20 / 20Future of work

AI capability should be visible in the systems, controls and outcomes a business can inspect.

AI claims are easy to make. A useful business system needs more than a polished screen or a demonstration that works once.

Before commissioning AI work, ask how the proposed system will use business data, where its output will go and who will be responsible for reviewing it.

01Define the business problem

Start with a specific task and a measurable outcome. Reducing the time to review a service request is clearer than a promise to “transform operations with AI”.

Agree on the starting process, the people affected and the result that would justify the investment. That gives the team a basis for deciding whether AI is even the right tool.

02Show the working system

A useful demonstration follows information through the real workflow. Show where content is created, where customer or operational records are held, which system owns each decision and how approved output reaches its destination.

Design and code should be tested together. A screen that looks finished can still fail with long content, a narrow viewport or a missing permission.

03Make controls visible

Show what the system may do automatically and what requires review. Staff should be able to inspect a proposed action, correct an error and see the state of the work before it is published or sent.

These controls belong in the workflow itself. A policy document alone cannot tell a user whether a particular draft was checked.

04Test real conditions

Evaluate the system with representative data and people. Test uncertain inputs, failed integrations, inaccessible content and the cases that require a human decision.

Record what failed and what changed. This makes the result easier to trust and maintain than a collection of ideal-case screenshots.

05Prove the outcome

Measure the result against the original task: time saved, errors caught, approval quality or another outcome the business actually values. Recheck the result after launch as the data and workflow change.

Evidence of AI capability is a system that people can understand, use and improve—not a badge attached to a mockup.

SourcesResearch trail

References used for this article

  1. 01Figma State of Design 2026
  2. 02Figma Code to Canvas
  3. 03Figma 2026 AI Report