Quick answers to common questions about SmartAssistant — what it can and can't do, why it behaves certain ways, and how to troubleshoot when something isn't working as expected.
Some actions are excluded from V1 - either too risky to automate or outside SmartAssistant's scope. Specifically, SmartAssistant will not:
Delete campaigns, email accounts, or leads
Change your billing plan, subscribe, or unsubscribe
Perform bulk destructive operations on multiple items at once
Modify existing email sequences after the initial creation (you'll need to edit these manually in the sequence editor)
Connect or disconnect email accounts on your behalf (OAuth flows require your direct authentication)
Edit DNS records (those live with your domain registrar)
Process payments or refunds (these go through support)
Guarantee deliverability (it diagnoses and advises, but final placement depends on factors outside Smartlead)
See data from other users in your workspace (strict account isolation)
If you ask SmartAssistant to do any of these, it explains what's blocking the action and guides you to the right screen so you can do it yourself.
SmartAssistant classifies every action by risk level:
Low risk — done immediately without asking (navigating, highlighting elements, pulling data)
Medium risk — done immediately but easy to undo (opening modals, typing into fields, applying filters)
High risk — always requires your explicit confirmation
High-risk actions include launching a campaign, pausing or resuming a campaign, large batch operations, and anything that sends emails or affects active leads. These can't be easily undone, so SmartAssistant stops and asks before proceeding. You can always say no and ask for more details first.
SmartAssistant handles most product questions and workflows, but some issues fall outside what it can resolve - technical bugs, account-specific billing issues, complex deliverability problems, or edge cases it doesn't have context for.
When this happens, SmartAssistant opens a support ticket via Gleap and pre-fills it with the full context of your conversation. You'll see a "Submitted via SmartAssistant" note in the ticket so the support team can pick up exactly where SmartAssistant left off. You get a chance to review the pre-filled ticket before it's submitted.
Support operates in IST and typically responds within 48-72 hours.
Not in V1. If you close the SmartAssistant panel, refresh the page, or navigate away from Smartlead, the conversation resets. Anything SmartAssistant already completed (campaigns, leads, sequences) stays in Smartlead - only the chat history clears.
Conversation persistence is on the V2 roadmap. Until then, if you're running a long workflow, complete it in one session.
Not yet. In V1, SmartAssistant doesn't store per-user preferences between conversations. If you always want a casual tone or prefer 3-email sequences by default, you'll need to specify these in each new chat.
Per-user memory is on the V2 roadmap. Once it ships, SmartAssistant will remember your communication style, common ICPs, and frequent workflows without you having to re-specify them.
Most common reasons:
It's waiting on a clarifying answer. Check the last message - SmartAssistant often asks a question mid-flow (like "which email accounts should I assign?") before continuing.
It hit a step limit. SmartAssistant executes actions in a loop with a safety cap of 15 steps per task. Very complex workflows can hit this. Break the request into two or three smaller asks.
It's waiting for the page to load. After navigating or clicking, SmartAssistant pauses briefly. If it looks stuck, give it a second.
A high-risk action needs your confirmation. See question 2.
If it really gets stuck, click the refresh icon in the panel header to reset the conversation and try again with a more specific prompt.
A few things that help:
Be specific. "Create a campaign for SaaS founders" works better than "make me something."
Front-load context. Passing ICP, product, tone, and email count in your first message speeds things up considerably. SmartAssistant carries that context through the whole flow without stopping to re-ask.
Use full sentences, not keywords. SmartAssistant is optimized for natural language. "Find CTOs at B2B SaaS companies" works better than "CTO B2B SaaS."
Break complex requests into steps. If your request has 5+ moving parts, SmartAssistant might miss one. Split it: find leads first, then create the campaign, then write the sequence.
Rephrase if the first try misses. If SmartAssistant interprets your request unexpectedly, say "actually, I meant..."and clarify. It will adjust.
If SmartAssistant consistently misunderstands a type of question, leave a thumbs-down with feedback. Corrections feed directly back into the system - it'll be better at the same question within a day or two.
Currently there's no per-user toggle to disable SmartAssistant entirely. You can close the panel using the X in the header any time - the floating button minimizes and SmartAssistant won't run in the background until you open it again.
On the V2 roadmap based on what we've heard from early users:
Conversation persistence - conversations saved to the database and resumable across sessions
Streaming responses - token-by-token streaming for faster perceived speed
Quick-reply chips - suggested follow-up actions as clickable buttons in chat
Per-user memory - SmartAssistant remembers your preferences, common workflows, and preferred tone
Expanded automation - deeper click-and-type support for Master Inbox, Smart Agents, Power Dialer, and CRM (currently guidance-only)
Proactive monitoring - SmartAssistant surfaces recommendations when you hit plan limits, deliverability drops, or features go underused
Smart Agent integration - create and manage AI reply agents entirely through chat
Comparative insights - compare your campaigns against similar users on the platform (e.g., "users targeting the same persona who send between 11am-12pm get 40% more replies than your 3pm-4pm window")
If there's something specific you want prioritized, leave a thumbs-down with your feedback on any response - that's the fastest way to get it into the roadmap queue.
