The AI cloud.
Where your agents go to work.
Anyone can get an agent working. Hosting it securely, governing what it is allowed to do, and managing a fleet of them as they grow is the part that takes a year and a team — and that part is ours. Yours is the knowledge, the judgment, and the product your customers pay for.
Every account starts with a conversation — so the first build is the right one.
Where are you starting from?
Cloud did it for software. We do it for AI.
Organisations stopped building data centres and put their engineering into what sets them apart. AI is at the same point. DoThat runs the platform underneath — models, security, governance, metering — so your teams build AI products, not infrastructure.
You used to buy servers, rack them, patch them and staff a rota to watch them — before you had written a line of your own product.
Now you run agents and agentic workflows without hosting a model, an index, or the governance around them.
The cloud did not make you a worse engineering team. It moved your engineers off the undifferentiated part.
Your teams build the AI products and solutions your customers pay for, not the isolation and metering underneath.
You stopped sizing hardware for a peak that might never come, and paid for what you used.
You set a ceiling per team and per agent, and know what each one spent.
The proof of concept took a fortnight
Making it safe enough to run took the rest of the year — and it still is not live.
Nobody can say who would see what
So nothing gets approved. The pilot stays a pilot and the business stops asking.
The bill has no ceiling
"It depends on usage" is not an answer anyone signs off.
Everything a demo leaves out.
Strict isolation between teams and clients. Role-based control over who can change what. Usage metered per team, so spend is visible before the invoice. Access granted when someone joins and revoked the day they leave. Prompt and response screening. Credentials held in a vault, never in a configuration file. An audit trail that answers “who changed that, and when”.
None of this is why your customers choose you. All of it is what an AI product needs before it can run in production — and it is in place before your first Goblin is built.
Your data is never used to train models
Agents that answer. Agentic workflows that do the work — planning, calling your tools, checking their own draft, and asking a person when it matters. Connected to the systems you already run, reaching people through the channels they already use.
One price covers the models, the cloud, the storage, the security team and the people who keep it running. Start with one, add the rest when it earns it.