Audit the AI your business is betting on.
Auditly is the independent AI audit and assurance platform, grounded in academic
research, that proves whether your AI systems are controlled, evidenced, tested and
operating within risk appetite.
} onClick={H.goToContact}>Book a demo
Get started
);
}
function MarqueeBand() {
return (
);
}
const WHY_CARDS = [
{ icon: "waves", title: "AI changes the control environment", body: "AI risk is dynamic and probabilistic. Systems that pass launch tests still drift, degrade, hallucinate, amplify bias and fail under adversarial pressure." },
{ icon: "scale", title: "Assertions are not assurance", body: "Boards and audit committees need independent assurance over AI governance, risk and controls, not management's word that controls exist." },
{ icon: "folder-check", title: "Trust runs on evidence", body: "Trust depends on evidence: registers, controls, testing, oversight logs and red-team results that are sufficient, reliable and traceable." },
];
function WhyAudit() {
return (
AI makes the decisions. Trust has to be proven.
AI systems influence decisions, process sensitive data and increasingly execute workflows.
If they can't be independently audited, no one can say whether AI risk is controlled, or
already outside appetite.
{WHY_CARDS.map((c, i) => (
{c.title}
{c.body}
))}
);
}
/* ── Pinned Assurance Model tour ── */
function OntologyPanel() {
const nodes = ["Business Use Case", "Audit Program", "AI System", "Auditing Algorithm"];
return (
);
}
function AssuranceModelTour() {
const stages = [
{ num: "01 / 03", title: "Ontology-driven audit structuring", body: "Auditly maps the AI system through connected ontologies. Business use case, audit program, AI system and auditing algorithm are structured into one audit file that links purpose, risks, controls, evidence and findings.", panel: },
{ num: "02 / 03", title: "AI risk appetite & control assurance", body: "Auditly tests whether AI risk-taking is deliberate, measurable, authorised and monitored, then links appetite thresholds to control requirements, evidence, lifecycle stages and assurance coverage.", panel: },
{ num: "03 / 03", title: "Automated oversight under pressure", body: "Continuous monitoring, adversarial stress testing and AI red teaming check whether systems hold up against real failure modes, prompt injection, drift, disclosure, unsafe tool use and agentic overreach.", panel: },
];
return (
The Auditly Assurance Model.
An assurance model anchored in academic research, built on ontologies, AI risk appetite,
risk and control mapping, automated oversight, and regulatory compliance assurance.
);
}
/* ── Differentiators ── */
const DIFF = [
{ icon: "flask-conical", title: "Research-based assurance logic", body: "Auditly translates evidence-based academic research into a practical AI audit model for enterprise assurance teams." },
{ icon: "network", title: "Ontology-driven audit files", body: "Every audit is structured around connected AI system, business use case, audit program and testing ontologies." },
{ icon: "route", title: "Risk appetite → control traceability", body: "Auditly links AI risk appetite to control expectations, evidence requirements, testing procedures, exceptions and conclusions." },
{ icon: "swords", title: "Automated oversight & red teaming", body: "Continuous monitoring, adversarial stress testing and AI red-team evidence are built into the assurance view." },
];
function Differentiators() {
return (
An audit engine built for AI.
{DIFF.map((d, i) => (
{d.title}
{d.body}
))}
);
}
/* ── Product screenshots: a look inside Auditly ── */
function AuditlyMark({ size = 22 }) {
return (
);
}
function AppWindow({ route, system, status, statusTone, capTitle, capMeta, children }) {
return (
AuditlyL3{route}{system}{status}
{capTitle}
{capMeta}
{children}
);
}
const ONT_NODES = [
{ code: "ONT-01", icon: "target", name: "Business Use Case", desc: "Purpose, value and the people each AI decision affects." },
{ code: "ONT-02", icon: "box", name: "AI System", desc: "Models, agents, data sources and vendors in scope." },
{ code: "ONT-03", icon: "clipboard-check", name: "Audit Program", desc: "Risks, controls, tests and evidence requirements." },
{ code: "ONT-04", icon: "workflow", name: "Auditing Algorithm", desc: "Procedures that generate findings and the opinion." },
];
const ONT_CHAIN = ["Purpose", "Risks", "Controls", "Evidence", "Findings", "Opinion"];
function OntologyShot() {
return (
);
}
const SCREENS = [
{
ey: "Ontologies", title: "Every audit is built on a connected model",
body: "Auditly structures each engagement as four linked ontologies. Because the business use case, AI system, audit program and auditing algorithm are modelled as related entities, purpose, risks, controls, evidence and findings stay traceable to one another across the whole audit file.",
pts: [["One linked file", "change a risk and the controls, tests and evidence it touches stay connected."], ["Full traceability", "every finding traces back to a use case and forward to an opinion."], ["No orphaned work", "nothing in the engagement sits outside the model."]],
shot: ,
},
{
ey: "Risk appetite", title: "Risk appetite, mapped to measurable thresholds",
body: "Each AI risk domain is tied to its enterprise appetite and a measurable threshold, then scored against live evidence. Auditly shows whether a system is within appetite, near threshold or in breach, so AI risk-taking is deliberate, authorised and monitored rather than implicit.",
pts: [["Enterprise-aligned", "every AI domain maps to an existing enterprise risk appetite."], ["Measured, not asserted", "thresholds are quantified KRIs scored from evidence."], ["Breaches surface early", "status escalates the moment a threshold is crossed."]],
shot: , flip: true,
},
{
ey: "Control effectiveness", title: "Controls assessed on three independent axes",
body: "A control that exists on paper isn't assurance. Auditly rates every applicable control on whether it is deployed, whether its design is effective and whether it is operating effectively, so partial deployment and design-only controls can never be mistaken for working ones.",
pts: [["Deployed isn't effective", "three separate verdicts, never a single tick."], ["Evidence-backed", "operating effectiveness is graded from sampled evidence."], ["Gaps become findings", "weak operation routes straight to a rated finding."]],
shot: ,
},
];
function ProductScreens() {
return (
A look inside the product.
Three of the screens assurance teams work in every day: how Auditly structures an audit,
tests risk appetite and grades whether controls actually work.
{SCREENS.map((s, i) => (
{s.title}
{s.body}
{s.pts.map((p, pi) => (
{p[0]}: {p[1]}
))}
{s.shot}
))}
);
}
/* ── Critical questions ── */
const QUESTIONS = [
"Could a single AI failure become tomorrow's reputation crisis?",
"Are high-risk AI systems operating within approved risk appetite?",
"Are AI controls designed, deployed and operating effectively?",
"Is human oversight meaningful, or just a rubber stamp?",
"Have these systems been adversarially tested before deployment?",
"Can we evidence our AI position to the board, customers and regulators?",
];
function CriticalQuestions() {
return (