AI can prepare the follow-up while the meeting is still fresh.

A useful follow-up begins with what was actually decided.

The meeting title rarely explains the outcome.

The important sentence may be halfway through the call: the renewal price is changing, Maya owns the security questionnaire, and the customer expects a reply on Thursday.

An agent can read the transcript or notes, separate decisions from discussion, and return a short list with the supporting language beside each item. That makes it easier to spot a false assumption before it becomes a task with someone else's name on it.

One call can produce several different next steps.

A sales call may need a customer recap, a product issue, an internal handoff, and a calendar hold. Sending the same summary to everyone makes the customer read internal detail and leaves the product team without a precise problem statement.

Strawberry can prepare each artifact from the same source material: an external email with the agreed next step, an internal note with risks, and task drafts for the people who own the work. The meeting lead decides which artifacts are real and which should be discarded.

The follow-up should preserve the context behind the promise.

“Send the updated plan” is not enough context three days later.

Which plan, what changed, and who said the timeline was feasible are all buried in the call unless the follow-up carries the relevant detail forward.

A companion can use the tabs and connected work context you choose to check the prior proposal, the account history, or the project plan before it drafts. That helps it write a follow-up that names the March scope change instead of a vague promise to “circle back.”

Approval is where meeting automation stays trustworthy.

A transcript can be wrong, a speaker can change their mind, and a call can include a sensitive aside that does not belong in a follow-up. The agent should do the sorting work, not quietly publish its interpretation.

Use a saved meeting-close skill for the recurring parts, then make it a routine after a weekly pipeline review or project stand-up. Keep sending, task creation, and any update to a customer record behind an approval step when the consequence belongs to another person.

Frequently asked questions