Morse AI
// one assistant, your organisationOne AI assistant for everyone in your organisation, behind your own sign-in and on your own domain. Nobody needs a personal account with a model vendor, and nobody pastes work into a consumer chatbot to get it done. Your conversations stay on your infrastructure, and where the model itself runs is your decision to make.
Your people are already using AI. The only question is whether they're doing it on your terms, or pasting client data into a browser tab you can't see.
Morse AI gives them one place to do it: sanctioned, branded, signed in with the account they already have, and logged where you can see it.
It runs in the Morse Workspace suite alongside mail, chat, documents and meetings, under the same identity, or on its own.
It is also where the rest of the suite's AI is heading: one gateway every request goes through, so there is a single place that knows what was asked, of which model, by whom, and at what cost.
Shadow AI, minus the shadow
The problem isn't that staff want an assistant. It's that the good ones live outside your walls, need a personal account, and quietly keep whatever gets pasted into them.
One door, the account they already have
- Sign in with your existing identity provider: no new account, no invite sprawl
- No personal API keys to hand out, rotate, or find on someone's laptop after they leave
- Sign-up is closed: access follows your directory, so leavers lose it when they lose everything else
- Served from your own domain, as an ordinary internal tool
The conversations stay yours
- Chat history lives on your infrastructure, not in a vendor's account
- Search across past conversations, indexed on your own systems
- Nothing is retained by a consumer product with its own ideas about training data
- When someone leaves, their conversations are yours to keep or delete
Unmistakably your toolWhitelabel
- Your name, logo and branding throughout, with no vendor's product showing through
- Model choices appear as plain names your staff understand, not vendor SKUs
- Sits next to your other apps as one more thing the company provides
- Adoption is the point: a tool that looks sanctioned gets used instead of the browser tab
Good at the boring things
- Drafting, rewriting, summarising, translating: the daily grind of written work
- Explaining a contract clause, a config file, or an error nobody recognises
- A capable frontier model by default, with a faster one for quick questions
- Deliberately plain: a chat box that works, not a wall of controls
- The platform underneath does far more, agents, chains, tools, code execution. It stays out of the way until you want it, and turning it on is a setting, not a rebuild
Where the model runs is a decision, not a default
This is the part most AI products don't let you have an opinion about. The interface, the identity and the history are already yours. The inference is the last piece, and it's the one your posture should decide.
Run the model yourselfOn the roadmap
- Point it at a model running in your own environment and no prompt ever leaves your estate
- The route in is an open, standard model API, so a local model is a configuration change
- The path for air-gapped and classified environments, where nothing else will do
- The trade is candid: today's self-hostable models are good, not frontier, so you choose which matters
Or bring your own provider
- Use your own account with a model vendor, on your commercial terms
- Business tiers, not consumer ones: no-training and no-retention terms, in writing
- Frontier capability where the work needs it and the material allows it
- Your contract, your data-processing agreement, your relationship
Both, split by sensitivityOn the roadmap
- Route by classification: routine work to the capable model, sensitive work to the local one
- Material above a level you set never reaches an outside provider, or is refused outright
- The same protective markings the rest of the suite already understands
- Enforced server-side, so the rule holds regardless of what a client asks for
Deployed how you deploy everything else
- On-premises, in your cloud, or fully managed by us
- Containers, your orchestration, your network, your certificates
- No phone-home, no vendor telemetry, no usage reporting back to us
- Runs standalone, or inside Morse Workspace under one identity
Self-hosted models and classification-aware routing are on the roadmap. If your posture needs them before you adopt, tell us and we will sequence accordingly.
What runs in your tenant
Four moving parts, all on your infrastructure. The model is the only thing that might not be, and that is the decision above.
Deployable on-premises, in your cloud, or fully managed. No telemetry, and no usage reported back to us.
Where it's going
A sanctioned assistant today; a sovereign one next. What's shipped is on the left; we mark the rest as intent, not inventory.
Now
- Unified AI chat for the whole organisation, on your own domain
- SSO sign-in through your identity provider; no personal accounts or API keys
- Conversation history and search held on your infrastructure
- Whitelabelled throughout, with neutral model names
- Runs standalone or inside Morse Workspace under one identity
In progress
- Self-hosted models: point it at inference in your own environment, so no prompt leaves your estate
- Bring your own provider account, with no-training terms enforced
- Per-tenant isolation: separate instance, credentials and data per organisation
Sanctioned AI, inside your own perimeter.
or fully managed.
Your identity, your history, your call.