Ennie AI

Ask which segment is shrinking, which channel drove conversions, or how last week compared to the week before. Ennie answers from the campaign, audience, content and performance data already in your account rather than from general knowledge, so the answer is about your client, not the category.

Answers are shaped by where you ask them. Ennie is passed the screen you are on and the data on it, so the same question asked on Audience and on Campaigns comes back framed for that work, with different follow-up actions offered. On Attribution, an answer can carry a note listing which of your data sources it drew on.

Alongside answers, Ennie suggests actions it could take from the current screen. Nothing runs until you confirm it.

• In the product — a chat panel you open from Campaigns, Content, Audience, Insights, Funnels and Workflows. It is labelled with the area you are in and is given that screen's data as context.
• Analytics and Attribution — Ennie has its own tab alongside the usual reporting views, so a question sits next to the charts rather than in a separate tool. On Analytics the tab also carries conversational reporting on message volume, sentiment and recurring topics.
• WhatsApp — a verified number talks to the same assistant. Replies can include tappable buttons or a short menu, so you can pick an option instead of typing it out.
• Slack — install the app to your workspace and direct-message Ennie. Replies come back formatted for Slack, in-thread, with up to five action buttons you can press without leaving the channel.

Under Settings you choose which channels you want to be reachable on: an email brief with a real-time, daily or weekly cadence, WhatsApp updates against a verified number, and Slack once the workspace app is installed. The Dashboard carries a setup entry point rather than a chat panel of its own.

Conversations with Ennie are not private to whoever started them. Every conversation, whether it began in the product or on a messaging channel, shows up in a shared inbox as a thread the team can find. You can filter by status, by unread, or by which assistant is involved, search the text, and narrow by date, so a colleague can pick up a thread from Monday without you forwarding anything.

Threads have named participants. You invite a teammate onto a thread as an owner, a collaborator or a viewer, and that role decides what they can do — a viewer can follow along, a collaborator can act. Roles can be changed and people can be removed later. Threads can be marked read, archived when they go quiet, or resolved when the work is done.

Anything Ennie produces inside a thread waits for a person. Drafts and assets sit in the thread for review and are approved or rejected by a named participant who holds that permission, and the decision is recorded against them. Campaigns waiting on a decision also collect in an approval queue, where you can see what is pending, who raised it and how long it has been sitting, then approve or send it back.

Ennie works on one brand at a time. Ask about revenue while your coffee client is active and you get that client. There is no way to phrase a question that pulls another client's numbers into the answer, which matters when the same team runs several accounts.

On WhatsApp and Slack the brand is part of the conversation. You can name a brand, switch to a different one mid-conversation, or clear it — and once it is cleared, Ennie waits for you to say which brand before it will answer anything else. Each switch is kept with the time, the brand you moved from, the brand you moved to and the reason given, so the thread shows which client each answer belonged to.

The practical rule: switch the brand before you ask, not after. An answer only ever describes the brand that was active at the moment you asked.

On the messaging channels Ennie holds the thread of a conversation rather than treating each message as a fresh start. It keeps the recent back-and-forth in view, carries short summaries of the past week's conversations for the brand you are in, and builds a picture over time of how you work — the accounts you spend time on, the questions you come back to and the metrics you tend to ask for.

What that buys you: you can say "and last week?" instead of restating the whole question, and you can come back the next morning without re-explaining who you are or which client you mean. An idle conversation eventually closes, but the summaries and the picture of how you work outlast it, which is why the second conversation is shorter than the first.

Campaign funnels. From Funnels, describe what you want and Ennie asks for the missing pieces — objective, target audience, budget, channels — then generates a full funnel: named stages across awareness, consideration, decision and retention, the steps inside each stage with a channel and platform, suggested headline, body and call to action per step, a budget split across the stages, and target metrics. Generation runs in the background with a progress indicator and drops you into the finished funnel. The same window can instead take a competitor's URL and build a funnel from what it finds there.

Automations. From Workflows, Create An Agent with Ennie AI opens the chat with your existing agents as context. Describe the job in plain language — monitor something, alert on a threshold, analyse, optimise or report — and say which data it should watch, what should set it off, and where you want to be told. Ennie turns that into a working automation, checks it over and shows you warnings and suggestions before it can go live, and lets you preview it first.

Describing what you want is the recommended route into an automation today. Starting from a template is the alternative, and bringing in an automation file you built elsewhere is not available yet.

1. Ennie only knows what you have connected. With no commerce, ad or email platform linked to the account, there is little to ask about and the answers will be thin.
2. Answers cover the active brand only. If a number looks wrong, check which brand is active before you check the data.
3. The longer memory sits with the messaging channels. The in-product panels carry the current screen and the conversation you are having, which is a shorter memory than WhatsApp or Slack have.
4. Ennie recognises common date phrases and the usual metric names when it tracks what a conversation is about. Unusual wording may not be picked up as something to carry forward.
5. Searching back over past conversations matches on wording rather than meaning, so a differently phrased question may not surface an earlier answer.
6. Generated funnels and automations are drafts. They come back complete enough to edit, not signed off — check budget splits, targets and copy before anything runs.
7. Building an automation needs an AI provider configured for the account. Without one the request stops and tells you to set one up in settings.
8. Conversations are kept. Messages are stored per account and per conversation for history, for the shared inbox and for the conversational reporting.