The platform for commercial intent from humans and AI agents.
The Intent Graph is the platform. It connects behavioural signals, agent-declared purpose, identity, policy scope, forecasts, decisions, actions, outcomes and causal audits without flattening humans and agents into the same object.
Not a warehouse. Not a dashboard. Commercial identity, purpose and outcome context.
Most teams store activity in separate tools: ads in platforms, accounts in CRM, orders in commerce, behaviour in analytics, and agent requests in workflow systems. The Intent Graph joins those pieces into a living map of who or what is acting, what they appear to want, what is likely to happen next, what the business should do, and whether that action later created value.
Know who or what the signal belongs to
A page view, ad click, CRM update, order, support request, or agent action is not useful on its own. The graph attaches it to a workspace-scoped person, visitor, account, campaign, product, transaction, experiment, agent, principal, mandate, or policy scope.
Fuse behaviour with declared purpose
Humans may reveal intent through ambiguous behaviour. Agents may arrive with structured purpose, delegated authority, and a requested action. Levered treats these as different signal classes in one graph.
Attach probability, horizon and support
The graph carries the forecast, confidence interval, effective segment n, model version, data freshness and expiry window that decide whether a recommendation is eligible.
Make the next action explicit
Every traversal can end in a decision: scale, cut, hold, wait, retest, route, approve, challenge, reject, or insufficient evidence.
What the graph connects
Every node carries commercial meaning. The value is not in storing the nodes; it is in understanding the relationships between them and updating those relationships as outcomes arrive.
Visitor, lead, user, customer, or buyer identity.
Workspace-scoped visitor before known identity resolution.
Buyer organisation, client, pipeline opportunity, or workspace customer.
AI shopping, research, reporting, routing, checkout, or commerce agent.
The party represented by an agent, where applicable.
Delegated authority, declared purpose, scope, and policy basis.
Allowed scope, consent state, or delegated authority.
Meta, Google, TikTok, email, lifecycle, affiliate, or partner stimulus.
Ad, angle, hook, landing page variant, offer, or message.
Plan, SKU, service, category, margin profile, or offer.
Country, city, store catchment, sales territory, or treatment market.
Page view, click, checkout, purchase, demo request, CRM update, or agent request.
Interpreted event such as pricing intent, comparison, urgency, risk, or declared purpose.
Probability, horizon, expected time-to-outcome, interval, support, and freshness.
Scale, cut, hold, wait, retest, route, approve, challenge, reject, or insufficient evidence.
Revenue, margin, CAC, ROAS, pipeline, conversion, retention, or payback.
OpenLift grade of whether acted-on recommendations created incremental value.
The object that turns activity into a forecastable decision.
Every recommendation starts from an inspectable intent object. It carries actor, purpose, policy scope, forecast, support, model version, freshness and expiry instead of hiding the decision behind a score.
Human, account, campaign or agent.
Party represented by the actor, where applicable.
Declared or inferred commercial purpose.
What the actor appears to want done.
Probability, horizon and expected time-to-outcome.
Uncertainty calibrated to support and data quality.
Effective sample size at recommendation granularity.
Scale, cut, hold, wait, retest, route, approve, challenge or reject.
Intent is not static. It strengthens, weakens, compounds, and decays.
A useful graph has to behave like the market. It should understand timing, repetition, proximity to money, declared purpose, policy scope, commercial constraints, and feedback from outcomes.
Fresh signals carry more weight. A pricing page visit today matters more than a whitepaper download six months ago.
Repeated intent across channels is stronger than one isolated event. The graph compounds patterns across web, CRM, ads, commerce, and agents.
Signals close to money, such as checkout, demo request, quote request, or stock check, are weighted differently from early research.
A channel claiming credit is not enough. Levered can connect signals to OpenLift evidence so the business knows whether activity caused incremental value.
The graph understands budget, stock, margin, sales capacity, region, campaign limits, and policy scope before recommending action.
Outcomes flow back into the graph. When a recommendation succeeds or fails, the next decision becomes better calibrated.
The economics the graph unlocks
The Intent Graph is valuable because it connects commercial behaviour to money. It helps teams decide where to spend, what to route, when to hold, which markets to test, and which actions should be allowed.
Spend moves faster
Shift budget toward campaigns, regions, creatives, and audiences with rising intent and evidence of incremental lift.
Efficiency becomes explainable
Understand whether CAC is improving because demand quality is changing, creative is working, or a platform is simply over-crediting itself.
Sales works the right accounts
Route accounts when website behaviour, CRM movement, fit, and timing indicate readiness instead of relying on static lead scores.
Growth respects profit
Tie product demand to margin, stock, return rate, region, and discount pressure before recommending scale or promotion.
Automation gets boundaries
AI agents can request reports, pricing, routing, or checkout actions, but the graph checks purpose, principal, scope, evidence, and policy.
Meetings become decisions
Replace weekly dashboard interpretation with live recommendations that show the signal, evidence, economics, and proposed action.
How questions become answers
Every recommendation starts as a traversal through the operating loop. The graph stores relationships such as visited, clicked, purchased, represents, delegated by, requested, permitted to, forecasted as, recommended, acted on, caused and audited by.
Decisions businesses can see
Levered is designed to make the output plain enough for a team to act on and detailed enough for finance, leadership, and operators to understand probability, support, uncertainty, risk and auditability.
Scale London prospecting by +£800/day because treatment markets show Tier 3 lift, intent is rising, and CAC remains inside target.
Route Acme Corp to sales today because pricing intent, return visits, account fit, and product engagement crossed the readiness threshold.
Hold discounting on Product B because demand is rising organically, stock is limited, and margin would fall below the goal constraint.
Approve a scoped report-generation action because the agent has a verified principal, a narrow purpose, and supporting evidence.
Businesses do not need more disconnected data. They need a better decision surface.
Marketing sees which spend is creating incremental value, not just which platform claimed the conversion.
Sales sees which accounts are genuinely warming up and why now is the right moment to act.
Commerce sees demand, margin, stock, and campaign pressure in the same decision path.
Operations can let AI agents act inside boundaries instead of giving them broad, risky access.
Leadership gets a clearer line from signals to economics to decisions, which makes performance easier to defend.