The Agentle Framework
Most digital products were designed for one kind of user. That assumption is now a liability.
AI agents are already browsing, evaluating and transacting autonomously on behalf of the humans who deploy them.
By 2028, Gartner predicts 90% of B2B buying will be agent-intermediated. The agent is not a future user. It is a current one.
The Agentle Framework is the first named, auditable methodology for Agent Experience Design. It answers a question no existing UX or technical standard has addressed: not whether your product can interact with an agent, but whether it was designed for one.
6 principles.
If an agent cannot find your product, it will not consider it. Findability measures whether your product announces itself clearly to agents searching on behalf of a customer: through structured data, consistent naming and explicit capability signals. Invisibility at this stage is not a technical problem. It is a revenue problem.
Findability
Being found is not enough. An agent must be able to extract accurate information from what it finds. Legibility measures whether your product's content is structurally readable, not merely visually present. Pricing in a graphic, features buried in JavaScript, ambiguous labels: each one is a point at which an agent stalls, infers incorrectly, or misrepresents the product to the customer it serves.
Legibility
A customer's agent that cannot complete a transaction will not retry. It will move on. Executability measures whether your product can be transacted with: whether APIs work reliably, whether workflows can be completed without human intervention, and whether the product's capabilities match what its documentation promises.
Executability
Agents encounter errors. What matters is what happens next. Recoverability measures whether your product fails gracefully: whether errors are explicit rather than silent, whether state is preserved, and whether the agent is given enough information to try again or route around the problem. A product that fails without explanation loses the transaction permanently.
Recoverability
The first four principles describe what a product must do. Considerateness describes what it chooses to do. A considerate product anticipates what an agent needs at each decision point and volunteers it: eligibility criteria before sign-up, rate limits before workflow initiation, stock levels before checkout. It is the difference between a product that technically works for agents and one that was built with them in mind.
Considerateness
An agent's recommendations are only as reliable as the information it encounters. Fidelity measures whether your product's data is accurate, current and consistent across every surface an agent might access: website, API, documentation, structured data feeds. Poor fidelity does not cause agents to fail. It causes them to make confident mistakes, at scale, on behalf of real customers.
Fidelity
What an audit produces
Every Agentle audit delivers four outputs: a scored report across all six principles with evidence and commentary; a prioritised recommendation list ordered by impact and feasibility; a 90-day agent-readiness roadmap with clear ownership; and an executive summary written for board-level consumption, translating findings into business risk and commercial opportunity.
Why it matters now
Gartner predicts that over 40% of agentic AI projects will be cancelled by 2027 due to unclear value and escalating costs. The missing variable in most of those failures is not infrastructure. It is design. Products that were not built to serve agents will fail agents, reliably and expensively, until someone fixes the design layer.
That is what the Agentle Framework exists to do.
Request a framework briefing or discuss an audit.