Human-centered design systems and product experiences
Core services
01UI/UX Design02Web Development03Software Development042D/3D Experiences05App Development06AI Agents & Automation07Website DesignTechnical services
Frontend Engineering
React DevelopmentNext.js DevelopmentWeb ApplicationsEnterprise DashboardsSaaS PlatformsBackend & Infrastructure
API DevelopmentCloud InfrastructureDevOps & CI/CDBackend SystemsCustom SoftwareIndustries we serve
Technology & Software
AI Meeting Notes and Follow-up Agent Development
Decisions written, actions assigned
An agent for what happens after a meeting.
An agent for what happens after a meeting. From the transcript or your notes it pulls out the decisions and who owes what, proposes the updates to your CRM and drafts the follow-up message for you to check and send.
The value of a meeting is decided in the hour after it. Either someone writes down what was agreed, updates the record and sends the summary, or the next meeting starts by reconstructing this one. Transcription tools made the recording easy. They also produced pages of text that nobody reads, with the three sentences that mattered buried on page six.
This page describes a workflow we propose to build as a custom project. It does not exist as software yet. It would reuse the transcript handling from our voice demonstration and the proposed-then-approved pattern from our invoice demonstration. Extracting decisions and tasks, proposing CRM updates and drafting follow-ups are not built.
How the Work Is Done Today
Transcripts nobody reads
Forty pages of text, and the decision is somewhere in the middle.
Tasks without owners
"We should look into that" and nobody does.
The CRM a week behind
The deal stage still shows what was true before the meeting.
Follow-ups sent late
The summary arrives after the customer has forgotten the details.
Summaries that invent
A tidy summary includes a commitment nobody made.
Recording without consent
A call is recorded and one participant was never told.
Six jobs it carries through
Decision extraction
Finds what was decided.
- Each decision as a sentence
- The words that support it, quoted
- Who agreed
- What was explicitly left open
Task extraction
Finds who owes what.
- The task, the owner and the date
- Tasks for your side and theirs
- Tasks with no owner flagged
- Nothing invented to look complete
CRM update proposal
Proposes changes to the record.
- Stage, next step and close date
- New contacts mentioned
- Notes on the account
- Shown as proposals for approval
Follow-up draft
Drafts the message to the customer.
- What was agreed and what happens next
- In your tone
- Only statements supported by the meeting
- Saved as a draft
Task hand-off
Puts tasks where work is tracked.
- Creates tasks after approval
- Assigns the named owners
- Links back to the meeting
- Avoids duplicates of existing tasks
Consent and retention
Respects recording rules.
- Uses only meetings recorded with consent
- Applies your retention period
- Excludes meetings you mark private
- Deletes transcripts on schedule
From the First Signal to the Record
An example of one case, in 6 steps. The same steps play in the picture at the top of this page.
Meeting ends
A sales call ends. Its transcript is available, and all participants were told it was recorded.
Decisions found
Two decisions are extracted, each with the sentences that support it.
Tasks listed
Three tasks are found. One has no clear owner and is marked as such.
Updates proposed
The agent proposes moving the deal a stage forward and adding a new contact.
Draft written
A follow-up email summarises what was agreed and lists who does what by when.
Approved and sent
The rep corrects one date, approves the CRM update and sends the email.
Walk One Real Case Through It
Thirty minutes, one case from your own process. We map it against the workflow on this page and tell you plainly what can be reused from our demonstrations and what has to be built for your systems.
The Systems Around It
The systems named here are the kind this agent works with. Which ones it connects to in your case is settled in the scope.
Meetings
- Zoom
- Microsoft Teams
- Google Meet transcripts
- or typed notes
CRM
- Deals, contacts and account notes
Tasks
- Your project or task system
Mailbox
- Where follow-up drafts are saved
Policy
- Recording consent and retention rules
Templates
- Your follow-up format and tone
Nothing Leaves the Meeting Without Your Check
A summary that adds a commitment nobody made is worse than no summary. Every decision and task the agent lists is tied to the words that support it, and anything unclear is marked unclear. CRM changes are proposals, and the follow-up is a draft.
It processes only meetings that were recorded with the participants' knowledge, and it follows the retention period you set.
- A supporting quote for every decision
- Unclear owners marked unclear
- CRM changes as proposals
- Follow-ups sent by a person
- Consent before processing
- Transcripts deleted on schedule
Straight Answers
How is this different from a transcription or note-taking tool?
Those tools produce the transcript and a general summary. This agent starts where they stop: decisions with their evidence, tasks with owners, the changes your CRM needs and a follow-up in your format, each waiting for your approval.
How is this different from the meeting preparation agent?
Preparation happens before and reads your history with the customer. This agent works after and reads the meeting itself. What it records becomes part of the history the next brief is built from.
Does it join and record our calls?
It works from the transcript your meeting tool already produces, or from notes you type. Whether and how meetings are recorded, and how participants are told, stays with your meeting tool and your policy.
What exists today?
Not as software yet. This page describes a proposed workflow. It would reuse the transcript handling from our voice demonstration and the proposed-then-approved pattern from our invoice demonstration. Extracting decisions and tasks, proposing CRM updates and drafting follow-ups are not built. A project starts by mapping your process, then proving the first steps on test data before anything touches a live system.
Send Us Three Transcripts You Already Have
We will show the decisions, tasks and follow-up the agent would have produced, next to what was actually sent.