Your AI assistant.
Set up by us. Talks to you on Telegram.
Set it and forget it. Once it's wired in, you don't need to ask us again. Your digital assistant handles the rest.
You know AI can help your business. You don't know how to set it up. That gap is the whole problem. We close it for you.
Book a 30-min discovery call
"The fire of fires." That's how Daniel Miessler describes what AI is doing to the SaaS stack. Cheap models plus a private assistant mean a mid-size business can often replace several subscriptions instead of adding another one.
See The Fire of Fires, Daniel Miessler. We don't claim every tool gets replaced. For a typical Digital Assistant client, three to eight do.Everyone knows AI can help. Almost nobody knows how to wire it in.
You've tried ChatGPT. You've watched the YouTube videos. You've read about AI agents (assistants that run tasks for you in the background). You still do your own email triage, your own follow-ups, your own admin. The gap is not understanding AI. The gap is somebody who actually connects it to your business and makes it do things for you.
A full AI setup, built for you end to end.
Your own private system
Your own dedicated system. Not shared with anyone else. Your data stays yours.
Connections ready on day one
A concrete, named list of tools we can wire in during setup. See the list just below the feature grid.
Custom agents for your work
Agents (assistants that run tasks for you) built around how you actually operate. Not generic. Not a template.
Telegram interface
Talk to it like you text a friend. No new app to learn. Works from your phone.
90-day support window
Something breaks or needs tuning in the first 90 days, we handle it at no extra cost.
Documented handoff
If you ever hire an assistant or ops person, they can pick up where we left off from the written docs we leave you.
Connections ready on day one
Concrete, not promisey. These are the tools we already have wiring for today. We'll tell you during the week-one engagement which of your tools fall inside the included five and which are outside.
- Gmail Email
- Google Calendar Calendar
- Google Drive Files
- Notion Docs
- Slack Comms
- Telegram Comms
- HighLevel CRM
- Stripe Payments
- Airtable Data
- DigitalOcean Infra
- GoDaddy Infra
Note on the AI models. Your system, your data, and your logins are yours. The AI company we use, Anthropic (the company behind Claude), runs as a public cloud service, and the messages we send them are covered by their terms of service. We link the current vendor terms in our privacy policy. You can review them before signing.
Compare $18,000 to hiring one person.
The cheapest admin hire in Ontario runs about $55,000 a year with benefits and onboarding. They take three to six months to get productive. They work 40 hours a week. They get sick. They take vacation. They eventually leave.
Your digital assistant is productive on day one because the playbook is already built. It does not sleep. It does not quit. It scales with your workload. Year one cost with us is $18,000. Year one cost with a human is roughly $55,000 plus the ramp cost.
| Dimension | Human assistant, year one | Digital Assistant, year one |
|---|---|---|
| Direct cost | ~$55,000 salary plus benefits | $18,000 one-time |
| Ramp time | 3 to 6 months | Live by Friday of week one |
| Availability | 40 hours per week, business hours | 24/7, responds in seconds |
| Sick days / vacation | 10 to 20 days per year | None |
| Turnover risk | Real. They eventually leave. | None. It doesn't quit. |
| Institutional memory | Walks out with them | Stays with you, documented |
| Onboarding load on you | Heavy, for months | A few hours across setup and training |
| Scales with workload | No. Hire again. | Yes. Parallel work is built in. |
Not an argument for firing anyone. An argument for setting up the AI layer BEFORE you're forced to hire.
Why $18,000 and not $1,800.
What you are buying is a year of research packaged into a five-day engagement that leaves you with a working system. Brandon spent the last twelve months figuring out how to wire private AI into a small business from end to end. Which connections work. Which fail. How to handle API-key rotation (an API key is the password that lets two tools talk to each other). How to set up Telegram without the bot dying. How to keep things running through reboots and software changes. Which agents (an AI agent is an assistant that runs tasks for you) actually earn their keep versus which burn money on nothing. The $18,000 skips all of that learning and hands you what works, already wired for your tools.
The reason the Digital Worker tier exists at $1,800 a month is for people who do not yet trust that claim. Start there, cancel anytime, upgrade when you see the returns.
What it would take to build this yourself.
If you try to assemble this on your own, here is roughly what the year looks like. Plain numbers from doing it:
- Time. 6 to 12 months of evenings and weekends learning Claude Code and the tools that make agents work. Not the surface tutorials. The edge cases.
- Infrastructure spend. $5,000 to $15,000 in server fees, API usage, false-start tooling, and side accounts while you figure out the shape of a real private setup.
- Dead ends. 5 or more connections that looked promising and turned out to be the wrong approach. You keep the lesson. You don't keep the time.
- Telegram rebuilds. Several. The bot quietly dies for reasons nobody explains. You learn why. You write a launcher that handles it.
- Software changes from the AI provider. The provider ships a change that breaks things. Your assistant falls over at 2am. You fix it, document it, and wait for the next one.
- Keeping it alive. The tools underneath keep changing. The version of this that works in month one is not the version that works in month ten. Staying on top of that is the permanent tax.
The $18,000 is a year of that work, already done. Five days of our time with you, instead of a year of yours. You get a working system by Friday and skip the losing year of figuring out which 30 things don't work.
"Start thinking about what you're paying for from various tech/SaaS vendors, and start thinking about whether it's possible to replace them with better/cheaper options or by remaking them yourself if relatively simple."
Daniel Miessler, Unsupervised Learning No. 525, Recommendation of the Week. Independent framing from someone who's been tracking this for years. Not our pitch, his.You're not hiring a seat. You're hiring infrastructure.
Fabio Hauser put it well: AI changed what we build, and it changed who we hire. A small business used to hire a person with a resume and hope they learned fast. Now the equivalent move is setting up infrastructure that compounds. The assistant you're hiring with Digital Assistant doesn't need to be trained by you. It shows up knowing Gmail, Calendar, Notion, HighLevel, and the common patterns of small-business admin on day one.
That seems like the better shape of a first hire in 2026. See Hauser's piece here for the longer argument.
Your own information matters more in the AI era, not less.
Tiago Forte, who wrote the PARA method, argues that organizing your own information actually matters MORE now, because AI performs better when your structure is clean. That's part of what the $18,000 buys: we sit with you during discovery and set the shape of your information up properly, so the assistant has something real to work from. Garbage in, garbage out still applies. See Forte's piece on PARA in the AI era if you want the longer version.
Day 0 to Day 90.
5-day engagement, live Friday
Brandon sits with you and your team all week. Half the time is discovery, mapping your workflows, tools, people, and bottlenecks. Half is building, wiring the assistant into what you already use. Live demo by Friday.
Handoff
90-minute training. Written docs. Telegram paired to your phone. You're using it live in your day-to-day.
Support
Fixes, tuning, new agents added. We stay on the project until the system is part of how you work.
The Gemini question.
Google is rolling Gemini into Gmail and Photos globally. That means Google is reading your data to build Google's context. A private Digital Assistant reads the same data to build YOUR context, sitting on infrastructure you control with credentials you own. Same raw material, different beneficiary. That difference seems like it will matter more, not less, over time.
What you pay us vs what you pay your own vendors.
The $18,000 covers our work, not the underlying infrastructure your instance runs on. Those you pay directly. We help you set them up during onboarding.
What Obsidian AI Labs provides for $18,000
- 5-day on-site or over-video engagement (Monday to Friday, week one)
- Full setup of the private PAI system on your infrastructure
- Up to 5 connections wired in during setup
- Up to 3 custom agents built for your specific work
- Telegram channel configured and paired
- End-to-end testing against real scenarios from your business
- 90 minutes of walk-through training plus written docs
- 90 days of support: fixes, tuning, new agent additions
What you pay directly to vendors (approximate monthly)
- DigitalOcean droplet (your dedicated instance): ~$12 to $15 per month
- Domain (if you don't already have one): ~$15 per year
- Claude API usage (actual use-based): ~$20 to $100 per month depending on load
- Optional: email-send or SMS services if we wire them in (Postmark, Twilio, etc): pass-through, your account
Total ongoing: typically $50 to $150 per month. You own every credential. You can pause or cancel any of these independently.
What's included in the $18,000
- 5-day on-site or over-video engagement (Monday to Friday of week one, Brandon embedded in your operation)
- Private PAI setup provisioned on dedicated infrastructure
- Up to 5 connections wired in during setup
- Up to 3 custom agents built for your role-specific work
- Telegram channel configured and paired to your phone
- End-to-end testing against real scenarios from your business
- Walk-through training session (90 min) and written docs
- 90 days of support: fixes, tuning, new agent additions
Payment terms: $9,000 to start, $9,000 on delivery. Interac e-transfer or wire. No subscription, no surprise fees.
Owner-operators tired of doing their own admin.
Small business owners
Drowning in email, bookings, invoices, and follow-ups. You want a buffer that handles the noise so you can focus on the work only you can do.
Sales-driven founders
You live in your CRM. Lead follow-up, meeting prep, and proposal drafting eat your week. You want that automated so you can sell more, not admin more.
Consultants and service providers
You bill for time you'd rather spend on clients. A personal AI layer handles the scheduling, invoicing, and onboarding in the background.
What people ask before booking.
How long does setup take?
A five-day engagement gets you live. Monday to Friday of week one, Brandon is embedded with you and your team. Half the week is discovery (mapping your workflows, tools, people, and bottlenecks). Half is building (wiring the assistant into the tools you already use, so it's live by the Friday demo). Then 90 days of support to tune what isn't landing and add what's missing.
What if I already use specific tools (HighLevel, Monday, Keap, etc.)?
We connect to most common small-business tools. If it has an API (a way for outside software to talk to it) and you use it often, we can wire it in. If we cannot connect to a specific tool, we tell you up front during the week-one engagement and refund if it is a blocker.
What happens after 90 days?
Two options. You can take it over fully using the handoff docs we leave you. Or you can extend to a monthly care plan for ongoing updates, new agents, and support. Pricing for the care plan is discussed once we see how you use the system.
What's the difference between this and the Obsidian Executive Assistant or Digital Worker?
Obsidian Executive Assistant is the privacy-first flagship: a local AI model runs on hardware we ship you, and nothing leaves your machine. For buyers who cannot send messages to a public cloud service. Digital Worker is the lightest version: sign up yourself, use our standard agents, and supply your own API keys (the logins that let tools talk to each other) using our guides. Digital Assistant sits between: done-for-you on a standard template running on Anthropic's Claude, with us setting up the hard parts.
Do you sign an NDA?
Yes. Mutual NDA before the engagement starts. Included at no extra cost. We also sign a data-handling agreement covering how your logins, API keys, and business data are stored and who has access.
What if it doesn't work?
If after 30 days of live use the system is not doing what we agreed it would do during discovery, we either fix it or refund the second payment. We have never had to refund.
Ready to see if this fits?
30-minute call. We learn your business. You learn whether this is the right move for you.
Book a 30-min discovery call