Operations
Operations assistant on WhatsApp and email
Watches alarms, readings and daily numbers, sends alerts and morning summaries to the team on WhatsApp and email, and answers follow-up questions in the same thread.
We build AI that does real work for your business.
From a single assistant to a team of agents, we design, build and run AI systems around your data and your tools, reachable on the web, WhatsApp and email. Our own platform, M-Drive, keeps them on your servers.
- You: Every weekday at 9 am, send the team the invoices that are still unpaid.
- Agent step: Read the invoices sheet
- Agent step: Created a routine: Monday to Friday, 9:00 am
- Agent: Unpaid invoices today Overdue more than 30 days2 Due this week5
- Agent: Done. It will post this list to your team's WhatsApp group every weekday morning.
At a glance
- AI systems built
- 5
- Building operations, finance, support, software and documents
- Model families
- 4
- Gemini, Claude, OpenAI and open models run locally
- Ways to reach agents
- 5
- Web, Android app, WhatsApp, email and voice
- Hosting
- Yours
- M-Drive runs on your own Linux or macOS server
What we build
We take on any AI or AI-agent work, from a quick prototype to a system your team uses every day.
-
AI agents
Agents with their own role, memory and tools that research, answer, draft and follow up, alone or as a team that hands work between specialists.
-
RAG and knowledge assistants
Assistants that answer from your documents, manuals and policies, and show the source so people can check the answer.
-
Automation and integrations
Scheduled and event-driven workflows connected to WhatsApp, email, spreadsheets, APIs and the systems your team already uses.
-
AI products and apps
Web and mobile apps with AI built in, from chat and voice interfaces to dashboards that explain what the data means.
-
Model training and fine-tuning
Models trained or fine-tuned on your data for classification, prediction and domain-specific language, with evaluation before anything ships.
-
Advice and pilots
Help deciding where AI fits and where it does not, then a small working pilot on real examples before a larger commitment.
What we have built
A few systems we have designed and run. Client names are left out on purpose.
Finance
Trading research agent
Pulls live market data, runs a trader's own option strategies and backtests, and sends buy and sell alerts to WhatsApp. It never places orders.
Customer support
Multi-agent help desk
A coordinator agent reads each request and passes it to billing or technical specialist agents, then replies with one combined answer.
Software
Developer agent
Works through a coding task in a project: reads the code, makes the change and runs the tests, then hands it to a person to review.
Knowledge
Document assistant with sources
Answers questions from a library of documents and shows which passage each answer came from.
Our product
M-Drive
One place to run all your AI agents, on your own server.
Many agents, one place
Each agent has its own role, instructions, memory and tools. Your team works with them from the browser or the Android app.
- Your choice of model: Gemini, Claude, OpenAI or local open models
- Knowledge base: agents answer from the documents you give them
- Quality checks: review how agents answer and spot weak replies
Replies on WhatsApp and email
Agents work where your team already does. They post updates, answer questions and reply to messages on WhatsApp and email, on a timetable or as messages arrive.
- WhatsApp groups and chats, including voice notes
- Email: reads and replies, in Gmail or other email accounts
- Scheduled routines, such as a 7:00 am daily summary
Example: the same operations assistant on WhatsApp (left) and email (right)
Your server, your rules
Conversations, documents and API keys stay on your own Linux or macOS server.
- Roles and access: decide who can see, use and change each agent
- Backup and restore: move an agent with its memory and history
- Install on your own hardware or cloud machine
How a project runs
Small steps, each one something you can see and try.
Discover
You describe the task and what a good result looks like. If AI is not the right fit, we say so.
Pilot
We build a small working version and test it on your real examples.
Build
We connect it to your data and systems, and set up access, schedules and alerts.
Run and improve
It goes live on M-Drive or your own stack, and we tune it from real use.
Common questions
Anything else, just ask.
Does our data leave our servers?
Not with M-Drive self-hosted. Your documents, conversations and memory stay on your server. If you use a hosted model such as Gemini or Claude, the text of each request goes to that provider; with a local open model, nothing leaves.
Which AI models do you work with?
Gemini, Claude, OpenAI models and open models run locally. We pick per task, and you can switch later without rebuilding the agent.
Can agents reply on WhatsApp and email?
Yes. Agents can post to and reply in WhatsApp groups and chats, and read and reply to email, in Gmail or other email accounts. You decide which groups and inboxes each agent may use.
Do we need M-Drive to work with you?
No. We also build agents, RAG systems and models that run inside your own applications and infrastructure.
How does a project start?
Send us a short email about the task. We reply with questions, then usually suggest a small pilot so you can judge the result on real examples.
Can you train a model on our data?
Yes, when it is the right tool. Often retrieval over your documents works better and costs less, and we will tell you which one fits.
Have a task in mind?
Send us a few lines about it. A short description is enough to start.
Email us about your taskOr write to mathan@km-ts.com