An update can change the tool people knew

What we found in the news

The July 2024 NIST Generative Artificial Intelligence Profile (opens in a new tab) sets out actions organizations can consider when managing generative AI risks. One recommendation is particularly practical:

“Verify adequacy of GAI system user instructions through user testing.”

— National Institute of Standards and Technology

This is voluntary guidance, not evidence that a particular product performs well. It directs attention to something easily overlooked: instructions are useful when they fit the tool people are actually using.

Why this matters to us

Someone learns a route through a service. They find the right page, recognize a familiar button, and keep a small instruction card nearby. Then the tool changes.

The new version may be better in several ways. Yet the instruction card can become wrong overnight, leaving a person unsure whether the problem is theirs.

Imagine a community resource assistant receiving an update. A staff trainer and participant repeat the task of finding next week’s activity information. The answer now appears in a different place, and the earlier step card no longer matches.

At genus, we want that kind of change to be noticed as part of introducing the update. People’s effort in learning the service deserves to carry forward.

Keep a few familiar tasks ready to try

Choose examples that represent what people depend on: finding a resource, checking where information came from, or returning to a useful page after an interruption.

Use fictional information where private details would otherwise be involved. Record the task and the relevant conditions, so someone else can repeat the exercise without guessing what was tested.

In our imagined update, the team can compare the familiar route with the new one. Does the task still reach the same useful result? What help is now required?

For an AI-assisted step, look at meaning as well as placement. A different answer may be acceptable, but an invented opening time or an unsupported recommendation needs review before the team treats it as equivalent.

Update the help people can actually find

An announcement that a version has changed may tell staff little about how to support the next person who asks for help.

Prepare guidance around the affected task. A short revised card, a demonstration, or a practice session may be useful, depending on the audience. Check that old instructions are removed or clearly distinguished where they would cause confusion.

Ask someone who uses the task to try the revised guidance. Give them room to point out an unclear step. Their difficulty can reveal a gap between what the writer assumed and what the screen now shows.

Make sure colleagues who provide occasional cover can find the same explanation. The person asking for assistance should not need to wait for the one staff member who attended the demonstration.

Decide what happens when the task no longer works

Some changes will need more than a rewritten card. A feature may have moved behind a different permission, a response may have become less useful, or an expected route may have disappeared.

Name who gathers those reports and who decides the next action. A team might seek clarification from its supplier, adjust the configuration, or provide another agreed route while the issue is resolved.

Tell people what is known and what remains uncertain. Avoid asking them to keep retrying an unhelpful step without explaining what will be different.

Use the findings to refine the next change review. The useful record is the task affected, the action taken, and the remaining limitation—not simply a note that the update was installed.

What this means for the people you serve

Learning a tool is work. When the tool changes, people deserve guidance that respects that effort and help that fits the new experience. They should be able to report a problem without being made to feel that they have forgotten how to use the service.

The genus connection

Today, genus care-app and care-wall experiences are configured around an organization’s audiences, content, and enabled functions. That configuration helps define which everyday tasks should be checked when a related feature changes.

We would bring staff guidance and participant-facing resources into the same review. Confirm the actual behavior, revise the relevant help, and agree who responds if a familiar task becomes difficult. The useful unit of progress is a person completing the intended task with appropriate support.

One question to take away

Which familiar task will we try with someone who uses it before we call this update ready?