Mindlyft docs
Reference documentation for how Mindlyft runs post-call execution: what each concept means, how the approval layer decides, and how writes reach your CRM, tickets, and email with a human in the loop.
Concepts
06 docs
- Human in the Loop, On the Loop, In CommandThree oversight models get used interchangeably and mean different things. In the loop approves each action, on the loop supervises and can intervene, in command decides whether the system runs at all. Which one you have decides what an agent can do to you.
- Post-Call ExecutionPost-call execution is the work a customer conversation creates: the CRM update, the ticket, the follow-up, the handoff. Here is what it means and the exact steps Mindlyft runs to close that loop with a human in it.
- The Commitment TaxonomyA commitment is any obligation a call creates. Mindlyft classifies commitments into a fixed taxonomy so each one maps to a concrete, executable action. Here are the types and how they turn into work.
- AI Agent Audit TrailAn AI agent audit trail is the record of every action an agent took, who approved it, and how to undo it. Here is how Mindlyft logs each write in a tamper-evident trail and keeps it reversible.
- Scoped Agent IdentityA.S.T.R.A. never runs as a shared service account or a borrowed human login. Each agent gets its own scoped identity with only the permissions its job needs, so every write is attributable and any one agent can be turned off without touching the rest.
- Trust and Security: The Architecture Behind A.S.T.R.A.Mindlyft's trust story is architectural, not a certificate: your systems of record stay yours, the credentials A.S.T.R.A. uses are scoped to your org and granted by your admin, consequential actions wait for a human unless your admin opts a low-risk type into automation, every write is logged in a tamper-evident trail, and the controls are tested in published security reviews. Here is how the pieces fit together.
How it works
15 docs
- The Approval Layer: How Approval WorksThe approval layer is the gate between an AI agent's draft and a real write to your CRM, email, or customer. Here is how Mindlyft's A.S.T.R.A. decides what needs a human, and why every write stays reversible.
- Post-Call CRM Updates, With ApprovalAutomating post-call CRM updates means turning a transcript into drafted field and stage changes that a human approves before they write. Here is exactly how Mindlyft does it in Salesforce and HubSpot.
- CRM Hygiene AutomationCRM hygiene automation keeps records clean without a rep doing data entry: deduplication, field enforcement, and activity capture, run continuously with a human approving the risky merges.
- Sales-to-CS Handoff AutomationThe sales-to-CS handoff is where context and promises get lost. Here is how Mindlyft carries the commitments made during the sale across the close, so customer success starts with the full picture instead of a blank record.
- Renewal Risk FlaggingRenewal risk flagging surfaces the accounts likely to churn while there is still time to act. Here is how Mindlyft turns signals from calls and product usage into an early, reviewable flag on the record.
- Customer Onboarding AutomationCustomer onboarding automation runs the steps between a signed deal and an adopted account, tasks, follow-ups, and internal coordination, with a human approving anything the customer sees.
- Approve, Edit, or Reject a Drafted ActionEvery consequential action A.S.T.R.A. drafts lands in a review queue as a diff: the current value, the proposed value, and the line of evidence behind it. Here is exactly what you see, what each decision does, and what happens after you make it.
- Follow-Up Emails After a Sales CallThe follow-up email after a sales call is one of the four things Mindlyft executes post-call. Here is what a good one contains, when it should go out, and how Mindlyft drafts it from the call while keeping a human in front of Send.
- Jira Tickets From Customer Calls: How Mindlyft Creates ThemA Jira ticket is one of the four things Mindlyft executes after a call, alongside the CRM update, the follow-up email, and the internal handoff. Here is what triggers it, what goes in it, and why it does not wait on the same approval gate as a customer-facing send.
- How Approval Requests Reach You in SlackAnything Mindlyft drafts that needs a human yes, a follow-up email, a CRM stage change, a record merge, arrives in Slack instead of a separate approvals dashboard. Here is what shows up, who sees it, and what happens if nobody responds.
- AI Agent Governance: The Control PlaneHow AI agent governance is implemented as a control plane over agentic GTM systems. Covers the control surfaces, the agent inventory, what agent-aware governance means, how to audit it, who owns it, and the failure modes that show up in practice.
- Migrating a Lead Spreadsheet Into a Live CRMData migration into a live backend, moving a spreadsheet of leads or accounts into Salesforce or HubSpot, is one of the six request types the Mindlyft subscription covers. Here is exactly how a migration is scoped, mapped, deduplicated, and approved before it writes.
- What Happens to a Commitment Nobody Acted OnA commitment drafted from a call doesn't vanish if nobody reviews it, and it doesn't execute on its own either. Here is exactly what happens to a stage change, a follow-up, or a ticket that sits unactioned, and how an unmet commitment becomes a renewal-risk signal.
- Tracking Commitments Across Multiple Calls on One AccountA real account is rarely one call. Here is how Mindlyft carries the commitments, context, and open next steps from call one forward to call five, so later calls build on what was already promised instead of starting over.
- Commitments to Multiple Stakeholders on One CallA call with the champion, the economic buyer, and an end user on the line rarely produces one commitment, it produces several, each owed to a different person. Here is how Mindlyft keeps that distinction instead of collapsing it into one generic account note.
Reference
04 docs
- Integrations: Tools and How Writes WorkMindlyft executes inside your own stack through scoped integrations: Salesforce, HubSpot, Jira, Gmail, and Slack, coordinated over MCP. Here is what connects, which direction data moves, and where the human sits.
- Which A.S.T.R.A. Actions Need ApprovalNot every action A.S.T.R.A. drafts waits for a human. Reversible, internal writes run on their own and get logged; anything that touches a customer or a system of record waits for a yes first. Here is the line, item by item.
- How the Mindlyft Subscription Queue WorksMindlyft runs on a flat subscription with unlimited requests in the queue and one active task shipping at a time. Here is exactly what counts as a request, how billing and rollover work, and what happens if you pause or cancel.
- Acquire vs Expand: Choosing a Mindlyft MembershipMindlyft sells two memberships, each with its own price: Acquire for everything before a customer signs, Expand for everything after. Here is what each one actually covers, and how to decide which one your team needs first.
The through-line across every doc
Whatever the concept or integration, the rule is the same: an AI agent drafts the work, and a human approves anything that reaches a customer or changes a system of record. Every write is logged and reversible. Start with a free 30-minute GTM Engineering Review.
Get your free reviewFAQ
Questions
What is post-call execution?
Post-call execution is the work a customer conversation creates after it ends: the CRM fields that need updating, the ticket that was promised, the recap email, the handoff note. On most teams that work scatters across several tools and survives on one person memory. Post-call execution is the practice of capturing those commitments from the call itself and turning each one into a tracked action in the system where it actually has to happen.
Does A.S.T.R.A. take actions without a human approving them?
Not for anything that reaches a customer or touches a system of record. Those writes are held for a human yes, and that gate is structural rather than a setting someone can switch off. Reversible internal writes can run unattended, and even those record a receipt that can be rolled back.
Which systems does A.S.T.R.A. write to?
Salesforce, HubSpot, Jira, Gmail, Slack, Google Calendar, Zoom, Gong, Apollo, Clay, n8n, Zapier, Make, PostHog, and Amplitude, with OpenAI and Anthropic as the model layer. The work runs inside your own accounts under permissions you scope, so an agent cannot widen its own access.
Do we have to migrate onto a new platform?
No. A.S.T.R.A. is not a stack you move onto. It runs inside the tools your team already works in. That constraint is deliberate: a system that requires migration before it delivers anything has to be right about your future stack, and it usually is not.
What happens to the systems if we cancel?
They keep running. Everything is built inside your own accounts, your CRM, your automation tools, your repositories, so cancelling ends the engineering relationship rather than the automation. There is no minimum term, and billing runs in 4-week cycles.