The conversation does the work
ANVL's conversation now knows what you have already made, puts a whole plan in one card, makes images in your style, fact-checks a deck, changes your brand and invites a colleague. An assistant that does the work instead of describing it.
Most AI assistants are good at telling you how to do something. To organise your files, start by creating folders for each client. Thank you. The work is still yours.
We want ANVL's conversation to be the other kind: the one where you say what you want done, see exactly what will happen, and press once. It already worked that way for single actions. Over the past weeks it has learned to do a lot more of the job.
It knows what you have made
What have we written about before? and which social posts should I do this month? are answered from your real decks and posts, not from guesses. Advice comes back numbered, so do number two in the next message builds it. Ask have we done anything on the autumn campaign? and the conversation searches the text in your projects, not just the titles, and answers with the project, the slide and an excerpt.
A thin request gets a follow-up question before anything is built, at most three, with suggestions you can tap. What is already in your brand kit, it never asks about.
A whole plan in one card
Help me organise all my content into folders used to give one card at a time, and after the first one the conversation ran out. Now the whole plan comes in one card: the folders with the material that goes in each, new folders marked, and what you might want to throw away as its own group. Every row has a checkbox, so you can leave out what should stay put. You approve once, and the card then shows which rows are done. Anything moved to the bin can be restored.
Images, made and chosen in the thread
Ask for an image, a café in autumn light, and you get a card with the subject. It says what happens before you press: one credit, your brand's image style, and that the image is saved as a draft in your image bank. If no image comes out, you get the credit back.
Then ask for something with it, can we use it on a poster, and it is already in the proposal. Proposals have a row for images where you add from the bank, upload or generate, and remove with the cross on the thumbnail.
Jobs around the workspace
Some of the most useful things are not about making material at all.
- Fact-check a deck. Fact-check yesterday's deck points out claims without a source, slide by slide, using the same reviewer that runs when a deck is built. It costs no credit. If one of them is actually right, say so, and ANVL remembers it.
- Change the brand. Change the primary colour to #1a5c3a, use Inter for headlines or make the tone more direct gives a card with before and after. A colour is checked against the background, and if it does not pass, the card suggests the nearest shade that does. Every change keeps a version, so it can be taken back.
- People. A workspace admin can invite someone, change a role or send a new activation link. The card says the address, the role and whether an email goes out, before you press.
- Rewrite in your tone. Paste a text and ask for it in your voice. You get the rewrite on its own with a copy button, and underneath what changed and why. Nothing is added that was not in your text.
Why it matters
If you are evaluating ANVL: the difference between an assistant and a tool is whether the work is done when the conversation ends. Here, it is. And because every action is a card that states exactly what will happen, you stay the one who decides.