Adaptive CX Glossary

The working vocabulary of Adaptive Customer Experience. Each term is defined as Kairos uses it in practice — in diagnostics, behaviour specs, and client engagements. Definitions are deliberately short so they can be lifted, cited, and shared.

Last updated: April 2026

Core concepts

Adaptive Customer Experience (Adaptive CX)

Adaptive Customer Experience is a design approach where services respond to what is actually happening for each customer — using real-time signals, defined decision rules, and calibrated levels of autonomy — rather than following a single assumed journey. It turns static flows into experiences that change behaviour when conditions change.

Static CX

Static CX is the traditional model of customer experience, where journeys are mapped once, flows are fixed, and every customer is routed through the same ideal path. Static CX works when conditions are stable. It leaks value when customers, context, or systems diverge from the assumed path.

Related: Adaptive CX · Moment

Moment

A moment is a specific point in a service where a customer's situation changes and the experience needs to respond. Moments are the unit of design in Adaptive CX — smaller than a journey, larger than a touchpoint. A moment can be moved from static to adaptive without re-platforming the whole service.

Related: Adaptive Moment · Moment Brief

Adaptive Moment

An adaptive moment is a moment that has been re-designed to respond to real-time signals at a calibrated level of autonomy, with a defined fallback and governance model. Adaptive moments are the smallest shippable unit of Adaptive CX work.

Related: Moment · Behaviour Spec

Service Intelligence

Service Intelligence is the capability a service has to recognise what is happening for a customer, decide what is true enough to act on, and respond usefully. It sits across data, service design, and operations — not just one of them. Service Intelligence is what makes a service adaptive rather than reactive.

The Adaptive CX Model

The Adaptive CX Model is a four-stage continuous loop — Detect, Decide, Respond, Learn — that replaces fixed journey logic with a repeatable operating pattern. Each stage has its own artefacts and its own failure modes. Moments can enter the loop at any maturity level and progress as signal and behaviour mature.

Related: Detect · Decide · Respond · Learn · Full framework

Detect

Detect is the stage where a service reads real-time signals to understand what is happening for a customer. Good detection uses signals that are reliable, timely, and specific. Assumed personas and survey data are weak detection. Behavioural events and system-of-record truth are strong detection.

Related: Signal · Signal Spec · Signal Maturity

Decide

Decide is the stage where a service evaluates the detected signal against confidence thresholds and decides whether to act, how to act, and at what level of autonomy. The Decide stage is where governance lives — it is where 'we know enough' becomes 'we are allowed to act'.

Respond

Respond is the stage where the service changes its behaviour in the moment — through content, channel, routing, automation, or a human handover. A good response is proportional to the signal strength and matches the customer's current need, not their historical segment.

Related: Behaviour Spec · Fallback Path

Learn

Learn is the stage where every resolved moment feeds back into the system, sharpening signal quality, validating decision rules, and unlocking higher autonomy over time. Without Learn, adaptive services plateau at their launch configuration and lose their edge against changing customer behaviour.

Related: Signal Maturity · Audit Gate

Maturity

Service Intelligence Maturity Model

The Service Intelligence Maturity Model places any service moment on a grid with two axes: signal maturity (what the service knows) and behaviour maturity (how the service responds). The grid exposes mismatches — for example, rich data paired with static behaviour — and sets the next credible move rather than jumping straight to full autonomy.

Signal Maturity

Signal maturity describes how reliably a service can read what is happening for a customer in the moment. It progresses from assumed (personas, segments) through observed (behavioural events) to integrated (cross-system truth in real time). Signal maturity caps what a service is safely allowed to do.

Related: Behaviour Maturity · Truth Gate

Behaviour Maturity

Behaviour maturity describes how a service responds when a signal is detected. It progresses from static communication through rules-based variation to adaptive automation with learning. Behaviour maturity must be earned by signal maturity — responding strongly on weak signals is how AI CX programmes lose trust.

Autonomy

AI Autonomy Levels

AI autonomy levels describe how much control a service gives an AI system in a moment. Kairos uses four levels — Observe, Advise, Assist, Act — to calibrate autonomy against signal strength, business risk, and recovery capability. Most AI CX failures come from skipping levels rather than from the technology itself.

Related: Observe · Advise · Assist · Act · The Four Gates

Observe

AI reads silently. No customer-facing action.

Advise

AI recommends to a human operator.

Assist

AI drafts. Human reviews before sending.

Act

AI executes. Governed by thresholds and audit.

Observe

Observe is the lowest autonomy level. The AI reads signals and logs context but does not surface anything to the customer or the agent. Observe is where most services should start — it builds signal confidence before any customer-facing risk is taken.

Related: Signal Maturity · Advise

Advise

Advise is the second autonomy level. The AI surfaces an insight, score, or recommendation to a human operator, who decides whether to act. Advise keeps the customer experience unchanged while the AI is tested against real decisions.

Related: Observe · Assist

Assist

Assist is the third autonomy level. The AI drafts, proposes, or suggests a response that a human can accept, edit, or override before it reaches the customer. Assist is the highest level of autonomy compatible with tight human oversight on every interaction.

Related: Advise · Act

Act

Act is the highest autonomy level. The AI takes an action — changing content, routing, messaging, or executing a transaction — without per-instance human review. Act is governed by sampling, thresholds, fallback rules, and audit trails rather than approval on every decision.

Related: The Four Gates · Fallback Path · Audit Gate

Artefacts

Moment Brief

The Moment Brief is the short document that frames a single adaptive moment before design work begins. It captures the customer situation, the business risk, the outcome the service is trying to achieve, and the constraints. It is the input to the Signal Spec and Behaviour Spec.

Signal Spec

The Signal Spec defines exactly what the service needs to detect to act in a moment — which signals, from which sources, with what confidence threshold, and with what handling when signals conflict or are missing. The Signal Spec makes 'we have data' specific enough to engineer.

Related: Signal · Detect · Truth Gate

Behaviour Spec

The Behaviour Spec (or Service Behaviour Spec) defines how the service responds when the Signal Spec fires — the response, the channel, the tone, the autonomy level, the fallback, and the stop rules. It is the bridge between service design and platform build.

Platform Translation Pack

The Platform Translation Pack maps a Behaviour Spec onto the specific platforms a client uses — for example, Bloomreach, Agentforce, Gemini, or Vertex. It tells the engineering team which platform capability delivers which part of the specified behaviour, so designers do not have to learn every platform.

Related: Behaviour Spec

CX Agent Card

A CX Agent Card is a one-page specification for a single AI agent in a service — defining the intervention moment, customer signals, psychology at play, customer value, governance, business value, and brand warnings. Agent Cards make AI roles designable and auditable at the moment level.

Related: Behaviour Spec · The Four Gates

Governance

The Four Gates

The Four Gates are the governance checks that determine whether a moment is safe to operate at a given autonomy level. The gates are Truth, Safety, Recovery, and Audit. Moments that pass all four can scale autonomy. Moments that fail any gate are capped until the gap is closed.

Related: Truth Gate · Safety Gate · Recovery Gate · Audit Gate

Truth Gate

Are the signals reliable enough to act on?

Safety Gate

Is the blast radius of a wrong action contained?

Recovery Gate

Can the service detect and correct a bad action?

Audit Gate

Can the service explain itself after the fact?

Truth Gate

The Truth Gate asks whether the signals feeding the decision are reliable enough. It checks signal provenance, freshness, completeness, and confidence. A failing Truth Gate means the service is at risk of acting on weak or stale data, so autonomy is capped at Observe or Advise.

Related: Signal · Signal Maturity

Safety Gate

The Safety Gate asks what happens when the service acts wrongly. It checks whether the blast radius of a bad action is contained — by segment, by channel, by reversibility, or by sampling. A failing Safety Gate means the cost of an error outweighs the value of autonomous action.

Recovery Gate

The Recovery Gate asks whether the service can detect and correct a bad action. It checks fallback paths, human handover quality, and customer-side recovery — for example, can the customer easily reach a human or reverse an action? A failing Recovery Gate means errors compound instead of being absorbed.

Related: Fallback Path · Safety Gate

Audit Gate

The Audit Gate asks whether the service can explain itself after the fact. It checks whether every automated decision is logged with its inputs, thresholds, and outcome, in a form an auditor or a regulator would accept. A failing Audit Gate is a regulatory risk regardless of in-market performance.

Related: Learn · The Four Gates

Activation

Activation Readiness

Activation Readiness describes whether an organisation has the operational capability to ship an adaptive change once it has been designed. Readiness spans three surfaces: communication (content and messaging), interaction (channels and touchpoints), and action (systems of record and fulfilment). Weak readiness caps design ambition.

Related: Activation Audit

Activation Audit

An Activation Audit is a short diagnostic that assesses readiness across the three activation surfaces. It surfaces where the organisation can ship adaptive change today, where it cannot, and what the minimum credible first move looks like. Its output is the Activation Roadmap.

Activation Roadmap

The Activation Roadmap is a 90-day plan that sequences the work needed to move from current readiness to shipping a first adaptive moment. It prioritises low-tech interventions where signal or platform capability is weak, rather than waiting for a full re-platform.

Supporting terms

Signal

A signal is any piece of data that tells the service something about the customer's current situation. Signals range from the weak (persona, segment) to the strong (behavioural event, system-of-record status). A signal's value in Adaptive CX is judged by reliability, timeliness, and specificity.

Related: Signal Spec · Detect

Fallback Path

A fallback path is the defined route a service takes when an adaptive decision cannot be made confidently — for example, when a signal is missing, a threshold is not met, or a model is unavailable. Every Behaviour Spec has an explicit fallback. Unspecified fallbacks become unplanned outages.

Related: Behaviour Spec · Recovery Gate

Service Intelligence Delivery Loop

The Service Intelligence Delivery Loop is the 10-step sequence Kairos uses to take a moment from discovery to deployment — from Moment Brief through Signal Spec, Behaviour Spec, Platform Translation Pack, to proof and iteration. It operationalises the Detect-Decide-Respond-Learn model.

Frequently asked questions

What is Adaptive CX in simple terms?

Adaptive CX is the practice of designing customer experiences that change based on what a customer actually needs in the moment, using real-time signals rather than fixed journeys. The service senses, decides, responds, and learns — rather than following the same script for every customer.

How is Adaptive CX different from personalisation?

Personalisation usually varies content for a known segment or profile. Adaptive CX varies behaviour — content, channel, autonomy level, and fallback — based on what is happening in a specific moment. Personalisation is a subset of Adaptive CX, not a synonym for it.

What are the four stages of the Adaptive CX Model?

Detect, Decide, Respond, Learn. The service detects a real-time signal, decides whether and how to act, responds through content or automation, and feeds the outcome back in to sharpen future decisions.

What are the Four Gates of AI CX governance?

Truth, Safety, Recovery, and Audit. Each gate is a check that determines whether a moment is safe to scale to a higher autonomy level. Moments that fail any gate are capped until the gap is closed.

What are the four AI autonomy levels in Adaptive CX?

Observe, Advise, Assist, Act. Observe is the AI reading signals silently. Advise is the AI recommending to a human. Assist is the AI drafting for human review. Act is the AI executing without per-instance approval, governed by thresholds and audit.

What is a Behaviour Spec?

A Behaviour Spec is the document that defines how a service responds in an adaptive moment — the response, the channel, the tone, the autonomy level, the fallback, and the stop rules. It is the bridge between service design and platform build.

Put the vocabulary to work

The Adaptive CX Framework shows how these terms fit together in practice. Start with a Reality Check to see where your service stands.

Book a Reality Check Explore the framework