HerdAI
HerdAI is your conversational teammate inside Herd — ask it questions about your program, and have it draft and publish trainings, phishing simulations, and tracks, all from a chat.
Overview
HerdAI is the conversational assistant built into Herd. Instead of clicking through pages to build a training, look up a risk score, or stand up a phishing campaign, you can simply ask — in plain language — and HerdAI does the work alongside you.
It can answer questions about your program (completion and engagement rates, risk scores, compliance status, leaderboards), and it can take action: drafting and publishing trainings, building tracks, generating phishing email templates and landing pages, scheduling campaigns, assigning content, and sending nudges. Everything it creates is shown to you as a preview first — nothing is published or sent until you approve it.
To open it, click HerdAI in the left-hand sidebar. HerdAI is available to every signed-in user, though what it can do for you depends on your permissions — see What you can ask HerdAI to do.
You can also start a HerdAI conversation straight from the Dashboard — type a request into the HerdAI prompt at the top of the page and Herd takes you into the chat to finish the job.
The chat
When you open HerdAI with no conversation in progress, you'll see a welcome screen headed "How can I help you today?" with a prompt box and a set of example prompts.
Example prompts. Four suggestions appear as clickable chips below the prompt box, sampled from a larger set of common questions. Click one to send it immediately. Examples you might see include:
What is the training with the highest engagement rate?
Who are the top 5 users with the highest risk score?
Which users have not completed the SOC2 Training Track?
What was the phishing simulation with the highest click rate in the last month?
Type your own. Enter a request and press Enter to send. Use Shift + Enter to add a new line without sending.
Responses stream in. HerdAI replies in real time. When it answers a data question, it can render the result as an inline chart — a bar, line, pie, or gauge visualization — with a "View raw data" table for the exact numbers.
A couple of keyboard shortcuts speed things up:
Cmd/Ctrl + K
Jump to the prompt box
Cmd/Ctrl + Shift + N
Start a new conversation
HerdAI can make mistakes. Don't send sensitive information in the chat, and always review what it drafts before publishing.
Choosing a model
A model selector sits next to the send button. It controls the model HerdAI uses when generating a training, trading speed against capability:
Haiku
Fastest — drafts a training in about 40 seconds. Best for quick iterations.
Sonnet
Balanced — around 90 seconds. The default for most trainings.
Opus
Most capable — slowest and priciest. Use for the most complex trainings.
What you can ask HerdAI to do
HerdAI works in two modes: answering questions about your program, and taking action on your behalf. Action tools respect your role — HerdAI will only offer to do something you have permission to do.
Ask questions
Analytics & effectiveness — "What training has the highest completion rate?", "What's the difference between completion rate and engagement rate?", "How did our last phishing campaign perform?"
Risk — "Who are the top 5 users by risk score?", "Show me our highest-risk users in Sales.", "Why is this user's risk score high?"
Users & status — "Has Jordan finished the SOC2 track?", "List users who haven't completed onboarding."
Your team (managers) — "Does anyone on my team have outstanding training?", "How did my team do on the last phishing campaign?" Managers with direct reports can ask about their own team's outstanding trainings and their direct reports' phishing simulation results (clicks, reports, credential submissions, and per-campaign outcomes). This covers only your own reports — no admin permission needed, and never anyone else's team.
Compliance — "What's our overall compliance score?", "Which trainings are overdue?"
Gamification — "Show me the points leaderboard for this month.", "What's Priya's rank?"
Threat intelligence & knowledge — HerdAI can pull from recent threat events and your security knowledge base to ground its answers. If your organization has connected Notion as a knowledge base, HerdAI can search and read your Notion pages too.
Integration health — "Is our Okta sync healthy?", "Any Teams delivery failures?" HerdAI reports connection status, last sync, sync errors, user/group counts, and delivery health for Okta, Azure AD, Slack, Microsoft Teams, and Notion (requires the org-integrations permission).
How Herd works — "How do I set up a smishing campaign?" HerdAI answers product questions from Herd's own published documentation (this site), citing the page it drew from, and points API questions at the API documentation.
Any answer HerdAI presents as a table can be exported as a CSV — just ask ("export that as a CSV"). On the web you get a download link, in Slack the file lands in the thread, and in Teams you get a time-limited download URL. Exports require the Reporting Dashboard permission.
Create and take action
Trainings — "Build a 3-minute training on spotting invoice fraud." HerdAI drafts a full training (steps, quizzes, tags) for you to review, then publishes it on your approval.
Tracks — "Create an onboarding track with our phishing, password, and data-handling trainings."
Phishing simulations — "Draft a phishing email that looks like a shared-document notification." HerdAI can also generate a matching landing page and schedule a campaign to a named audience. Generating several variants at once is supported via batch generation.
Campaign lifecycle — beyond creating campaigns, HerdAI can launch, pause, resume, and report the status of phishing campaigns. Creating never sends anything (campaigns start as drafts or scheduled), and launching or resuming always shows a preview and asks you to confirm first.
SMS (smishing) templates — "Write a smishing text that mimics a delivery notification." HerdAI previews the message and publishes it to your SMS template library on approval.
Posters & videos — "Make a poster about clean-desk policy.", "Generate a short video for step 2 of this training." Longer videos are planned as multiple narrated scenes with per-scene progress ("Scene 2 of 5") and stitched into one clip. When a video finishes, HerdAI previews it right where you asked: an inline player in the web chat, a link unfurl in Slack, or a playable card in Teams.
Assignments & nudges — "Assign the new phishing training to the Finance group.", "Nudge everyone who hasn't finished the SOC2 training."
Training from a policy — attach a policy document and ask HerdAI to turn it into a training (see File attachments).
Archiving — "Archive the old GDPR training." HerdAI can archive and unarchive trainings and tracks (always reversible).
Digest subscriptions — "Subscribe me to the monthly org recap." HerdAI manages subscriptions to Herd's recurring digests — list, subscribe, and unsubscribe, choosing the cadence and delivery channel in chat.
Feature requests — tell HerdAI what you wish Herd did ("I'd love a weekly PDF report") and it captures the request for the Herd team after showing you a preview and confirming. Available to every user, members included.
Actions that send or change things — assigning content, nudging users, and deleting trainings — always ask you to confirm first. HerdAI shows you who will be affected and waits for an explicit "yes" before going ahead.
Preview, then publish
Anything HerdAI generates — a training, a track, a phishing template, or a landing page — appears as a preview card inside the conversation. Nothing is saved or made live until you approve it.
Training preview shows the title, description, tags, and every step rendered as it will appear to learners, including quiz questions with the correct answer marked.
Track preview shows the title, description, tags, and the ordered list of trainings in the track.
Phishing template preview shows the email in a Gmail-style view, with a Preview / HTML toggle and a Desktop / Mobile toggle so you can see how it renders. HerdAI can also generate header details, template variables, and images for the email.
Landing page preview renders the generated page in a sandboxed frame, again with Preview/HTML and Desktop/Mobile toggles.
On each preview card you have two choices:
Publish — save the artifact and make it available (a training becomes assignable, a template becomes usable in campaigns, and so on). Once published, the card turns into a link such as View Training, View Track, or View Template that opens the new item in Herd.
Continue Chatting — dismiss the card and keep working in the conversation.
To refine a draft, just keep chatting — tell HerdAI what to change ("make step 2 shorter", "use a more urgent tone") and it regenerates the preview. There's no separate edit form; the conversation is the editing surface.
HerdAI won't publish in the same breath as it drafts. It always pauses on the preview and waits for you to say "publish", "save", or "looks good" before committing anything.
File attachments
You can attach files to a message and HerdAI will work from them — for example, turning a policy PDF into a training, or reading a screenshot you paste in.
Attach files with the + button in the prompt box, or drag and drop them onto the chat.
Supported file types:
Images
PNG, JPEG, GIF, WebP
Documents
PDF, DOCX (Word), TXT
Limits:
Up to 10 MB per file.
Up to 5 files per message.
Images are sent to HerdAI so it can see them directly. For documents, Herd extracts the text when you upload the file and provides that text to HerdAI. Each pending attachment shows as a chip with a thumbnail (for images) or a file-type icon, its name, and its size; remove one with the × before sending.
Conversation history
HerdAI keeps a history of your past conversations so you can pick up where you left off.
Open the History panel from the chat header to browse previous conversations, grouped by Today, Yesterday, Previous 7 Days, Previous 30 Days, and Older, or switch to a flat list. A search box filters by title or last message.
Each conversation shows its title, a preview of the last message, when it was last active, and its message count. Select one to reopen it.
Pin a conversation you keep coming back to — pinned conversations are marked with a pin icon so they're easy to spot. Use the same control to Unpin.
From a conversation's ⋮ menu you can Rename or Delete it.
Start a fresh conversation with the New Chat button — in the History panel, in the chat header (once a conversation has started), or with Cmd/Ctrl + Shift + N.
Conversations are titled automatically from your first message, and re-titled if the topic shifts — unless you've renamed one yourself, in which case your title sticks.
Saved prompts
If your team relies on the same kind of request often — a particular style of training, say — you can save it as a reusable prompt.
Open Saved prompts from the chat header to see the prompts your team has saved. For each one you can:
Use it — drops the prompt text into the composer so you can send it (and tweak it first if you like).
Edit or Delete it.
To capture a new one, type your prompt in the composer and choose Save current prompt, then give it a:
Name (required) — e.g. "PhD-style 3-minute presenter training".
Description (optional) — a short hint about when to use it, shown in the list.
Prompt (required) — the full text to send.
Saved prompts are shared across your team, making it easy to standardize how you generate content with HerdAI.
Last updated