What FACE is
FACE by Runink (Fulfilment Autonomous Claims Engine) is for the people who chase logistics losses: claims, cold-chain excursions, customs holds, deductions and invoice disputes. Instead of reading across systems by hand to build each case, they ask FACE a question, get the answer with the money at stake, and decide what happens next. Every figure is cited to the record it came from.
FACE reasons over an organisation’s own operational data and turns it into decisions a person reviews, approves and acts on. It is a Go gRPC backend and a Flutter web app.
The journey: Ask → Briefs → Options → Why → Actionable Twins
flowchart LR
A[Ask<br/>a question, the answer first] --> B[Briefs<br/>the same questions, on a schedule]
A --> C[Options<br/>what could we do]
B --> C
C --> D[Why<br/>the checks and rules behind it]
D --> E[Actionable Twins<br/>approve, reject, follow up]
E -->|approved| F[Your systems and people]
- Ask. You ask a question of your connected systems, by typing, speaking or attaching files. FACE reads your sources through a connection, either directly or through a runner you control, and answers first: what it found and how many decisions wait for you. See Ask.
- Briefs. The questions each persona cares about are asked for you, every week or when your data changes, and the people responsible are told what needs a decision. See Briefs.
- Options. For each decision the run’s checks stand behind, FACE compares the options on one measure and recommends one only when it is a clear winner. See Options.
- Why. The money line (what is owed, at risk and waiting), every check FACE ran on the rows it read, what your records and public pages say about each counterparty, and the business rules behind them. See Why.
- Actionable Twins. The digital twins draft the decisions: claims, disputes, orders, route re-plans. Each waits for a named person to approve or reject it, and follow-ups chase what comes back. See Actionable Twins.
Behind each run, records are grouped into data domains and kept as plain text that FACE searches and reads directly (Direct Corpus Interaction: no embeddings, no vector index), so its reasoning is grounded in your records rather than in a model’s general knowledge. Language and vision models served inside the platform work alongside classical statistics and machine learning. See Analysis & evidence.
Where things are in the app
The left rail is the only menu: three sections, with Settings at its foot.
| Section | Pages |
|---|---|
| Fetch | Ask, where the app opens, and Briefs. Cameras, devices, field readers and live sensors are connection types, opened from their card on Ask. |
| Simulate | Options and Why. |
| Twins | Actionable Twins: the decisions in lanes, with a Map view of the same decisions, site plans and twin entities as panels over the map, and Follow-ups. |
| Foot | Settings: Account, Personas, Integrations (channels, systems, contacts, runners) and Subscription. |
An assistant sits at the bottom right of every page, and a card narrates any run while it works. See Using FACE.
What it is used for
FACE recovers or protects money on a set of plays, claims first:
- Claims recovery. Cross-checking deliveries, weights and counts against shipping documents, watching claim windows, and drafting the claim and its evidence.
- Disruptions and routes. Late or short shipments, and route re-plans over the stated stops.
- Fulfilment and inventory. Stock below its reorder point, and the purchase orders that answer it.
- Cold chain. Temperature readings against stated limits. FACE never assumes a threshold the source did not provide.
- Maintenance and fleet. Predictive maintenance findings per vehicle.
- Invoices, deductions and contracts. Overbilling, duplicate invoices and deductions the documents do not support.
- Custody, customs and cover. Bills of lading, customs entries and insurance, checked before a file is bound or a claim is paid.
Principles you can rely on
- Sovereign AI. Every model call goes to the one inference plane your deployment
names (
INFERENCE_REMOTE_URL), run by the platform. FACE links no AI vendor’s SDK and sends your data to no outside AI service. - Your data stays yours. Connection credentials are sealed and never returned by the API. Data at rest is encrypted, and FACE refuses to start without its encryption key. See Security & trust.
- Honest outputs. “Not measured” and “could not be read” are different things, and neither is ever shown as zero. A document is produced only when the evidence supports it.
- FACE never acts alone. Every decision waits until a named person approves or rejects it, unless a standing rule that two different admins approved covers later decisions with exactly the same conditions. FACE recommends an option only when it is a clear winner. A decision is stored with the authenticated identity of the person who made it.