AXD for Product Leaders

Accept that your product now has two users - the human who sets intentions and the agent who executes them. Both need designed experiences.

Shift your success metrics from engagement to outcomes - in agentic systems, the best experience may be no visible experience at all.

Design your product's trust architecture before its feature set - trust is the load-bearing structure of every human-agent relationship.

Build delegation as a first-class product feature, not an afterthought - users need clear, controllable ways to grant and revoke agent authority.

Plan for the absent state - the most consequential moments in your product will happen when the user is not watching.

Map every agent action in your product to an autonomy level - from 'suggest and wait' through 'act and report' to 'act silently' - based on stakes and reversibility.

Design autonomy progression paths that allow agents to earn greater independence through demonstrated reliability, not just time elapsed.

Implement context-sensitive autonomy - the same action may require different autonomy levels depending on amount, counterparty, or user history.

Build autonomy controls that users can adjust without technical knowledge - sliders, presets, and natural language constraints.

Design your product to degrade gracefully when autonomy is reduced - pulling back agent authority should never break the workflow.

Start your roadmap with high-visibility, low-stakes agent actions that build user confidence before introducing high-stakes autonomous behaviour.

Design trust milestones into your product roadmap - specific moments where the product demonstrates reliability and earns expanded authority.

Build trust recovery features before you need them - the first agent failure in your product will define user perception of the entire system.

Instrument trust metrics - track delegation frequency, scope expansion, manual overrides, and trust recovery rates as core product KPIs.

Plan for trust regression - users who experience agent failures will reduce delegation scope, and your product must support this gracefully.

Audit your product for machine customer accessibility - can an agent discover, evaluate, and use your product without a browser or visual interface?

Design API-first experiences for machine customers - agents need programmatic access to everything humans can do through your UI.

Build machine-readable product specifications that go beyond what humans need - agents require precise attributes, constraints, and compatibility data.

Design your pricing and policies for machine negotiation - agents will compare your product against alternatives at speeds and scales humans cannot match.

Prepare for the machine customer funnel - discovery, evaluation, trial, purchase, and retention all look different when the customer is an algorithm.