Your team probably chased the wrong AI update this week.

Fable 5 got the headlines. Fair. It is the world’s new most powerful model, and it came with a 30-day enterprise privacy problem hiding in plain sight.

But the bigger business story was quieter.

AI tools are starting to collapse real work into the chat box: translation, charts, agents, video, and email.

That is the real FOMO. If competitors wire these into work first, y’all are not behind on AI news. You are behind on workflow design.

This week’s lesson is simple: model power matters, but distribution into daily work matters more.

That’s what we tackled today on Everyday AI Friday Features, 7 fresh AI features that can change how your team communicates, builds, visualizes, creates, and responds next week.

Let’s get featurey.

Anthropic made the loudest move of the week with Claude Fable 5, the generally available Mythos-class model.

Fable sits above Haiku, Sonnet, and Opus as the safer public version of the Mythos family, while Mythos Preview and Mythos 5 stay limited to Project Glasswing.

It is available to paid Claude users on Claude.ai, Claude Desktop, the API, AWS Bedrock, Vertex AI, and Microsoft Foundry.

But the enterprise catch is ugly.

Fable 5 has a mandatory 30-day data retention policy, and prompts and outputs can be reviewed by a human.

Microsoft reportedly stopped using it for that reason, and we said large enterprises should probably sit it out.

Fable 5 also falls back to Claude Opus 4.8 in high-risk areas like cybersecurity, biology, and chemistry.

If you are building 3D worlds, simulations, games, videos, websites, or individual coding artifacts, this thing is still worth hammering before included access ends June 22.

Try This

Fire up Fable 5 with one build-heavy task this week: a website concept, simulation, 3D world, game idea, or artifact.

If sensitive data is involved, do not touch it until the retention issue changes. If not, use the included access window and see where it beats your current model.

Google’s Gemini 3.5 Live Translate is the kind of AI feature that sounds small until your global team actually needs it.

It is a new audio model for near real-time speech-to-speech translation across more than 70 languages.

It auto-detects the spoken language and preserves intonation, pacing, and pitch.

The unlock is continuous translation, staying only about a second or two behind instead of waiting for the speaker to finish.

It is available through the Gemini Live API and Google AI Studio, with preview access for businesses using Google Meet and users in the Google Translate app on Android and iOS.

In Google Meet, it moves from a handful of languages to more than 70, and from English-only pairs to 2,000 language combinations in one meeting.

It can also use a Chrome tab as the input in Google AI Studio, which matters for live streams, videos, trainings, and anything else happening in the browser.

Global meetings, customer calls, travel, and voice apps just got a lot less clunky.

Try This

Pick one multilingual meeting, training clip, customer call, or browser-based video and test Gemini 3.5 Live Translate against it.

Watch for delay, tone, and whether the translated speech still feels natural enough for business use. If it works, language stops being a meeting blocker.

ChatGPT finally made data inside chat less painful.

It can now turn some answers into rich interactive bar, line, pie, and scatter charts directly in the conversation.

You can ask for a chart, or ChatGPT can add one when the response has data worth visualizing.

Any interactive chart can be switched back to a static image, and the feature is available to any user.

The same update also brought full-screen writing edits, tables of contents for long chats, and editing messages with attachments.

The business value is not fancy charts.

It is killing the spreadsheet detour when your team just needs to eyeball trends, compare data, or make sense of uploaded numbers inside the same workflow.

Try This

Rerun one data-heavy prompt and ask ChatGPT for an interactive chart.

Use it on scattered numbers, benchmark comparisons, or uploaded data. If the visual gets you to the decision faster, bake that into the workflow.

Microsoft rebuilt Copilot Studio for the exact place agents usually fail: multi-step work.

Copilot Studio is the no-code and low-code builder for agents and workflows inside the Power Platform and Copilot Studio stack.

The new version has a streamlined authoring experience, a modern AI core, a new agentic orchestrator, a new agent building interface, and support for skills.

It also works with Microsoft IQ, the system that helps company data follow you around the Microsoft 365 Copilot ecosystem.

Microsoft said the rebuilt version strengthens instruction following and complex task handling for more consistent results.

The rollout is generally available worldwide, ready for production use, and goes into existing Copilot Studio environments.

The numbers that matter: about 20% better evaluation performance, about 50% lower net token usage, and configuration tabs dropping from nine to four.

If you build internal agents for IT, ops, business analysis, or enterprise workflows, this is the version worth rechecking.

Try This

Open Copilot Studio and rebuild one fragile multi-step agent in the new authoring experience.

Start with a workflow that currently breaks halfway through. If the new orchestrator holds up, move it from “interesting demo” to an actual internal process candidate.

HeyGen’s HyperFrames connector is niche until your team needs more video than your video team can handle.

HyperFrames itself is not new; it was announced back in April as an open-source HTML-to-video framework.

The new part is the official Claude connector.

It lets Claude and Claude Code turn prompts, specs, and documents into code-first composable scenes that render shareable MP4s.

That means Claude’s reasoning can drive video creation without your team living inside a traditional video editor.

It runs through MCP, and Fable 5 looks especially useful here because it is strong at individual artifacts.

The obvious plays are training videos, sales videos, customer support clips, marketing content, creator content, and developer-led video workflows.

That is the unlock: video drafts from structured instructions, not another editing bottleneck.

Try This

Give HyperFrames one small video job: a training explainer, sales clip, customer support walkthrough, or product promo.

Feed it a prompt, spec, or document instead of opening a timeline first. If Claude gets you a usable MP4 draft, fam, that is a new content workflow.

Microsoft Copilot quietly added a Team of Advisers experience.

It is a roundtable of AI-generated adviser voices that appear on screen when you ask a question.

Each adviser brings a different perspective, expertise area, and way of reasoning.

They challenge, build on, and rethink each other in real time, which is the whole point.

You are not asking one chatbot to be wise; you are forcing multiple perspectives to stress-test the answer.

Access appears to be for paid Copilot online users with Copilot Labs enabled, under the Experiences tab, not Microsoft 365 Copilot.

Microsoft also included prebuilt modes like historical figures roundtable, product discussion, brand naming workshop, startup idea clinic, business war room, and general decision making.

Use it for messy decisions, not simple lookup work.

Try This

Bring one decision your team keeps circling and run it through Team of Advisers.

Pick a prebuilt roundtable instead of building personas from scratch. If it surfaces better objections than your first prompt did, keep it in the strategy toolbox.

ChatGPT can now close the loop on one of the most annoying admin workflows: email.

If Gmail or Outlook is connected, paid subscribers can ask ChatGPT to draft and send emails directly from the same conversation.

Drafting and inline editing happen inside writing blocks, which now replace the old dedicated Canvas flow in this part of the experience.

The new step is the send button.

You still need write access authorized in the Gmail or Outlook connector.

On a team plan, an admin may need to enable the tool call.

Right now, it is web only and not available on mobile.

For long threads, missed replies, triage, and ugly copy-paste formatting, this is one of those small features that can actually save real time.

Try This

Ask ChatGPT which emails are waiting on your response, then give it three bullets for one reply.

Edit the writing block before sending. If it saves you from inbox bouncing and formatting cleanup, make it part of your morning triage.