ITHACA Autopilot: AI agents with memory, set up and run for you
AI is clever — but it does not know your business. We give it a lasting, governed memory and let it do real, recurring work — with approvals, an audit trail and one accountable partner: us.
- A lasting, readable memory of your business — yours to keep
- Agents for recurring work: quotes, follow-ups, reports, document filing
- Approval before every irreversible action (email, invoice, post)
- Audit trail: who asked, which AI did it, what changed, from which source
- Backup and restore with a preview
- Works with the AI you already use (Microsoft Copilot, Anthropic Claude)
- A fixed monthly fee and one accountable partner — us
Most people use AI in bursts: they open a chat, explain their business from scratch, get an answer, and next time explain it all over again. There is no memory, no record, and nobody dares let the AI touch real things — email, invoices, posts.
ITHACA Autopilot changes that. We set up a lasting, readable memory of your business (what it is, how it works, what its rules are) and put agents on top of it that do the recurring work. We run and supervise them, every irreversible action goes through approval, and every step is logged. You own the result and the memory; we own the responsibility. We already run ITHACA itself the same way.
Why AI on its own is not enough
It does not remember
Every conversation starts from zero. Your business knowledge stays in your head, not in the tool.
It does not account for itself
Without a record, you do not know who asked what, which model did it and what changed. Hard to trust with real work.
You are afraid to give it access
Letting a chatbot send an email or issue an invoice unchecked is a risk. That is why we add approvals and boundaries.
Three packages, by how much you want it to run on its own
All on a fixed monthly fee, with an initial setup. The final price is given in a written quote after a short baseline.
Memory
We set up your business's governed memory and knowledge base, with one or two agents (e.g. a knowledge assistant, inbox triage). Monitoring and small changes.
Operations
Several agents on real workflows: quotes, follow-ups, reports, document filing, with approvals and a monthly report. Connections to your tools (e.g. Microsoft 365).
Autopilot
Agents across several areas, full governance and audit, recovery, and a quarterly "AI governance" review with management. For those who want a large part to run on its own, safely.
How we keep it safe
Approvals where it counts
Sending, publishing, paying or deleting wait for your approval. The AI proposes; you decide.
Boundaries per task
Each agent has a clear scope: what it may touch. Anything outside it is proposed, not done.
Instructions hidden in documents are not obeyed
Commands hidden in emails or documents are reported, not obeyed — a defence against malicious content.
Your data stays yours (GDPR)
With our ISO 27001 and NIS2 background, we set access boundaries and transparency over where data goes.
Restore with a preview
Before bulk changes we keep a checkpoint; you see what would change before it happens, and every restore is logged.
How we work
Baseline
We understand how your business works, which recurring tasks eat time and where AI genuinely helps.
Set up memory & agents
We build the memory, the rules and the first agents, with approvals and boundaries from the start.
Monthly running & oversight
We run them, monitor them, improve them and add new tasks as it matures.
Reporting
A short picture: what the agents did, what you saved, what needs a decision from you.
Frequently asked questions
What is ITHACA Autopilot?
It is a managed service: we set up a lasting, governed memory and AI agents that do recurring work for your business, and we run and supervise them on a monthly basis.
Do I need technical knowledge?
No. We set it up, run it and explain it in plain language. You give direction and approve the important things.
Is it safe to let AI send emails or issue invoices?
Every irreversible action goes through your approval — the AI proposes it, you approve it. Every step is logged and there is a restore.
What about my data?
The memory is yours and readable. We set access boundaries and transparency, with our ISO 27001 and NIS2 background, so data stays under control and in line with the GDPR.
Which AI does it work with?
With the AI you already use or that suits you — mainly Microsoft Copilot and Anthropic Claude. We do not lock you into one vendor.
How is it charged?
With a one-time setup and a fixed monthly fee, depending on scope. We agree it in a written quote after a short baseline.
Ready to see where your business stands?
A short first conversation, no commitment.