Dual CX — customer service when both sides have AI
Customer service has two designers now: the brand, and the customer's AI. Dual CX is the framing for what that means — four interaction modes, three design dimensions, and the gaps most companies haven't designed for.
Dual CX is the framing for customer service when both the customer and the brand have AI in the loop. Until now, brands have been the sole designers of the customer experience. With consumer AI agents now contacting service channels on customers' behalf, customer service has two designers — the brand and the customer's AI. Dual CX names the design problem this creates. Coined by Neos Wave in 2026. The Dual CX framework is the basis of the open Service Handshake standard.
The four interaction modes
Every service interaction now sits in one of four modes — defined by who's on each side of the conversation.
Customer (human) ↔ Brand (human)
The model we know.
Declining as a percentage of all interactions. Risk: over-investing here as the volume shifts.
Customer (human) ↔ Brand (AI)
The model most companies are building.
Chatbots, IVR, self-service. Risk: containment over resolution — measuring deflection, not outcomes.
Customer (AI) ↔ Brand (human)
The model almost nobody is designing for.
AI assistants calling your contact centre. Verification fails. Patience assumptions break. Tone-dependent steps stop working.
Customer (AI) ↔ Brand (AI)
The model that changes everything.
Two AIs negotiate outcomes directly. No human in the loop. Policy gaps exposed instantly. Needs an explicit protocol.
14% of UK consumers already use AI agents to interact with brands — projected to reach 37% by end of 2026 (Braze, 2026).
The three design dimensions
Across the four modes, three design dimensions determine where you're exposed. Twelve cells in total — four modes by three dimensions. The cells where you've designed nothing are where the shift will hit you first.
Goals alignment
Does your service model have a shared understanding of what each party is trying to achieve?
Data protocol
Do you have a framework for what data is exchanged, permitted, or requires human sign-off?
Fallback design
When the interaction fails or exceeds scope, what happens? Who picks it up?
Where are you — designed, drifting, or blind?
The Readiness Grid is how we score an organisation across the four modes. Most companies sit in drifting for modes 2 and 3, and blind for mode 4.
Why it matters now
Mode 2 has had a decade of investment. Mode 3 and Mode 4 are happening in production, not in roadmaps. The gap is not theoretical:
- 14% of UK consumers already use AI agents to interact with brands. Projected 37% by end of 2026 (Braze).
- Consumer agents already pass verification. We tested it — an AI-side agent called a live UK contact centre, passed identity checks, got a resolution. The human agent didn't know. See the recording.
- No commerce protocol covers service. Google A2A, Anthropic MCP, IBM ACP, Mastercard, Stripe — all assume human involvement is a failure state. Service can't work that way.
- That gap is where the Service Handshake sits. An open standard for structured AI-to-AI interactions in service, with a human always reachable. Read the standard.
Designed for the Dual CX shift — or drifting into it?
Three ways to find out. None of them are sales pitches.
Designing for the Dual CX Shift
A half-day working session with your CX and operations leaders. Pre-work, real exercises on your operations, three takeaways: your Exposure Map, two redesigned journeys, and your Readiness Grid.
Enquire →Mode 3 / Mode 4 Readiness Test
We deploy AI agents against your live service channels in a controlled environment. You see how your operations perform before your customers' AI does it for real.
Enquire →Public Dual CX briefings
Quarterly public briefings on what we're seeing — across sectors, modes, and the regulatory horizon. Sign up to be notified when the next one is scheduled.
Notify me →See it in action
Three pieces of evidence behind the framework.
A customer AI called a real contact centre
38-second clip from a 4-minute live call. The human agent didn't know.
Watch the recording → Mode 4 simulationTwo AI agents resolve a customer issue
A delivery problem resolved in one message in, replacement dispatched. Agent-to-agent exchange: 4.2 seconds.
Watch the simulation → ResearchThe State of Dual CX — 100 brands audited
How ready is the UK service economy for modes 3 and 4? Sector breakdown, named brands, and the readiness cliff.
Read the audit →Frequently asked questions
What is Dual CX?
Dual CX is the framing for customer service when both the customer and the brand have AI in the loop. Until now, brands have been the sole designers of the customer experience. With consumer AI agents now contacting service channels on customers' behalf, customer service has two designers — the brand and the customer's AI. Dual CX names the design problem this creates.
Who coined the term Dual CX?
Dual CX was coined by Neos Wave in 2026 to describe the moment customer service became a two-sided design problem. Neos Wave is a UK-based service design firm that wrote the open Service Handshake standard for how customer service works when one or more parties is an AI agent.
What are the four modes of Dual CX?
Mode 1 — Customer (human) ↔ Brand (human): the traditional model. Mode 2 — Customer (human) ↔ Brand (AI): chatbots, IVR, self-service. Mode 3 — Customer (AI) ↔ Brand (human): the customer's AI agent contacts a human service team. Mode 4 — Customer (AI) ↔ Brand (AI): two AI agents negotiate outcomes directly. Most companies are designing for modes 1 and 2. Modes 3 and 4 are happening now.
What are the three design dimensions of Dual CX?
Goals alignment — does your service model have a shared understanding of what each party is trying to achieve? Data protocol — do you have a framework for what data is exchanged, permitted, or requires human sign-off? Fallback design — when the interaction fails or exceeds scope, what happens? Applied across the four modes, these three dimensions form a 12-cell exposure map.
Why does Dual CX matter now?
14% of UK consumers already use AI agents to interact with brands, projected to reach 37% by end of 2026 (Braze 2026). Consumer AI agents are calling contact centres, navigating self-service, and starting to negotiate directly with brand-side AI. Most service operations were designed for modes 1 and 2. The mode 3 and mode 4 gap is being exposed in production, not in planning.
How is Dual CX different from omnichannel CX?
Omnichannel CX is about channel consistency — the same customer reaching you through phone, email, chat, social. Dual CX is about who's on the other end of those channels. In omnichannel, the customer is always human. In Dual CX, the customer can be a human, an AI agent acting on their behalf, or both at different points in the same interaction.
What is the Service Handshake?
The Service Handshake is the open standard Neos Wave published for how service works when one or more parties is an AI agent. It defines six declaration elements every party exchanges before the interaction begins: goals, options and constraints, data permissions, fallback rules, cost parameters, and authority level. It is licensed under CC BY 4.0 and is the protocol layer that makes Mode 4 (AI ↔ AI) interactions auditable and safe.