CRMy Raw Context Test Fixture
Source type: call transcript
Suggested source label: Northstar pilot planning call
Suggested subject records: Northstar Labs, Maya Patel, Northstar Agent Context Rollout
Occurred at: 2026-05-27T18:00:00Z

Use this transcript to test Add Context. It is designed to produce Signals for
stakeholders, risks, commitments, next steps, evaluation criteria, timeline,
business impact, and governed writeback readiness.

---

[00:00] Cody Harris, CRMy:
Thanks for making time, Maya. I want to make sure we are focused on the parts of
the Northstar Agent Context Rollout that matter before the pilot decision. Last
time you mentioned governed writebacks, scoped agent tools, and human approval
before anything reaches HubSpot or Salesforce.

[00:21] Maya Patel, Northstar Labs:
That is still the core evaluation. Our revenue leadership wants agents to help
the team keep customer context current, but nobody is comfortable with agents
writing directly into Salesforce without a clear review step.

[00:42] Cody:
Understood. Should we treat safe writeback governance as a required pilot
criterion?

[00:49] Maya:
Yes. If CRMy can show that raw customer context becomes evidence-backed Signals,
and only trusted Memory can drive writeback requests, that would answer a big
concern. The approval trail matters too. Finance and RevOps will ask who
approved each change and what evidence supported it.

[01:18] Cody:
Who else needs to be part of that review?

[01:23] Maya:
Elena Ramirez from Security needs to review the data residency story. She is not
blocking the pilot yet, but she will block production if we cannot keep customer
context in our environment or prove that secrets are never exposed in logs.

[01:47] Cody:
So Elena is the security reviewer, and data residency plus secret handling are
the main risk areas.

[01:55] Maya:
Correct. Also, David Kim in Finance needs to see a business case. He is asking
whether this saves seller time or just creates another admin surface. If we can
show fewer manual CRM updates and faster follow-up after customer calls, I think
he will support the pilot budget.

[02:25] Cody:
Is David the budget approver?

[02:30] Maya:
Not final approver. He can recommend approval, but the final pilot spend needs
sign-off from Priya Shah, our CRO. I can sponsor the evaluation internally, but
Priya owns the final go/no-go.

[02:54] Cody:
That distinction helps. Maya is evaluation sponsor, David influences Finance,
and Priya is final executive approval.

[03:02] Maya:
Yes, that is right. Please do not mark me as the economic buyer. I can push the
evaluation, but Priya owns the buying decision.

[03:18] Cody:
Got it. What would make the workshop successful next week?

[03:24] Maya:
Bring a live example with Northstar data. Show a call transcript creating
Signals, show two sources supporting the same Signal, then confirm it as Memory.
After that, show a governed writeback request that updates Salesforce only after
a human approval. That is the story our team needs.

[03:58] Cody:
We can do that. Would Thursday, June 4, 2026 work for the workshop?

[04:04] Maya:
Yes, June 4 works. I can do 10:00 AM Pacific for 60 minutes. Please include our
RevOps lead, Sam Torres, and Elena from Security. David may join if you include
the business impact section.

[04:29] Cody:
I will send a follow-up with the agenda today and include Sam, Elena, and David.
We will prepare a Northstar-specific demo showing Raw Context to Signals,
Signals to Memory, Handoff approval, and Salesforce writeback.

[04:48] Maya:
Perfect. One more thing: we are comparing this against a generic agent memory
prototype that our data team built in Databricks. It stores summaries, but it
does not have record ownership, evidence, or approval gates. If CRMy can show
the operational layer between agent inference and CRM action, that is the
stronger case.

[05:20] Cody:
That is useful. So the competitive point is not generic memory. It is trusted
operational memory with evidence, policy, and writeback controls.

[05:31] Maya:
Exactly. I want the system to help reps act quickly, but I also need RevOps to
trust that agents are not inventing CRM truth. If a claim is uncertain, keep it
as a Signal. If it is supported and confirmed, make it Memory. If it affects
forecast, budget, security, or CRM writeback, route it through a handoff.

[06:04] Cody:
That maps cleanly to CRMy. Anything else that could block the pilot?

[06:11] Maya:
Timeline could become an issue. We want to decide by June 20, 2026 because Q3
planning starts the following week. If the workshop slips, the pilot probably
moves to July.

[06:34] Cody:
I will treat June 20 as the target decision date and June 4 as the required
workshop date.

[06:41] Maya:
Yes. Also, please send the security architecture overview before the workshop.
Elena will not join cold.

[06:52] Cody:
I will send that by Friday, May 29, 2026.

[06:57] Maya:
Great. If you send the agenda and security overview, I will forward it to Priya
and David with my recommendation that we proceed to the pilot workshop.

[07:12] Cody:
Thanks, Maya. I will follow up today with the meeting invite, agenda, security
overview owner, and demo plan.

[07:20] Maya:
Thanks. This is promising if the governance pieces are real.
