Launched this week
Katalyst is the AI sales agent for teams on Salesforce. Hang up a call and it's already done: notes summarized, records created, fields updated, follow-up drafted, next step set. It runs 24/7, reading every call, email, and calendar, surfacing the right signals, prepping you for each meeting, and prompting you to act: follow up here, this one’s slipping. What's new: AI Resolution on every account, meeting recorder, hygiene scores, deal patterns. Built for enterprise sales teams.
For enterprise sales teams with complex Salesforce setups, custom objects, non-standard field names, and org-specific process rules, how much configuration work happens before Katalyst is actually writing to the right places? "Records created, fields updated" assumes the field mapping is solved, which on a heavily customized enterprise org is rarely trivial.
Katalyst
@ansari_adin Katalyst reads your live Salesforce schema directly, so custom objects and non-standard names surface as-is, no fixed template to map against. You just scope which record types and fields it can touch. Config is setting boundaries, not hand-mapping.
@divyansh_lohia Reading the live schema directly is the right approach. Curious what happens when the schema has 200 custom fields across a mature enterprise org, does Katalyst surface all of them and expect the admin to scope down, or is there some inference about which fields are actually active and relevant based on recent usage patterns?
Katalyst
@ansari_adin one detail that matters a lot on a heavily customized org: those boundaries aren't checked once at setup, they're re-checked at the moment of every single write. Each field has its own permission level, and if your admin flips a field to read-only after Katalyst has already drafted a suggestion for it, the write gets blocked at accept time. Same if Salesforce itself marks the field non-updateable. So config drift, which is constant on enterprise orgs, doesn't turn into wrong writes. The current state of your org always wins over whatever was configured earlier.
How does it actually handle calls when reps work from a softphone outside of Salesforce, does the recorder hook in at the carrier level or only on certain devices?
Katalyst
@hanifefndkpqon Good question, right now the recorder covers virtual meetings (Zoom, Meet, Teams, Webex) rather than softphone or carrier-level call capture. Dialer integration is on our radar, but I'd rather tell you straight where it is today than oversell it. What softphone is your team on? Helps us prioritize.
Katalyst
@hanifefndkpqon To add, dialer capture is one of the more requested asks we're tracking, so knowing which softphone your team runs genuinely shapes what we build next. For now the recorder shines on virtual meetings, but this feedback is exactly what moves it up the list.
the real question isn't whether AI can update Salesforce for you. it's whether sales reps will actually trust it enough to stop doing it themselves. the hardest habit to break isn't the admin work, it's the control. curious how long it takes before a rep stops double-checking what Katalyst wrote and just trusts the output
Katalyst
@tina_chhabra You nailed it, the barrier was always control, not admin. So we don't ask for trust up front: the rep stays in the loop, sees the reasoning behind every write, and watches it get their deals right. The double-checking fades on its own once accuracy's undeniable, usually fast
Katalyst
@tina_chhabra To add from the build side, this is exactly why we made the reasoning visible instead of hiding the write behind a black box. Trust comes from watching it get your deals right, not from being told to trust it. Reps stop double-checking when the accuracy earns it, and that's the only way it should happen.
How does the AI Resolution actually work when two reps on the same team have different takes on a slipping deal, does it surface both viewpoints or just pick one?
Katalyst
@toprakmdrk Good question, and there's a nuance that makes it cleaner than it sounds. Each rep's agent works off their own context, their emails, their calls, so AI Resolution is rep-specific by design: two reps on the same deal each get a read from their own vantage point, not a blended one. The underlying deal and pipeline data is shared at the org level, so everyone's working from the same source of truth, but the resolution reflects what that rep has actually seen. And when it comes to writing back, it maps to the opportunity and account owner, so the record has a clear source rather than two agents fighting over it.
Katalyst
@toprakmdrk To add, the key is that the shared data and the per rep read don't conflict, they layer. Everyone works off the same org level source of truth, but each agent's take reflects what that rep has actually seen. And since write-back anchors to the owner, you never get two agents overwriting each other on the same record.
the post-call automation angle makes sense. curious how it handles messy or overlapping deals where context isn't clean. that's usually where CRM tools break down
Katalyst
@shubham4real That's usually where tools break, agreed. Overlapping deals stay clean because signals and write-back key off account and opp ownership, so no duplicate or conflicting records, the owner anchors what's written. And when context is genuinely messy or thin, it surfaces its read with reasoning instead of guessing, leaving the unclear calls to the rep.
Katalyst
@shubham4real To add to Div, the ownership anchoring is what quietly does the heavy lifting here, it's why overlapping deals don't turn into duplicate or conflicting writes the way they do in most tools.
Katalyst
@ridhwikvinod Same place you did, and for the same reason, one bad write erodes trust you don't get back. So it's approve-before-write by default: the agent queues the change with its reasoning, rep confirms. The one thing we added is making it per-field, so an agent that's proven itself on some fields can write those directly while high-stakes ones stay approval-only. Trust gets earned field by field, not flipped globally.
Katalyst
@ridhwikvinod One thing to add since you run sales ops. Every queued change shows its receipts: the exact call or email it came from and why the agent thinks the field should change. So approving isn't another chore, the rep just glances and confirms. And on the write itself, Salesforce is always written first. If Salesforce rejects the change for any reason, like a validation rule, nothing updates in Katalyst either. The two can never disagree, so there's no cleanup job waiting for you later.
Finally tried a demo and the auto-drafting of follow-ups after calls is genuinely useful, not just templated fluff. Curious how it handles messy transcripts with multiple speakers on the same deal.
Katalyst
@cemlpsj Glad the follow-ups landed, human not templated is the bar. Multi-speaker calls hold up well, and since each summary shows its source, anything that drifts is easy to catch.
Katalyst
@cemlpsj To add, the follow ups reading human instead of templated is exactly the bar we set, so glad it landed. And on messy multi speaker calls, showing the source behind each summary is what keeps it trustworthy, you can always see where a line came from rather than taking it on faith.