildAcross a month, for a library, congregation or local; 401 when the room is ready
The talk, then the two workshops, then the build. The room that heard the explanation is the room that leaves running its own machine. Every participant who completes the set receives a certificate of completion; colleges now sell this syllabus by the seat, we teach it by the room. Continuing-education credit through a partner college on request.
Any session, for a teamTraining days and in-service
The same session for the people who work there, with a policy conversation added: what the organisation will and will not use these tools for, written down before you leave.
For the coordinator: the syllabus, one page per class with outcomes, contact hours and what the room leaves with, and the glossary, the catalogue words beside ours. Colleges now sell certificates in generative and agentic AI, twelve to eighteen weeks of Python and rented services at about three thousand dollars a seat. They teach building on somebody else’s machine. They do not address what these classes do: explaining it plainly, checking it, and running it in your own building.
Everything a coordinator needs to say yes
Presented by Brock Satter, Director of Infosoul Laboratories
(LinkedIn).
BA student in Comparative Studies of Science, Technology and Culture, University of Minnesota
— an individually designed major focusing on information technology,
biotechnology and artificial intelligence, thirty years before the current wave, with the anthropologist Harvey Sarles, the department’s former chair, as senior project adviser. In 1993 the MIT Media
Lab turned down his application to its science, technology and society programme, which he
mentions so you know how long he has been at this: in 1985 he was writing BASIC to automate
the Players Handbook and Dungeon Masters Guide, and then TSR’s
Top Secret and Marvel Super Heroes, and in 1987 a text-based slam-dunk contest
game. He has spent thirty years in front of rooms as a political candidate and community organizer:
debates, classes, conferences, rallies and media interviews. He builds these systems for a living, is
not a futurist, and is not selling anything from the stage. If an appliance is the right
answer for your organization we will say so afterwards, in private, and you are free to
ignore it.
The book: the talk, expandedA primer in the history of science and technology, for the uninitiated · coming soon · about 180 illustrated pages · paperback and ebook
“Everyone meets these machines at the same door, the engineer and the grandmother alike, and on the other side is something nobody fully understands and anyone can use. An enigma and an invitation.”
“Whether or not the data centres are shut down, whether or not a model is banned, whether or not a company fails or a law passes, the thing itself is now known, to the people of the whole world. The method is in the papers, and the file is on drives in every country, copied more times than anyone can count. A file of numbers can talk back to you. That is a fact about the world now, like the fact that a lever lifts, and nobody gets to unknow it.”
Seven questions, one to a chapter, read in one sitting. Why did it arrive all at once? What is it? What does it mean? What is it good for? When should you not believe it? Whose is it? Can you have your own? The one-page guide is the appendix.
Made in conversation with a machine, and honest about it. Chapter four opens with the prompts that shaped the book, the corrections, and what the machine got wrong; an appendix interviews the machine itself. The question is not when the machines become gods. It is what happens between a person and one, and who is in charge.
For hosts: a copy for every seat can be added to any booking at cost. Say so in the booking form.
One email when it is published. Nothing else.
Inside the book
Photographs from Wikimedia Commons: Alan Turing, Elliott & Fry, public domain; Apple IIc, Bilby, CC BY 3.0; clouds off Alejandro Selkirk Island, NASA, public domain. Diagrams by the author. All credits.
Book a session
Pick the session, give us two dates, and we reply within one business day with a confirmation and the invoice. No deposit.
Ready to see whether this is a fit?
Tell us what's wasting your time. We'll tell you honestly whether AI can fix it.
We reply within one business day. No pressure, no obligation.