The Service Handshake · SH-1.2
The open standard for service conversations when AI joins in.
Service only works when every party knows who they’re dealing with and what each side can agree to. The Service Handshake sets those terms whenever AI is part of a service conversation. Free to use. Vendor-independent.
SH-1.2 · 1 October 2026 · DOI 10.5281/zenodo.23181969 · First published 16 March 2026 · Standard CC BY 4.0 · Schemas and code Apache 2.0
Part of the Dual CX framework, the four ways people and AI now meet in service. See the framework →
Where it fits
The protocols get the agent in. The Handshake sets what it can agree to.
New agent protocols connect customers’ agents, carry the conversation, handle checkout and prove what was approved. Most of the work stops at the sale. Service is what happens next. The Service Handshake works alongside them on the service conversation: what each side may agree to, and what happens when it can’t.
When each one launched is on the Dual CX timeline.
- 1The sale
The purchase
CustomerBuy the kettle we looked at. Under £60, delivered Friday.CompanyOrder confirmed. Delivery Friday.SoldCheckout and payment approval done.- UCP · ACPCheckout and order confirmation
- AP2Bind approval to this item and price
- 2After the sale
It arrives faulty
CustomerIt's cracked. Get me a refund or replacement, not credit. Check with me before agreeing.Terms setThe agent can explore remedies, not accept one.- PoppyConnect and carry the request
- SHDeclare goals, constraints and authority
- 3After the sale
Credit is offered
CompanyWe can offer store credit.Customer's agentCredit is ruled out. Can you offer a refund or replacement?DenyThe credit offer stops. The case stays open.- PoppyCarry the offer and the refusal
- SHReject the conflicting proposal
- 4After the sale
A replacement is offered
CompanyA like-for-like replacement on Friday, with collection. No charge.Customer's agentThat fits the brief. I'll ask my customer.HoldAcceptance waits for the customer's approval.- PoppyPresent the replacement terms
- SHAssess the alternative, hold for approval
- 5After the sale
Resolved
CustomerYes to the replacement and Friday collection.CompanyReplacement delivered. Faulty kettle collected.AllowApproval and authority checked first.- SHCheck approval, then allow
- UCP · ACPUpdate the order status
Decline the offer. Keep working on the problem. Pause commitment until authorised.
Illustrative journey, not a tested integration. The company enforces the checks. AP2 provides approval evidence, not funds movement.
The declaration framework
Six things every party must declare.
Element 1
Goals
What this party is trying to achieve, in priority order. Including explicit non-goals. Without stated goals, an agent optimises for whichever proxy metric it sees first, usually completion, not resolution.
Element 2
Options & Constraints
What this party can and cannot do. Hard limits, soft limits, jurisdictional boundaries. Includes financial authority, data-sharing permissions, and specific actions that are explicitly out of scope.
Element 3
Data Permissions
What data may be used, for which service purpose, how long it is kept and how it is verified. Completing the service does not authorise marketing, profiling, training or sharing.
Element 4
Fallback Rules
What triggers clarification or escalation, how far automation can go, and who is responsible. Safeguarding routes cannot be waived, and restrictions travel with any human handover.
Element 5
Cost Parameters
Limits on interaction time, exchanges and computing resources. Sets the ceiling above which the agent must escalate. A budget is not permission to spend money.
Element 6
Authority Level
Who the agent represents, who is affected, what it claims it can commit to, and the evidence behind that claim. Setting up an agent is not authority over someone else's account.
What it doesn’t do: the Service Handshake doesn’t issue credentials, provide transport, decide whether a delegation is legally valid or carry out transactions. Those stay with the systems and people that control the action.
Adopting it
Start with your people. Add evidence where it matters.
Two levels. You state which one you meet, and for which services and actions.
Level 1
Participation
The starting point. Run by your staff, using your existing procedures and case records. No JSON needed.
- Disclose AI involvement at first contact.
- Help record the other side’s terms when they arrive without a declaration.
- Check authority. Missing information never counts as permission.
- Keep data to the service purpose and respond to safeguarding concerns.
- Offer a human route that keeps every restriction in place.
Level 2
Verified authority
Everything in Participation, plus evidence for the actions that need it, such as cancelling, switching or paying.
- Independently checkable evidence of authority, linked to the specific action.
- Checks that the evidence is still valid and hasn’t been withdrawn.
- Controls so an approval can’t be reused improperly.
- A record of how each action was assessed.
Implementing it? Read the Technical Specification, PDF.
See it in action
Two AI agents. One damaged delivery.
A customer’s AI contacts a brand’s AI. Watch it resolve when the handshake aligns, then watch what happens when it doesn’t. Narrated, about two minutes each.
Resolved in 4.2 seconds
Both declarations align. Replacement dispatched, tracking returned, full audit trail on both sides. No re-explaining.
Escalated cleanly
Refund only, and the brand can’t refund this item. It says why, hands over with the evidence attached, and a person approves the refund. No chasing.
Acting for someone else
A daughter sets up an agent for her father’s broadband. It can discuss bills and renewals, but can’t cancel or switch until the account holder approves. Setting up an agent is not authority over someone else’s account.
Why now
Customers’ agents are arriving. Most brands aren’t ready.
Without shared terms, every service conversation with AI in it starts from scratch. Agents over-commit, refuse without a reason, loop, or share data they shouldn’t. The cost shows up in resolution time, customer trust and regulatory exposure.
The Service Handshake is deliberately narrow. It doesn’t tell you how to build your agent. It sets what each side must be able to say, and hear, before it’s safe to go ahead.
Since 2 August 2026, the EU AI Act’s transparency rules (Article 50) have required AI systems to let people know they are dealing with AI. The Service Handshake adds an operational rule: every agent says it is AI, and whom it represents, at first contact, including agent to agent.
Adopting the standard doesn’t by itself make you compliant. GDPR purpose limitation, data minimisation and legal basis remain separate obligations. What it gives you is a reviewable record of declared purposes, authority claims and decisions.
The evidence
Our audit of 100 UK brands found 0/100 with a published AI-interaction policy at the time of fieldwork. 73% had no fallback path when an AI agent reached them.
Read the audit reportNew in SH-1.2
Easier to join. Harder to misuse.
- No declaration? No problem.Assisted participation replaces automatic rejection. Missing information never counts as permission.
- AI says it's AIAgents disclose that they are AI, and whom they represent, at first contact.
- Two levelsParticipation and Verified authority, so you can start without building anything.
- Confirmation is not approvalIrreversible actions need checkable delegation or approval from the person represented.
Full list in the changelog.
Resources
Read it, cite it, build on it.
- The standard, SH-1.2 (PDF)
- Technical Specification, SH-1.2 (PDF)
- GitHub: schemas, examples and tests (Apache 2.0)
- Pre-webinar brief for contact leaders, by Martin Hill-Wilson
- Zenodo: SH-1.2 · DOI 10.5281/zenodo.23181969
- Zenodo: SH-1.1 · DOI 10.5281/zenodo.19046746
Audio guide: SH-1.1 overview
16 minutes on the core ideas. Recorded for SH-1.1.
Using it? Credit “The Service Handshake, Dual CX. CC BY 4.0” and link back. How to credit us.
Questions
The Service Handshake, answered.
What is the Service Handshake?
An open standard for service conversations where at least one party is represented by AI. It sets six things each side must declare: goals, options and constraints, data permissions, fallback rules, cost parameters and authority level.
Is it free to use, and how is it licensed?
Yes, and it is vendor-independent. The standard and its documents are CC BY 4.0. The SH-1.2 schemas, examples and tests are Apache 2.0, so you can use them in commercial software. Credit "The Service Handshake, Dual CX" and link back. The Service Handshake name, logo and badges are not covered by either licence.
Do we need to build anything to start?
No. The Participation level is run by your staff using existing procedures and case records, with no JSON needed. Verified authority adds evidence checks for actions that need them, such as cancelling, switching or paying.
How does it relate to PAP, UCP, ACP, AP2, A2A or MCP?
It works alongside them. Those protocols connect agents, carry the conversation, handle checkout and provide evidence of payment approval. The Service Handshake covers what each side may agree to in the service conversation, and what happens when it can't.
Does adopting it make us compliant with the EU AI Act or GDPR?
Not by itself. It supports the AI Act's disclosure rules by requiring every agent to say it is AI, and whom it represents, at first contact. GDPR obligations remain separate. It gives you a reviewable record of declared purposes, authority claims and decisions.
Can a customer's agent act for someone else?
Only within the authority it can show. Setting up an agent is not authority over someone else's account. Actions the evidence doesn't cover are held until the person represented approves.
Use it. Build on it.
The standard belongs to whoever uses it.
Fork the repo. Reference the DOI in your architecture docs. Ship declarations in your agents. Contribute back what you learn.