The Assistant now works out loud. As it reaches an answer, you can watch the steps it takes and the metadata queries it runs — what it's looking for, what it found, and how it's connecting the pieces. The experience is far more interactive: instead of waiting on a finished answer that shows context, you're working alongside it.
​
It tells you when it needs clarification, flags when something is ambiguous, and shows you the reasoning you'd want to double-check before acting on a result.
​

The payoff is confidence. When you can see how the Assistant reached a conclusion, you can trust it — and when it's missing context, you'll know immediately and can steer it.
​

The redesigned interface is quicker to navigate and easier to live in for longer sessions. It also lays the groundwork for your Assistant to tap into a growing set of skills and tools over time, so it keeps getting more capable.
​

The Assistant and AI Admin are designed to work as a team, and the upgrade tightens that connection.

The AI Assistant is where you think. Use it to explore your org, weigh options, design a solution, and pressure-test the impact of a change — all the analysis and brainstorming that comes before you touch anything.
​

When you're ready to act, hand off to AI Admin without leaving the conversation. Because the two share context, AI Admin picks up everything you worked through with the Assistant — the components involved, the decisions you made, the plan you landed on — and goes to make the actual changes, with you in control of approvals at each step. (Working with Admin requires the appropriate permissions.)
​
Think of them as two teammates with handoffs:
​

AI Assistant is your analyst and thought partner. It reads your org, answers questions, surfaces impacts and risks, and drafts solutions. It doesn't make changes to your org.

​AI Admin is your implementer. Once you know what you want done, AI Admin makes the actual metadata changes — with you in control of approvals at each step.

The upgrade connects them: you can move from "what would this break?" to "go make these changes" without leaving the conversation.
​

The result: you do your designing and brainstorming with the Assistant, then hand a fully-briefed implementer the work — no switching tools, no re-explaining what you're trying to do.

Ask the Assistant: "I want to change the first stage of my Opportunity Stage field — what are the downstream effects?"

The Assistant runs through your metadata and surfaces components that could be broken by the change. Because you can see everything it searched, you know the metadata that it it evaluated.
​

Hand it off to AI Admin: Once you know what needs to change, kick off an AI Admin task directly from the conversation to make the updates. If you have the right permissions, you can assign the changes out and let AI Admin handle implementation while you stay in control of approvals.

Ask the Assistant: "I have two flows, Renewal Automation and Closed Won Automation, what does each one do, and are they redundant?"
​

The Assistant inventories the flows, lays out what each does, and flags redundancies.
​
​Follow up: "Walk me through a plan on how to combine the two" Then, if you're ready to act, hand the cleanup to AI Admin.

Ask the Assistant: "Why is the field Account.Territory__c being populated automatically?"
​

The Assistant traces the automation back to its source, showing you the flow, apex, or other automations that could be responsible — with the metadata queries visible so you can verify the trail yourself.

See Swantide's AI Assistant Use Case Ideas and the Best Practices & Prompt Library for prompts across every phase of your work.