Skip to main content

​
Overview

Experiences in Adopt AI are the Human-in-the-Loop (HITL) interfaces where human reviewers inspect, validate, and approve agent work. Pipelines power the data layer of these interfaces — you bind a Pipeline Outcome to a table, chart, or data component in the Experience Builder, and the component automatically displays the most recent data from that outcome. This means reviewers see real, structured business data (not a static mock) every time they open an Experience.

​
Why Bind Pipelines to Experiences?

Without Pipeline bindings, an Experience would either show static data or require the agent to pass all necessary context at runtime — which is inefficient and brittle. With Pipeline bindings:
  • Data is always fresh — the Pipeline runs on its schedule; the Experience always reflects the latest outcome
  • Dashboards load instantly — no live API calls when the reviewer opens the interface
  • Consistent data model — the same Pipeline outcome can drive multiple Experience components
  • Separation of concerns — data preparation (Pipeline) is decoupled from data presentation (Experience)

​
How Pipeline Bindings Work

In the Experience Builder, each data-displaying component (table, chart, metric card) has a Data Source configuration. You select a Pipeline Outcome as the data source, map the outcome’s columns to the component’s fields, and the component renders live data when the Experience is opened. The binding is configured once; the data refreshes automatically each time the Pipeline runs.

​
Setting Up a Pipeline Binding

​
Step 1 — Open the Experience Builder

Navigate to your Experience in the Orchestration section and open it in the builder. Screenshot 2026 04 15 At 11 41 11 PM 2

​
Step 2 — Add a data component

From the component palette, drag a Table, Chart, or Metric component onto the canvas. Screenshot 2026 04 15 At 11 39 05 PM

​
Step 3 — Configure the data source

Click on the component to open its configuration panel. Under Data Source, select Pipeline Outcome. Screenshot 2026 04 15 At 11 39 21 PM 2 A dropdown will appear listing all active Pipelines in the current project. Select the pipeline whose outcome you want to display.

​
Step 4 — Select the Outcome

Once a pipeline is selected, choose the specific Outcome (the named output table). For example, if your pipeline produces deals_over_10k, select that.
Screenshot 2026 04 15 At 11 42 34 PM 2

​
Step 5 — Map columns to component fields

The builder will display the columns available in the selected outcome. Map them to the component’s display fields:
  • For a Table: each outcome column becomes a table column
  • For a Chart: map outcome columns to X-axis, Y-axis, and series fields
  • For a Metric card: select the outcome column to aggregate (sum, count, average)
Screenshot 2026 04 15 At 11 43 04 PM 1

​
Step 6 — Preview and publish

Use the Preview mode to see how the component will render with real Pipeline data. Once satisfied, publish the Experience.

​
Example: Expense Review Dashboard

Use case: A finance team reviews AI-generated expense categorizations before final submission. Pipeline: QuickBooks Expense Aggregation — daily, produces aggregated_expenses with columns: category, total_amount, transaction_count, period. Experience layout: 
┌────────────────────────────────────────────┐
│  Expense Review — March 2026               │
│                                            │
│  [Metric: Total Expenses]                  │
│  Bound to: aggregated_expenses → SUM(total_amount)
│                                            │
│  [Table: Expenses by Category]             │
│  Bound to: aggregated_expenses             │
│  Columns: category, total_amount, count    │
│                                            │
│  [Agent Decision Panel]                    │
│  "The following categories were flagged..."│
│                                            │
│  [ Approve ]  [ Request Changes ]          │
└────────────────────────────────────────────┘
The finance reviewer sees current totals and a breakdown table driven by the Pipeline — no manual data preparation required.

​
Example: Risk Monitoring Dashboard

Use case: A compliance team monitors deals flagged for risk review. Pipeline: Deals Over Threshold — hourly, produces deals_over_10k with columns: deal_name, amount, risk_score, assigned_rep. Experience layout:
  • A Table showing all deals above the threshold, sorted by risk score
  • A Metric card showing total deal value at risk
  • An Approve / Escalate action panel for the reviewer
Because the Pipeline runs hourly, the dashboard always reflects the current state of the pipeline — new deals appear as soon as the next Pipeline run completes.

​
Multiple Components, One Pipeline

A single Pipeline Outcome can power multiple components within the same Experience. For example, aggregated_expenses could drive both a summary table and a bar chart on the same dashboard — both stay in sync because they’re bound to the same outcome.
Design your Pipeline Outcomes with the Experience in mind. If you know reviewers need to see data sorted, filtered, or grouped in a particular way, consider building that transformation into the Pipeline itself rather than relying on the Experience to do it.

​
Pipeline Data Refresh in Experiences

Experiences display Pipeline data as of the most recent successful Pipeline run. They do not trigger Pipeline runs themselves. If a reviewer needs the most current data possible:
  1. Navigate to the Pipeline and click Test Run (or wait for the next scheduled run)
  2. Reload the Experience after the run completes
Future versions of Adopt AI may support on-demand refresh directly from within the Experience. Check the product changelog for updates.

​
Next Steps

  • Best Practices — design patterns for reliable pipeline-experience integrations