Long 269 words · 1491 characters · profile or about page
RiskMandate creates and sells Agent Behaviour Policies.
An Agent Behaviour Policy is a working record, for one agent in one deployment, of four things: everything it can actually reach, what you authorised it to do, the gap between those two, and what stands in the way. Three of the four can be written down by the people who run the agent. The fourth, the gap, is worked out from the first two, and it is usually the one nobody has looked at.
It is a portable vault rather than a document, because four audiences need different things from one record. The operator needs instructions, leadership needs a decision, security needs evidence, and an insurance conversation needs explicit scope and open questions. One record, several views, with a named owner, a review trigger and every version kept.
A policy is a draft, and it is meant to be argued with. It goes to the people who built the agent, the people who own what it touches and the people accountable for it, and each corrects the part they know. It is never finished, because the agent is not finished.
The question it answers is not what model somebody bought. It is what the agent can do that nobody in the business has fully understood yet. That is not a property of the model. It is a property of this deployment, so a policy is fitted to a twin of the environment the agent is actually in. That is where a generic shape stops being generic and becomes yours, and it is the only point at which the exposure means anything.
The full version. The third paragraph is what is being sold rather than a description of it, and the last is the one that matters: risk is a property of the deployment, not the model.
Medium 85 words · 462 characters · directory or listing entry
RiskMandate creates and sells Agent Behaviour Policies: for one agent in one deployment, a working record of everything it can actually reach, what you authorised it to do, and the gap between the two. It comes as a portable vault rather than a document, with a view for each audience, a named owner and every version kept. Risk is not a property of the model but of this deployment, so the record is fitted to a twin of the environment the agent is actually in.
The one to send when no field limit is visible. Safe under almost any cap and reads as finished rather than truncated.
Profile 73 words · 396 characters · a company profile with a 400 cap
Create or buy Agent Behaviour Policies: for one agent in one deployment, what it can actually reach, what you authorised it to do, and the gap between the two. It arrives as a portable vault with a view for each audience, a named owner and every version kept. Risk is a property of the deployment rather than the model, so each one is fitted to yours. Open source, and sixteen are published free.
Cut to fit Web Summit’s 400-character company overview on 24 September 2026, which the 462-character version above overshoots. It is the medium one with the drafting dropped and the reader’s verb at the front, and it keeps the sentence about the twin, because that is still the half nobody else is saying.
Short 45 words · 258 characters · a card, a tile, a panel
RiskMandate creates and sells Agent Behaviour Policies. For one agent in one deployment, a portable record of everything it can actually reach, what you authorised it to do, and the gap between the two, fitted to a twin of the deployment it actually runs in.
Where there is room for a sentence and no more. Gives up the vault and the drafting, and keeps the twin, because that is the half nobody else is saying.
One line 31 words · 180 characters · strapline or badge
RiskMandate creates and sells Agent Behaviour Policies: for one agent in one deployment, what it can actually reach, what you authorised it to do, and the gap nobody has looked at.
A strapline field, and what this page called the floor until a form refused it: Web Summit’s exhibitor description rejected 180 characters on 23 September 2026. The two below exist because of that. They drop the company name rather than the claim, because a directory has already asked for the name in the field above.
Directory 65 words · 395 characters · a conference listing, written to start a conversation
Create or buy Agent Behaviour Policies: for one agent in one deployment, what it can actually reach, what you authorised it to do, and the gap between the two. Delivered as an encrypted vault with stakeholder views, fractal semantic graphs, prompts and toolkit. Risk is a property of the deployment rather than the model, so each one is customised. Open source with multiple commercial services.
The lead’s own line, and the only one written to be asked about rather than to be complete. It opens on what the reader does, and it names four things a technical reader will want to push on:
stakeholder views,
fractal semantic graphs,
prompts and
toolkit. Two of those run today and two are drawn;
the page they land on says which is which, node by node, so the curiosity has somewhere to go. Use
or rather than
and: it is a choice between two ways in.
Badge 20 words · 110 characters · a field with a hard cap
Agent Behaviour Policies: what one agent can reach, what you authorised it to do, and the gap between the two.
The shortest one that still carries all three objects and the verb on each. Reach for this first when a counter is ticking down.
Chip 12 words · 80 characters · the floor
Agent Behaviour Policies: what one agent can reach, against what you authorised.
Two of the three objects, with the third implied by against. Below this the sentence stops being a claim and becomes a category, which is the line the one above used to hold.
None of these names a customer, a funder or an insurer, and none attaches an adjective to
anybody else’s product. Every sentence is a definition of one of our own terms or a
statement about our own work. British spelling throughout, and no dash in any of them.