Skip to content
Actionable Twins

Actionable Twins

Actionable Twins is where you decide. It lists every decision the twins drafted for the selected run, in three lanes: what waits for you, what is on its way, and what came back. You approve or reject each one, follow it up, and see the same decisions where they sit on the map. Open it with Actionable Twins, under Twins in the rail (/today_command).

FACE never acts alone. A decision stays waiting until a named person approves or rejects it, and FACE recommends an option only when it is a clear winner (see Options). The one exception is a standing rule that two different admins approved, and it covers only later cards with exactly the same conditions.

List or Map

The toggle at the top switches between List (/today_command) and Map (/twins/map), the same decisions on the map. On the list, Collapse all draws every card as one row and Expand all opens them all. Open card and Minimize card change one card.

The lanes

LaneHolds
Waiting for youDecisions that still need a person.
In flightDecisions approved, with no final outcome recorded yet.
BackDecisions with an outcome recorded: recovered, partly recovered or denied.

Each lane heading shows how many items it holds and the money their cards state. Cards that state no amount are counted apart ({count} state no amount), never added as zero. An empty lane says why, for example Nothing waits for your decision on this run.

Inside a lane, decisions are grouped by category, with a count such as {count} actions · {waiting} waiting for you. Decisions whose documents state a deadline come first, soonest first. Decisions already decided are marked DECIDED.

Filter the decisions

  • The play chips filter by play, in this order: Claims recovery, Disruptions, Routes, Fulfilment, then Cold chain, Maintenance, Inventory and Fleet. Once the twins have finished, each chip shows its count, for example {play} · {count}.
  • Deadlines in the next 30 days keeps only decisions due in that window.
  • All bands, Medium and up and High only filter fraud and leakage indicators by likelihood.

While a filter is on, a notice titled Filtered to {filter} says so. Show all clears every filter.

The toolbar above the chips holds:

  • Options for this run, which opens Options;
  • Analyze document, which sends a csv, txt, tsv or json file of up to 3 MB to the twins and reads the list again when it is done;
  • Draft next steps automatically, with Recovery leakage or Fraud indicators: an agent drafts steps for a person to approve (Drafted for a person to approve. Nothing has been filed, sent, paid or held.);
  • Refresh. While a run is still working, Updating with the latest results… shows, and the list reads itself again until the run ends.

Read a decision

A card’s row shows its title and the money it states, for example {amount} at risk, {amount} revenue at risk or {amount} drafted order, or not measured: {reason}. Open the card to read:

  • the twin’s reason, its priority and the run that produced it;
  • its items, as on an order: route stops, missing documents, recommended vendors. An amount is printed only when the card states one;
  • Next:, the stage it waits on, for example Next: Approval — waiting for your decision;
  • its pipeline, one step per stage, each done, now, stopped, not yet or not recorded;
  • Open the evidence, which opens the decision artifact;
  • Show on the map, on a decision about a vehicle, a route or an asset;
  • Why, the figures behind it: what the twin stated, the cited records and the measured figures, or No measured figures were recorded for this action.

The pipeline stages

Every stage is drawn from what FACE records. A stage FACE keeps no record of says not recorded.

DecisionStages
Claim or disputeDetected → Evidence gathered → Dispute drafted → Approval → Sent → Response → Resolved
Sourcing or fulfilmentShortfall found → Order drafted → Approval → Placed → Confirmed
Any other decisionApproval, then what the approval carried out

What FACE does not record today:

  • Placed and Confirmed: FACE places no ERP order. An approval records not placed: {reason}.
  • A route approval records the new order. Nothing is sent to a vehicle or a TMS.
  • Sent is done only when the decision log records that the message went out, and stopped with not sent: {reason} when it did not.

Approve a decision

Open the card and read what approving does

WHAT APPROVING DOES says what this approval will and will not do. Read the drafts: the email, the calendar invite, the ERP change.

Select Approve

Your decision is sent at once (Sending your decision…) and the card moves to In flight. If the server cannot be reached, Decision not delivered yet: {reason} · retrying shows, and the app keeps trying. Nothing is recorded until the server confirms.

Read the outcome

A decision that needs two approvers shows 1 of 2 approvals recorded after the first. When the server records an approval but part of it could not run, Recorded, but not executed lists each part with its reason.

When the claims play is selected and more than one claim is ready, Approve all {count} ready claims · {amount} asks first, then approves each claim one after another, through the same approval as a single one, under your name.

FACE has no way to edit a draft before it goes out. To change one, reject it with your reason.

Reject a decision

Every rejection needs a reason: it is the only record of why.

Select Reject

The dialog Why are you rejecting this? opens, naming the decision (Rejecting: {title}).

Write your reason

Type it in Your reason. An empty reason is refused, and so is one too long to keep word for word (Room for about {count} more characters.).

Select Reject with this reason

The decision’s later stages are marked stopped, and the rejection is kept in its decision record. Cancel leaves it where it was.

A fraud or leakage indicator is dismissed the same way: Why are you dismissing this indicator?, then Dismiss with this reason.

Record what happened next

On a claim you approved, Record what happened next records its outcome: Filed, Awaiting reply, Recovered, Partially recovered or Denied. A recovery asks for Amount recovered (USD), recorded as you entered it. Nothing is billed on it. A denial asks why. The claim then moves to Back.

Make this a standing rule

On an approved card, Make this a standing rule proposes that later cards with exactly the same conditions (its kind, category, the actions approving does, the counterparty and an amount ceiling) get the tier you choose: Guarded — one named approver or Autonomous — runs with no approver. Give a reason under Why (shown to the approver) and select Propose.

The proposal changes nothing on its own: Standing rule {id} is waiting for a second admin’s approval. It is not in effect. It takes effect only when a second, different admin approves the exact text on Admin, and any admin can revoke it there.

Follow-ups

A follow-up is a flow that chases the items on FACE’s boards (work orders, claims, deductions) by calling or messaging the people responsible, on rules you approve.

Run a follow-up on selected cards

Select cards on the list, or on the map (Select its card for follow-ups). The selection bar shows {count} selected, with Follow-up and Clear. Follow-up opens a sheet: Pick a follow-up. FACE first tries it on each item and sends nothing; you then choose to run it. After the dry run, Run on {count} items runs it.

The Follow-ups page

Follow-ups (/twins/followups) opens from the sheet. It is not a rail entry: it lights Actionable Twins. Calls and messages that chase your work orders, claims and deductions, on rules you approve.

  • With no follow-up yet, Start with a ready-made follow-up offers starters, for example When a claim’s window closes in 3 days, message the carrier. Use this opens one as a draft. Nothing runs until you arm it.
  • Each follow-up is built from steps: When (what starts it), Reach people (a voice call, WhatsApp or SMS, or telling the addressee), Work orders (mark an order Draft or Sent; approving or rejecting stays a person’s decision) and Timing and checks (wait, or check before going on).
  • A follow-up is Not armed until you arm it, and runs only as a dry run until then (Runs only as a dry run until you arm it.). Once Armed, it Runs by itself when its trigger fires.
  • The assistant on the page drafts changes as pending steps; a person approves them.
  • Tasks lists the open follow-up tasks with their due time.

While no carrier is connected, the page says Calls and messages can’t go out until your carrier is connected., with Connect a carrier, and steps that need one show Needs a carrier.

Map view

The Map view (/twins/map) draws the selected run’s sites, vehicles and routes where your data places them, with their decisions on them. FACE geocodes nothing: a thing is placed only where its source states a position.

The banner says what is on the map from your data: On the map from your data: {sites} places, {vehicles} vehicles, {routes} routes. It also says which run the map shows, and when vehicle positions come from earlier runs or are live positions rather than this run’s.

What the map draws

  • Placed assets: sites and vehicles at their stated position. Vehicles come from a connection tagged telemetry, equipment or shipments; the banner names them, or says No connection is tagged telemetry, equipment or shipments yet.
  • Route proposals: The line joins this card’s stated positions in the proposed order. It is not a road. A route waiting for the route judge’s verdict cannot be approved yet.
  • Reefers: minutes out of band: cold-chain pins, with value at risk only when the card states it.
  • Money at stake, Cold-chain excursions and Detention & dwell: money layers, each with its total and what it could not place.
  • Fleet plan (live): planned stops from the vehicles’ current GPS positions and pending shipments, not from the selected run.
  • Live layers from each active connection that reports positions, and the providers FACE found near sites in trouble (Providers found), each with what the option costs against doing nothing.

Layers lists every layer with a switch, how many features it has on the map ({count} on the map), the attribution its source requires, and how old its data is (Old data · newest {time}). Fit to sites frames every site, route and provider.

Not on the map lists what your data names but the map cannot place, each with its reason: No location in the data, or Address only — waiting for the FACE geocoder.

The basemap is FACE’s own, served by your FACE instance. No third-party map, tile or geocoding service is called.

Select a place

Tap a site, a vehicle or a route, and its card opens beside it. With a keyboard, open On the map and choose an entry. Esc, the close button or a tap on the map closes the card. Show on the map on a list card opens the map on that card.

  • A site card tells its story first, for example how its demand moved in this run, then Outbound demand, Inventory safety nets, Cold chain, Policies, Verified shippers and Suppliers, each with its Basis. No model proposes a supplier or shipper here. Open the entity page opens the site’s own page.
  • A vehicle card shows Predictive maintenance and Cold chain for this run, then its pending decisions with Approve and Reject.
  • A route card shows its mode, its temperature band, the Re-plan (kilometres in the current order against the proposed order) and what triggered it.

A figure the run does not contain says not measured, and why.

Site plans

Site plans opens a side panel over the map with the plans FACE inferred from your BIM/CAD files: each plan’s levels, its elements (zones, docks, yard slots) and the twin entity each element is bound to. Open it from a place’s card, from a plan’s footprint on the map, or from Layers.

  • Plans come from IFC or DXF files. DWG is a proprietary format and is not read: export DXF or IFC from your CAD tool.
  • PDFs and pictures are read by the vision model, and what it finds is labelled proposed by the model until you confirm it.
  • Select an element to see its twin (This element is the twin {entityClass} {id}.), or This element is not in the twin. Only a confirmed zone, dock or yard slot is written. Open in the twin opens the entity.
  • 3D model shows the open plan in the map’s low-poly style: zones and yard slots as coloured floor areas, walls, columns and racks as tall as the file states, and docks and doors as pins. Overview, From above and Low angle reset the camera, Full screen opens the model on its own, and a tap selects the element on the page. An element whose height the file doesn’t state is drawn flat, one with no outline or position isn’t drawn, and the model counts both. A plan in pixels, or in units its file doesn’t state, has no 3D model.

Twin entities

Twin entities is a side panel over the map (/twins/map?panel=entities): every entity FACE discovered from your sources (sites, docks, assets, devices), each with where it came from. Open it from Layers.

All classes filters by class, Search id or values searches, Load more reads the next page and New entity adds one by hand. Select an entity to open its page. An empty twin says so; Could not read the twin means the request failed.

Decision artifacts

A decision artifact is the full record of one decision, with the drafts it would produce. Open the evidence on a card opens it, marked PENDING before approval. The page shows DECISION ARTIFACT #{id}, the title, the decision trail and the GENERATED DRAFTS. A link that carries no decision shows No decision artifact.

Remediation

Remediation: Compliance, Finance and Operations, above the lanes, open the remediation screen for that twin. Each shows the verdict, the server’s statement of which data it rests on, VIOLATIONS ({count}) or FINDINGS ({count}) with each item’s remediation and rule, and the AUDIT TRAIL. Unable to Assess is not a verdict: nothing could be assessed, and the summary says why.

Findings can become work orders. Remediation orders lists them; FACE sends no order anywhere: approving records a decision, and a person sends the order and marks it sent.

On the compliance screen, Routing table holds your own map from finding category to department. FACE holds no org chart and proposes no route. Enter a DEPARTMENT, a CONTACT or both, then Save rule. A rule applies to future orders only.

When the list is empty

You seeMeaning
The twins are working…The run is still working.
The twins finished and proposed nothingThe run’s data led to no decision that needs a person.
Digital twin reasoning…The decisions are loading.
Failed to fetch action queue: {error}The decisions could not be read. This is not an empty run.

RPCs behind this page

  • TwinsService.GetActionQueue, ExecuteAction, ListDecisionLog, ProposeStandingRule, UploadAnalyze (server stream).
  • MapFeatureService.ListMapLayers, StreamMapFeatures (server stream).
  • TwinGraphService.GetTwinSiteDetails, for a site’s card.
  • ComplianceService.AnalyzeCompliance, AssignRemediation, DecideRemediation.
  • runink.ui.siteplan.v1.SitePlanService, for the site plans.

See the API reference.