Just stumbled on this while digging through Sembly's workspace admin settings. The feature is buried, but you can actually define custom templates for what its AI spits out after a meeting.
Everyone complains about the generic "Key Points" and "Action Items" format. Turns out you can force it to follow your own structure. For example, our sales team needed:
- Immediate next step (owner/deadline)
- Client's stated pain point (direct quote)
- Competitive mention (yes/no)
- Sentiment shift during call
You build these templates in the workspace settings with a mix of plain text and their placeholder variables. It's not as flexible as a full scripting engine—you're mostly rearranging their existing data points—but it does let you prioritize what matters.
Biggest caveat: The UI for this is separate from the personal "Summary Preferences." If your admin hasn't set a workspace template, you're stuck with the default. Also, changing the template only affects new meetings, obviously.
Worth checking if your admin has access. Has anyone else set this up? I'm curious if you've found clever uses for the variable system beyond the obvious ones like `{action_items}` or `{key_points}`.
-- CRM Surfer
Your CRM is lying to you.
The fact that they buried this is the most telling part. It's the classic vendor playbook: ship a mediocre generic feature, then hide the configuration that would actually make it useful behind an admin wall.
You can bet the placeholder variables are just repackaging the same raw data they already extract. The "flexibility" is an illusion. It's just a different sorting algorithm for your data before they lock it into their format.
I'd be more impressed if you could pipe the raw transcript and analysis into your own system via an API, instead of just playing with their summary templates. That's where you'd get real value, but that's also where they'd lose control.
-- cost first
That's a fantastic find. I've been wishing for this exact feature. The generic summaries make it hard to standardize handoffs between teams.
For our customer success squad, we built a template that pulls in the `{sentiment_score}` variable to flag at-risk calls automatically. It looks for a drop below a certain threshold and places a warning at the top. We also use `{topics_mentioned}` to auto-tag the conversation for our internal knowledge base.
The biggest hurdle we hit was that the placeholders for speaker names and timestamps were a bit rigid. You can get `{speaker_1}` but formatting a clean participant list took some trial and error. Still, it's a massive step up from the default. Did you find any undocumented placeholders while you were poking around in there?
You're right about the control aspect, but I think the API limitation is more about operational cost than malice. Transmitting raw, structured data continuously at scale for every meeting is a significant bandwidth and processing overhead compared to delivering a cooked summary.
That said, the template feature actually does expose a limited API of sorts. The placeholder variables are the data points they're willing to guarantee. Building your template is essentially defining a contract for the output schema. It's a common first step before opening up a full JSON feed, because it forces them to standardize those internal data structures. The cynical view is vendor lock-in, but the practical one is that they're stress-testing their own data consistency before externalizing it.
Your point about piping the raw data into another system is the ideal end state for real workflow integration. The custom templates are a useful stopgap for teams that need immediate consistency, but they don't solve the problem of getting that data into a CRM or a project management tool automatically. The templates live inside their walled garden, as you noted.
Plan the exit before entry.
Agree on the template-as-contract point. That's exactly why we pushed for this feature last year - to force them to commit to data points like `{next_step_owner}` and `{client_quote}`.
The bandwidth argument doesn't fully hold up. They're already processing and storing the raw data. It's not about cost, it's about product scope. Exporting structured data becomes a whole new product tier they have to support.
They'll add the API when enough enterprise clients demand it and are willing to pay 5x more. Until then, the templates are the only lever you have. Use them to define what you need, then build a manual process around them. It's messy, but it works.
Optimize or die.