What is Opportunity Intelligence?

Why the category is emerging

Businesses already have enormous amounts of data. The problem is that someone still has to know what question to ask. Traditional analytics is reactive by construction: it answers questions a person already thought to pose. Most enterprise AI, including most "AI copilots," inherits the same limitation: it waits. Opportunity Intelligence exists because a growing share of a business's most valuable decisions are never identified in the first place, simply because nobody thought to look.

The category name is not exclusive to one company. Other organizations already use "Opportunity Intelligence" in adjacent senses: sales opportunity discovery, public tender discovery, grants and funding, general business-development signals. That existing usage is part of what validates the category rather than something Vink Intelligence invented; our aim is to publish one of the clearest conceptual frameworks for it, not to claim ownership of the term.

The four-way distinction

The clearest way to place Opportunity Intelligence is against the categories it sits next to:

Opportunity Intelligence does not replace the other three. It can sit above them and use their output as input. Read the full comparison with Decision Intelligence →

Signals

A signal is a normalized, source-independent observation: a usage decline, a contract renewal 53 days out, a cluster of negative-sentiment support tickets, a cost anomaly in cloud spend. Signals are the bridge between raw business data and opportunity detection: a usage_decline signal can come from a product analytics tool, a data warehouse, or a customer's own API. The detector reading it doesn't need to care which.

Company context

Signals only mean something in context. A usage decline alone might be noise; a usage decline combined with deteriorating support sentiment and an approaching renewal is a retention opportunity worth flagging. Opportunity Intelligence systems maintain a continuously updated model of each customer, product, and deal so detectors can reason about combinations of signals, not just one at a time.

Opportunity detection

A detector asks a narrow question: could something important be happening here? Detection is kept separate from validation and scoring. A detector's job is to notice a candidate and gather supporting evidence, not to decide how important it is.

Validation

Not every candidate a detector flags is real. Validation checks a candidate against corroborating signals, historical false-positive rates, and, where available, direct confirmation, before it reaches a scoring stage. This is what keeps an opportunity feed useful rather than noisy.

Scoring

Once a candidate opportunity is validated, scoring answers: how important is this? A typical formula weighs expected value, probability, urgency, and strategic fit against cost and risk, discounted by confidence. Different opportunity types reasonably use different scoring factors: a sales opportunity and a cost-saving opportunity are not scored the same way.

Evidence and provenance

Every opportunity a system surfaces should be able to answer "why are you telling me this?" with a clear chain back to the source records that produced it. Without provenance, an opportunity is an assertion, and assertions don't earn trust in an enterprise setting where a recommendation might drive real spending or outreach decisions.

Recommended actions

An opportunity is only useful if something can be done about it: drafting an email, creating a CRM task, opening a support escalation, updating a record. The action layer turns a scored, evidenced opportunity into a concrete next step, gated by the level of trust that action has earned: observe, recommend, assist, execute with approval, or, eventually, for the classes of action that have earned it, limited autonomy.

Outcomes and learning

Every opportunity should have a lifecycle: detected, shown to a person, accepted or rejected or ignored, acted on, and, eventually, an observed outcome. That accumulated dataset of context, opportunity, decision, and outcome is what lets a system learn which signals actually predict successful actions, rather than relying on a static rule set indefinitely.

Architecture

A well-built Opportunity Intelligence system separates concerns cleanly:

BUSINESS SYSTEMS
  → NORMALIZED BUSINESS CONTEXT
  → SIGNALS
  → OPPORTUNITY DETECTION
  → VALIDATION
  → SCORING
  → RANKED OPPORTUNITIES
  → RECOMMENDED ACTION
  → HUMAN / AGENT EXECUTION
  → OUTCOME
  → LEARNING

New data sources plug in without touching the detectors, and new opportunity types plug in without caring where the data came from.

Relationship to AI agents

Agents are becoming increasingly capable of executing well-defined tasks. That makes the upstream question more important, not less: which tasks are worth executing in the first place? Opportunity Intelligence is the layer that answers that question. It decides what deserves an agent's attention, so agents spend their growing capability on the actions that actually matter. Read how AI can detect opportunities automatically →

Examples

In practice, Opportunity Intelligence shows up as a ranked feed rather than a chat interface: a customer ready for an upsell, a support issue that risks a renewal, several customers independently requesting the same feature, an avoidable cloud cost pattern, a competitor weakness opening a window. Read our Asset Circles case study for what a closely related architecture looks like applied to energy and comfort management.

Opportunity Discovery: the service

We apply this architecture directly for clients through Opportunity Discovery, an expert-led service that maps a company's products, capabilities, and market environment to uncover evidence-backed opportunities for new customers, applications, and partnerships. It's powered by our proprietary Flock Discover research technology, paired with human review at every stage.