AI
AI replies from your notes.
On Growth and Business, a rule can write the reply. You set the goal, the instructions, the knowledge note, the length, and the fallback line.
3 minute read
What it uses
The model sees the incoming text and the note you saved. Put prices and links in the note. Do not expect it to know a price you never wrote down.
The default length cap is 300 characters. You can set it from 1 to 2,000.
Credits
Growth includes 1,000 AI credits a month. Business includes 2,500. A test draft reserves 1 AI credit; provider failures may still consume it. Free and Pro do not include AI.
Use AI as an optional writer inside an existing rule
DripDM AI is not a new permission to contact people. It is a way to compose reply text from a goal, instructions, pasted knowledge, and incoming text within an eligible automation. The live service can also include recent conversation history; do not claim that it never sees prior private messages. Growth and Business enable this capability, while Free and Pro retain fixed replies. Credit usage and provider availability are separate from the underlying ability to match a rule and send an eligible message.
A workflow to try
Start with a narrow topic such as service hours and booking directions. Supply the current hours, destination URL, and conditions you can confirm. Set the goal to answer the question and point to booking, not to invent availability or complete a reservation. Write a fallback for unsupported questions and a static reply for the non-generated path. Preview routine, ambiguous, and conflicting requests before turning on the rule. A preview reserves one credit; it is not a free unlimited playground.
Checks before you change the setup
Maintain knowledge when prices or policies change and review live text. Instructions improve the intended behavior but do not guarantee correct facts or prevent every invented discount. This public content does not claim website crawling, link allowlists, a special AI receipts dashboard, or an automatic human-takeover timer. Topic matching and generation can each reserve credits. At exhaustion, the service can use static text. Meta messaging eligibility, connection, and recipient constraints still govern whether any reply can be delivered.
A launch checklist for optional generation
Assign an owner for the facts, an owner for the campaign, and a person who can pause misleading replies. Keep one narrow goal, current knowledge, safe static text, and fallback text together. Test ordinary, missing-fact, and conflicting requests, then review credit movement and actual threads. If an action belongs to a booking or commerce system, say so explicitly. Maintain a recovery route back to fixed text if generation is not useful. This checklist makes AI an optional capability in a working messaging system rather than a dependency that hides missing facts, a broken destination, or an ineligible send behind fluent language.
Questions
Yes. A rule can fire when the message is about a topic you describe, not only when the text is an exact keyword.
Generated text can still be wrong. Explicitly prohibit unauthorized discounts in the instructions, maintain current prices in the note, and test difficult requests. Keep fallback and static text safe, and review live replies.
The supported workflows here compose replies and send configured destinations. They do not read a live calendar, hold inventory, finalize a purchase, or confirm a reservation merely by mentioning it. The linked booking or commerce system remains responsible for those actions and their confirmation.
Review it whenever a price, policy, date, location, or destination changes, and when real questions reveal missing facts. Keep a named owner for those updates and rerun relevant previews instead of relying on an old successful sample.