Trello added seven AI features. The smallest three may save the most time.

The cleanest feature is card merging. Drag one duplicate onto another and Trello creates a consolidated card, combines descriptions and checklists, preserves the original content, and archives the source cards. An immediate undo can restore them. Atlassian's support page says the feature works on web and desktop for Standard, Premium and Enterprise users with AI enabled; mobile is not supported yet.

That is a good shape for everyday AI: one familiar gesture, a visible result and a recovery path. The user does not have to open a chatbot, explain the board structure or approve a long plan. The system handles a reconciliation chore at the exact spot where the duplicate appears.

The due-date feature is even smaller. Type ‘Submit report by Friday’ in a card title and Trello turns Friday into the due date. No separate date picker. The focus-time feature then looks at a connected calendar and recommends an open block for cards due within five days. Trello says it can read the calendar to suggest time, but it does not change the calendar without the user's say-so.

None of this is science fiction. Good. Work automation becomes useful when it removes the little translations people repeat all day: sentence to date, card to calendar, duplicates to one current item.

Creating a full board from a prompt sounds more impressive. Describe an offsite, a 5K plan or a week of meals and Trello generates lists and cards on a private board. One click can also turn a card into an editable checklist.

Those features solve blank-page friction. They can also manufacture plausible work. A twelve-step checklist feels organized even when six steps do not apply, the owner is missing and the actual constraint is buried in an email. The time saved at creation can come back as review and deletion.

The right test is not whether the generated board looks complete. Give it one real project that already has awkward details: a vendor who only answers Tuesdays, a room that is unavailable, a dependent task owned by someone else. Then count what survives the first week. If people keep the cards, assign them and close them, the setup helped. If the board becomes a polished graveyard, generation only made abandonment faster.

This is where the small features have an advantage. They begin with work a person already decided mattered. They tidy or place it instead of inventing a larger plan around it.

The release is part of a broader Trello shift. A week earlier, Atlassian opened a Trello connection for AI assistants including Claude, ChatGPT, Gemini and Cursor. Through that connection, an assistant can create boards, add or move cards, update checklists, search, and work with Planner. Atlassian says nearly half of its MCP users are outside software teams.

Trello is also testing generated widget cards that can bring changing information such as headlines or schedules onto a board. That may be useful, but it changes the job of the surface. A board that once held decisions and commitments can become another feed asking to be checked.

There is a simple line worth protecting: information should enter the board because it changes a task, not because it can refresh. A daily headline widget that never changes a card is decoration with a maintenance schedule. A shipment status that unblocks a customer order belongs there.

AI assistants are useful when they carry context into the place where work closes. They are less useful when every old tool grows a new stream of generated material.

Jun Vega likes that the originals are archived after a card merge, but the immediate undo banner is not enough for a team that notices the mistake tomorrow. His version would put a plain ‘merged from two cards’ line on the new card with a durable restore path. Recovery should live beside the changed work, not disappear with a toast notification.

Cass Bell is more interested in subtraction. Seven new features are only a productivity win if one old ritual goes away. Stop manually deduplicating Friday's board. Stop copying dates from titles. Stop opening the calendar to hunt for a slot. If the team still performs every old check while also reviewing AI output, the feature list grew and the day did not shrink.

They pull in different directions in a useful way. Jun wants safe recovery after the clever shortcut. Cass wants evidence that the shortcut removed something. Together that is a better standard than ‘AI-powered’: can a normal person trust the change, and is there actually less work afterward?

Do not rebuild your whole system. Pick the board that already annoys you: client work, household repairs, a launch, the volunteer event nobody wants to coordinate.

First, merge only obvious duplicates. Open the result and check the title, earliest due date, links, attachments and checklists. Find the archived originals and make sure someone other than the person who merged them can restore one.

Next, put due dates directly in three card titles and schedule focus time for one task that has been floating for a week. Decline one bad calendar suggestion on purpose. The tool should make that correction cheap and should not quietly move the deadline when it moves the work block.

Skip generated backgrounds. Use a generated checklist on one task, delete every invented step and assign the parts that remain. At the end of the week, ask four blunt questions: Did fewer duplicates survive? Did one task get real time on the calendar? Did anyone recover a mistaken change? Which board-maintenance step can stop next week?

That is enough to judge the release. You do not need an AI project manager. You need a few chores to leave the room.