Meet Atlas & Ava

Overview

brayv.ai gives you two AI assistants, and they're built to complement each other:

• Ava Ask mode at the top of every page, edit mode lives inside the tools you already use. It answers questions about the platform and makes quick, targeted edits to the resource you have open.
• Atlas has its own page and works across your whole account. It plans multi-part configuration changes for your review, and it can investigate exactly why something happened with a contact.

A quick rule of thumb:

Atlas is generally the new go-to for changes, but Ava still has a real role. Here's the breakdown.

Ava: quick answers, quick edits

Where you'll find it

• Ask Ava — the button in the top bar and in the Help section. This is Ava's question-answering mode, backed by the full help documentation.
• The AI assistant chat in your Text AI Agent editor — edit the agent conversationally, including working on drafts.
• The AI assistant chat in your Voice AI Agent editor.
• Ava — Code Mode on booking rules (calendar selection) and lead-source assignment rules, where it writes and updates the selection logic for you.

What it's great at

• Platform questions. "How does X work? Where do I go for Y?" Ask Ava has the broadest knowledge of the whole platform — every feature, screen, and setting — and answers with citations to the help docs.
• Small, quick changes when you already know which piece to change. "I want an email when a lead says they're ready to buy" — said in your text agent's editor, that's one conditional tool, done in a minute.
• Experimenting on text agent drafts. The agent editor's draft system lets you play with changes without touching your live agent, and Ava edits drafts happily.

Ava works on the resource in front of it. Ava can't make the multi part edits needed to conclusively tie together more complex operations within an agent - how should the instructions, tools, and booking routing be set up to optimally perform together? It can edit each part individually and do it well, but relies on you to keep in mind what's needed in the parts it's not working on at the momement. It won't check whether a change to one agent leaves a loose end somewhere else — that's on you (or Atlas).

Atlas: account-level planning and investigation

Where it is

Atlas lightning bolt icon in the main left hand menu (only on desktop). It's a chat on the left and a Configuration Plan panel on the right.

Why use it

Atlas reads your account the way a specialist would — not one resource at a time, but how everything works together: which lead sources feed which journeys, what each agent is tasked with, what it books into, what the triggers move, and what happens after. That changes what you can ask for:

• Describe an outcome, not a setting. "If a customer wants a phone call instead of an on-site visit, book that instead." Atlas reverse-engineers the structure: the custom field, the extractor that fills it, the booking rules that route on it, the calendar, and the prompt language — staged together, with nothing left dangling.
• Changes that touch multiple resources. Where Ava would handle one piece and rely on you to tie the loose ends, Atlas stages every mutually required piece as one plan and cross-checks the result before you apply.
• Understanding your own setup. Ask "walk me through how my account is set up" and Atlas traces it in plain language: where leads come from, what happens when they arrive, and how they get to booked.

Nothing is written until you approve it. Atlas stages a plan, you run Review (it checks every change against your live configuration for conflicts and risks), you acknowledge any warnings, and only then does Apply write the changes. Applied conversations are kept as a permanent, read-only record of exactly what changed and why.

What Atlas can read

Everything that defines how your account runs: text and voice agents (full configuration), journeys, stages, and their calendar triggers, calendars and availability, lead sources and their journey routing, assignment rules, follow-up sequences, completion goals, custom fields, team members, your phone numbers and call routing, and your account's texting status.

What Atlas can change

Atlas stages changes to: text agent instructions, field extractors, conditional tools (including advanced AI decisions), booking routing, follow-up sequences, completion goals, custom fields, calendars (create and edit), journey calendar triggers, new text agents and journeys, voice agents, outbound voice calls in journeys, initial outreach messages, lead-source assignment rules, the account's default agent roles, and knowledge-base FAQs — it can even run the exact same knowledge-base search your agents run, with match scores, to prove why an answer wasn't found and to verify its fix actually works.

Atlas 'stages' all changes to the plan panel and explains them in the chat. You review and click a button for one final review pass from Atlas. After it passes the Atlas review, you click Apply to save the changes.

Every applied change goes to your live configuration. Atlas can read a draft you were working on and carry the good parts forward, but it doesn't edit drafts — that stays with Ava in the agent editor.

Investigate mode

Flip the toggle under the chat to Investigate and Atlas becomes read-only and forensic. It reconstructs what happened from your account's actual records: full message threads, what the agent saw and did on every turn (with the configuration that served that turn), voice call transcripts, appointment booking traces (who was eligible, who was chosen, and why), journey timelines, and lead submission records.

"Why did this Joe Smith get booked with the Amy Smart?" "Why didn't the AI reply to the most recent message from Tom Smith?" "Why did it offer 5pm as the latest time?" "Walk me through everything that happened with Jose Ruiz" — Investigate answers with evidence, and when the answer leads to a desired configuration switch, one click switches you back to Build mode to stage it.

Atlas has access to a lot of change history in the account to make auditing past events more accurate. Quite a few of these change histories are recent additions (as of Aug 2026), so historical events prior to that may have less for Atlas to work from but it is still good at inferring what happened and will let you know its confidence level in the evidence.

Using them together

They genuinely complement each other. A typical week might look like: Ask Ava where a setting lives, have Atlas investigate why a lead stalled, apply the fix it proposes, then hop into your agent editor and have Ava add one quick notification tool. Quick and targeted — Ava's right there. Broad, multi-part, or "what actually happened?" — that's Atlas.