Feature · Process Automation Technology

Process automation technology that lives inside the system of record.

Process automation technology is the stack underneath the visual flow builder: the event bus, the trigger / action library, the LLM augments, the typed webhook APIs, and the audit log. Strkr Flows ships the whole stack native to the CRM so automations run next to the data they operate on.

Start free trial See pricing
The architecture

What the technology underneath Flows actually is.

The sections below map the architecture of Strkr's automation stack. The point is not to brag about plumbing but to make the operational and engineering trade-offs visible so a technical buyer can evaluate honestly.

Event bus

Every write emits an event.

Every create, update, and delete on a Strkr record emits a typed event onto the internal event bus. The bus is the single source of truth for triggers. Flows subscribe to event types and fire when a matching event arrives. The architecture guarantees no missed trigger when a record changes.

Trigger library

A typed catalog of 100-plus starting events.

Triggers include deal stage change, lead score threshold, project stage change, invoice overdue, ticket SLA, form submission, email reply, meeting scheduled, and dozens more. Each trigger is typed, so the flow builder knows what fields are available in subsequent action steps.

Action library

The typed catalog of operations a flow can perform.

Actions include create task, send email, send Slack, update field, create project, assign owner, enroll in sequence, send SMS, generate document, and call webhook. Every action is typed and permission-aware; a flow cannot escalate privilege through an action.

LLM augment

An inline Strkr AI call inside the flow.

A flow can call Strkr AI as an action: summarize an email reply, draft the next-step note, classify a form submission, write the first-touch opener. The LLM call runs inside the flow with the deal context pre-loaded. No separate AI SKU, no credit meter.

Webhook ingress + egress

Typed endpoints for external systems.

A flow can trigger on an inbound webhook and act by calling an outbound webhook. The payloads are JSON-shaped and documented. External systems (industry SaaS, custom internal tools) wire into Strkr without a separate integration platform.

Audit + replay

Every flow run is captured and replayable.

Every flow execution logs its inputs, outputs, actor, timestamp, and outcome. The log is queryable by flow, by trigger, and by date. For investigation, a failed run can be replayed with the same inputs after the underlying bug is fixed. The replay closes the operational loop.

Execution guarantees

The reliability properties the platform makes.

The sections below describe the execution guarantees Strkr Flows ships, so an engineering buyer knows exactly what the system promises and what it does not.

At-least-once delivery

A trigger will fire; it may fire twice on retry.

The platform guarantees at-least-once trigger delivery. For most business workflows this is the correct trade-off. For actions where exactly-once matters (e.g., invoice send), the flow is responsible for idempotency, which Strkr makes easy with built-in deduplication keys on common actions.

Latency SLO

Trigger-to-first-action under five seconds.

Strkr's target trigger-to-first-action latency is under five seconds at the 95th percentile. The target holds across the common trigger types. Webhook-triggered flows add network latency on top. The SLO is published; the platform engineering team monitors against it.

Ordering

Events on the same record are processed in order.

Events on the same record are serialized through the event bus in the order they were written. Two rapid-fire stage changes on the same deal produce two flow runs in the correct order. The guarantee prevents the ugly race where the "Closed-Lost" trigger fires after the "Closed-Won" one.

Retries with backoff

Failed actions retry with exponential backoff.

A failed action (e.g., a webhook endpoint returning 503) is retried with exponential backoff up to a configurable limit. After exhaustion, the run is marked failed and surfaced on the admin dashboard. The failure is visible, not swallowed.

Per-tenant quotas

A noisy tenant does not degrade a quiet one.

Each tenant has quotas on concurrent executions and total daily runs. The quotas are generous for the base tier and lift on higher tiers. A runaway loop in one tenant cannot starve the shared platform. The quota is a safety fence, not a pricing meter.

Permission-scoped actions

A flow runs with the triggering user's permissions.

Actions in a flow run with the permission scope of either the triggering user or a configured service principal. A flow cannot bypass record-level permissions by using an action that the triggering user could not have performed manually. The security model is one, not a side policy.

LLM augments in detail

Where AI fits in the flow runtime.

The sections below describe how Strkr AI is wired into the flow runtime, so a technical buyer can evaluate what runs locally, what runs on the model, and what the cost / latency / privacy profile looks like.

In-flow LLM action

Call the model as a typed action.

A flow step can call Strkr AI as an action with a typed input and a typed output. "Classify this email reply" returns a known enum. "Summarize this deal timeline" returns a bounded string. The flow builder sees the output shape and can route on it.

Context pre-loaded

The deal, lead, or project record is in scope.

When a flow calls Strkr AI, the triggering record and its relevant history are pre-loaded as context. The LLM does not need the flow author to pass the full deal timeline; the runtime does it. The author writes the prompt; Strkr handles the context assembly.

No credit meter

The LLM action is on the paid tier, included.

Strkr does not meter the LLM calls per flow run. The LLM augment is included in the paid tier at the same seat price. Tenants running thousands of LLM-augmented flows per month pay the same as tenants running tens. The pricing is honest about the capability.

Data boundary

The model does not train on tenant data.

The LLM provider is contractually prohibited from training on tenant data. The Strkr Trust page lists the exact provider terms and the SOC 2 controls. For teams with stricter data residency, a dedicated model endpoint is available on the enterprise tier.

Cached results

Repeat prompts hit a cache.

Identical prompts with identical context hit a per-tenant result cache. The cache keeps latency sub-second for repeated patterns (e.g., the same deal re-triggering a summarize action) and reduces the model-call volume. The caching is transparent to the flow author.

Fallback paths

A model error falls back to a non-LLM path.

Every LLM action supports a fallback path for the case where the model is unavailable or returns an unusable output. The flow author sets the fallback (e.g., "if classification fails, assign to the default manual-review queue"). The flow does not stall on a model hiccup.

What this replaces

The technology stack a native automation layer retires.

The sections below are the trade-offs across the common process automation technology stacks and where Strkr lands.

Zapier

Great reach, hourly sync, per-task cost.

Zapier is the broadest integration network and the honest default for teams without a native alternative. The cost model meters per task and the latency sits in the minutes-to-hours range depending on tier. Strkr's native flows run at second-scale latency with no per-task meter inside the Strkr record scope.

Workato / Tray

Powerful, platform tier, engineering intensive.

Workato and Tray are credible mid-market automation platforms. The capability is deep and the price is a dedicated platform SKU. For automations that live within the Strkr record scope, the native flows retire the platform build. For cross-tool automations outside Strkr, the platforms still have a role.

Custom in-house service

Flexible, costs engineering indefinitely.

Rolling your own automation service is viable with an engineering team and a Pub/Sub style event bus. The ongoing cost is the on-call. Strkr's managed flow runtime removes the on-call for the record-centric automations and leaves engineering free to work on the differentiated code paths.

Salesforce Flow

Deep, admin-heavy, Salesforce-scoped.

Salesforce Flow is capable and well-integrated with the Salesforce opportunity model. The configuration demands admin certification and the automation scope stops at the Salesforce boundary. Strkr Flows is simpler to configure and spans the whole Strkr record set (CRM, Marketing, Projects) in one layer.

HubSpot workflows

Credible, metered on contact count.

HubSpot workflows are a credible automation layer inside the HubSpot bundle. The contact-count meter prices up on growth. Strkr's per-seat pricing with included flows is friendlier at scale for teams that have outgrown the free HubSpot tier.

Strkr

Native, typed, LLM-aware, no meter.

Flows runs on the Strkr event bus with typed triggers / actions, LLM augments, webhook bridges, audit + replay, per-tenant quotas, and permission-scoped execution. The capability is on every paid tier. No per-execution meter, no AI add-on, no admin certification.

Process automation technology that ships with your CRM.

Start a free trial, open the Flows editor, and watch the trigger library, action library, and LLM augments appear in the side panel. The whole runtime is included on every paid tier.

Start free trial See Flows automation
Common questions

What buyers ask about this feature.

What is process automation technology?

Process automation technology is the stack underneath a visual flow builder: the event bus that captures record changes, the typed trigger library that subscribes to those events, the typed action library that performs operations, the LLM augment layer that adds AI into a flow step, the webhook ingress / egress for external systems, and the audit / replay layer that captures execution history. Strkr Flows ships all of those pieces native to the CRM, so automations run inside the same workspace as the data they operate on.

What execution guarantees does Strkr Flows provide?

At-least-once trigger delivery (actions should be idempotent, which Strkr makes easy with built-in deduplication keys), ordering of events on the same record, retries with exponential backoff on failed actions, per-tenant quotas to isolate noisy tenants, permission-scoped action execution, and a published latency SLO of under five seconds trigger-to-first-action at the 95th percentile. Every flow run is logged and replayable from the admin dashboard.

How is Strkr AI wired into the flow runtime?

An LLM call is a typed action in the flow builder. The runtime pre-loads the triggering record and its relevant history as context, so the flow author writes the prompt and the system handles the assembly. Outputs are typed (enums, bounded strings) so subsequent flow steps can route on them. Results of identical prompts hit a per-tenant cache for sub-second repeat latency. Every LLM action supports a fallback path for the case where the model is unavailable.

Does Strkr charge per automation execution?

No. Flows is included on every paid tier at the same seat price as the CRM. There is no per-execution meter and no LLM-augment credit meter. Per-tenant quotas exist as a safety fence to isolate tenants from each other, not as a pricing lever. A tenant running thousands of flows per month pays the same seat price as one running tens.

Can Strkr Flows integrate with systems outside Strkr?

Yes. The webhook action POSTs to any external endpoint with a documented JSON payload. The webhook trigger accepts inbound JSON and fires a flow. The pattern covers integrations with industry-specific SaaS, custom internal tools, and systems not natively connected. For high-reach integrations, Strkr's native connector library covers the common tools (Slack, Microsoft 365, Google Workspace, Salesforce, HubSpot, Stripe, and others) with pre-built triggers and actions.

How does the data boundary work for LLM augments?

The LLM provider is contractually prohibited from training on tenant data. The Strkr Trust page lists the provider terms and the SOC 2 controls. For teams with stricter data residency requirements, a dedicated model endpoint is available on the enterprise tier that keeps all LLM inference inside the tenant's preferred region. The audit log captures every LLM call so compliance teams can inspect what was sent and what returned.

Related features

Pair this with...

Process Automation Process Automation Solutions Process Automation Examples AI Process Automation Automation Workflows Business Process Automation

Try it free. Bring your team next week.

No sales call, no migration consultant, no four-month implementation. Enter your card, get 14 days of the full Pro tier, cancel any time before day 14 with zero charge. Spin up a workspace, import your CSV, and have something useful before lunch.

Start free trial See pricing