Logistics · Insurance · Banking · Telecom · Marketing
You find out when it is too late to argue.
An action software takes on its own leaves nobody to ask afterwards.
So this one drafts, and waits.
The approval is the record: a name, a time, and the reason it was sent, kept together.
A reefer drifts warm overnight. A review threshold is raised to clear a backlog and never put back. A rating change is right for the promotion and wrong for a legacy plan. Every one of them was written down somewhere before it cost anything.
Runink FACE is the product behind that: it reads the records your systems already hold, compares each one against the rule that governs it, and puts a drafted action in front of the person who owns the decision.
- The records already existOrders, claims, payment instructions, call detail, contracts and sensor readings — in the systems you already run.
- Every one of them is readNot a monthly sweep over a sample. The comparison runs record by record, so an exception arrives as a named item rather than an estimated rate.
- A person decides what happensEach finding arrives with the rule it broke, the records behind it and a drafted action, for someone to approve, edit or reject.
What changes in the week
Three changes, and they are the ones that move the lines above.
-
You get the finding, not the data
A dashboard shows you a number and leaves the work to you. This arrives as one specific proposed action, ranked, with the rule it invoked and the records it cited attached to it.
-
A named person decides
A finding arrives as a drafted action and waits. Approving it is what sends it, and who approved it, when, and what they changed stays on the record — so the reason can be given later without assembling it again.
-
Your records stay on your machines
The files and the reasoning about them run on hardware you control. Nothing is sent to an outside model provider, which is usually the shortest route through a security review.
Read the FACE paper The long version: what it reads, what it produces, who approves it and where it runs.
Five industries, one shape of problem
Find the line that reads like your week. Each one opens onto a page written for that industry, including the measures to write your own figures against.
-
Logistics & Supply Chain
An entry held at the port for a missing paper while the daily charge runs. A freight claim still inside its filing window that nobody had the morning to assemble.
Operations directorChief financial officerTrade compliance
How it is read today
A container on a hold nobody has joined to the document it is waiting for, while free time runs out. A filing window that closes while the evidence sits in four systems.
How Runink reads it
Every entry is compared against the rule that governs it — held, under examination or detained, with days held above zero — so a hold arrives as a named entry with the charge counted from your own rate.
Where it shows up in your operating numbers
-
Demurrage, detention and the claims that expire
Where yours is
The accessorial lines on the carrier invoice, against the free-time clock on the entry. Claims filed against claims available.
-
Before it ships
Demand and fulfilment
A line starts moving weeks before the reorder point does, and the plan catches up after the stockout.
-
In transit
Cold chain
A container drifts warm overnight and nobody opens the door until morning.
Reactive logistics
The plan that was right at six in the morning is wrong by ten, and nobody re-runs it.
-
After it comes back
Reverse logistics
A return is worth the most on the day it comes back, and the grading call waits.
-
When the money moves
Claims and disputes
A claim expires because assembling one takes a morning nobody had that week.
Underwriting
The wording, the loss report, the reserve history and the authority limit sit in four places.
A container held at the port, from hold to cleared
-
The hold shows up
An entry comes back held, under examination or detained, and the count of days it has been held is above zero. That combination is the whole test — it is a fixed rule, not a judgement, and it runs against every entry rather than the ones somebody thought to check.
-
The cost is counted from your own rate
The days it has been held, multiplied by the per-day demurrage rate in your own agreement. That is the arithmetic in full. It is the figure that is already accruing while the entry sits in a queue nobody reads end to end.
-
The missing paper is named
The reason for the hold and the documents outstanding against it come off the entry record and are stated on the item, so the person who picks it up is not starting by finding out what is wrong.
-
The accountable party is checked
Separately, entries are read for an importer of record that is blank, or filled in with the consignee, or with a placeholder somebody typed once. Those carry duty and tax with nobody accountable for them, and they are raised as their own item with the amount at stake attached.
-
Every pound is counted once
Demurrage on a held entry and duty on an unattributed one are different money, and counting them as one number is the commonest way this sort of total gets inflated. They stay separate, deliberately, and there is a test that fails if they ever merge.
-
A named person decides
The item waits. Approving it is what sends anything, and who approved it, when, and what they changed is written down. If part of what was drafted could not be carried out, the result names that part rather than reporting success.
A reefer that drifted warm, from the door to a filed claim
-
The frame arrives
A photograph taken on a handheld at the door, or a still pulled off a yard camera. Before anything reads it, it is checked to be an image: the header is decoded on its own, the format has to be one of two, and the size is capped both in bytes and in pixels. A PDF, a video container or raw bytes are refused at that step.
-
A model reads it, on your hardware
The frame is reduced to a size the model can take and read by a vision model running on machines you control. What comes back is a written observation tied to the exact frame it was read from, so the sentence and the evidence for it stay together.
-
The paperwork is read beside it
The consignment record, the handover, and the condition the documents say the load should be in. The observation is set against what was already written down rather than against a threshold somebody picked.
-
What differs comes out named
The pallet, the crate, the container door — named, in the words a person would use. A severity score cannot be argued with in front of a carrier. A named part of a named consignment can.
-
The call reaches the person who owns it
A cue can be broadcast to whatever is subscribed to the yard's event stream, and the record keeps it as what it is: requested. What happens next is somebody's decision, and the record says so rather than implying a movement was stopped.
-
A named person decides
The item waits, exactly as the held entry does. Approving it is what sends anything, and who approved it, when, and what they changed is written down.
-
-
Insurance
A review threshold raised to clear a backlog, meant to be temporary, never put back and never decided either way. Reserve movements checked by sample because the stream is too long to read.
Claims operationsCompliance and riskInternal audit
How it is read today
A review threshold raised to clear a backlog, with the policy document still carrying the old figure and nobody deciding either way.
How Runink reads it
The figure that sets a second review is read out of the policy wording itself and compared against what the workflow applied, so a file reviewed on the wrong threshold comes back named rather than sampled.
Where it shows up in your operating numbers
-
Reserves and recoveries signed off on a sample
Where yours is
The share of files that took a second review, against the figure your policy wording states.
-
when authority is delegated
Delegated underwriting
You still answer for what is decided under the agreement, and the reports the agent sends back have to be read against it.
-
while the file is open
Second review
The workflow applies one figure while the policy document still carries another, and which of the two governs is settled the first time somebody is asked.
-
at period end
Reserving and recovery
Treaty terms set what you can recover, and whether a loss was presented under the right ones is a reading of the treaty against the file.
A second-review threshold, from raised in the workflow to decided on the record
-
The policy wording is read
The clause that sets the value at which a claim needs a second review is read out of the policy document. It comes back stated in plain English, with the document it was read from named beside it.
-
The workflow's own rule is read
Separately, the claims system is read for the threshold it actually applies to a file today. It is read where it sits: the connection to your claims data reads the records and leaves them as they are.
-
The difference is named as drift
Every rule lands in one of four states. Aligned: the policy says it and the systems do it. Drift: the systems do something adjacent. Shadow: logic is running that no policy describes. Missing: the policy describes a control nothing performs. A threshold raised in the workflow and left out of the document comes back as drift. Read that as a reconnaissance: the finding names the policy document and the implementation it was read out of, and carries a confidence, so a compliance officer is checking a specific claim against two named sources.
-
The finding reaches one queue
It carries the observation, the rule it invoked, the records it cited, a severity and one specific proposed action. This one turns on risk appetite — whether the raised threshold is the one you now want — so it waits, with the ambiguity named and the write-up already done.
-
A named person decides
Someone approves, edits or rejects the drafted action, and that decision is recorded as an event carrying who made it. Whether the wording is updated to the new figure or the workflow is put back to the old one, the observation, the rule, the records and the approval are kept as the work is done. When an auditor asks what this control did, answering is retrieval.
-
-
Banking & Financial Services
A break that grows inside the range you always clear, so no single month escalates and nobody reads the sequence. A supplier contract nobody has opened since signature.
Compliance and riskInternal auditFinance
How it is read today
A movement that sits inside the range every month, so nothing escalates and nobody reads the sequence the months make.
How Runink reads it
Each break is read against the sequence its own months make rather than the range it sits inside, so a movement that clears every month stops disappearing into the aggregate.
Where it shows up in your operating numbers
-
Breaks that clear inside the range you always clear
Where yours is
The age profile of open items, against how many months each one has been carried.
-
when a payment destination changes
Destination verification
Whether the rule on a payment destination was enforced in the system or followed by a person has a habit of surfacing after the event rather than before it.
-
after a contract is signed
Supplier obligations
The agreement carries obligations on service levels, sub-contracting, data handling and notification, and reading it back against the relationship you actually have is a comparison that gets made when something has already gone wrong.
-
when the fee is charged
Rate card check
Tiers, thresholds and product terms make the fee a reconciliation between what the terms say and what was charged.
A reconciliation break, from a run of clearings to a decision on the record
-
The month clears as usual
The movement sits inside the range that has been cleared before. It is cleared the way the month before it was cleared, and on its own it reads like that month.
-
The sequence is read, not the month
The comparison runs over the series rather than the latest figure. What stands out is the shape of a run of movements, each of which sits inside the range on its own.
-
The item is raised with its series
The item says the sequence is unusual against the history available, and cites the series it was read from. The months, the values and the range they sat inside travel with it, so whoever opens it starts with the evidence in front of them.
-
A named person decides
The item waits in one queue with the series it cites already attached. Approving it, editing it or rejecting it is that person's decision, and the cause is theirs to write: naming one takes somebody who knows what changed in the operation that month.
-
The record answers the next request
Who decided, when, what was cited and what was changed stay together, each entry hashed against the one before it, so a later change to the record shows. When the supervisor or the internal reviewer asks, the answer is read rather than reassembled.
-
-
Telecom
A rating change that is right for the promotion and wrong for a legacy plan, too small to move an aggregate. A sweep that returns an error rate when operations needs the accounts by name.
Revenue assuranceInterconnect settlementFinance and procurement
How it is read today
A rating change right for the promotion and wrong for a legacy plan, too small to move an aggregate and too late to re-rate once it does.
How Runink reads it
Each account's rated output is read against that subscriber's own plan terms, record by record rather than as a monthly sweep over a sample.
Where it shows up in your operating numbers
-
Rating errors that arrive as a rate, not as accounts
Where yours is
The error rate revenue assurance reports upward, and the sample size and month behind it.
-
when usage is rated
Revenue assurance
Rating is where the plan terms meet the usage, and a change made for one product lands on every account the configuration touches.
-
at period end
Interconnect settlement
Your traffic, the counterparty's record of the same traffic and the agreement that sets the rates are three sources, and the one difference that matters shows up at period end as an open item with an age on it.
-
as sites are built
Network rollout
Approved spend, purchase orders, equipment received and sites carrying traffic sit side by side only when somebody assembles that view by hand, and a site that took delivery but was never turned up sits inside it unnoticed.
A rating change wrong for one legacy plan, from the first wrongly rated call to an approved re-rate
-
The rating is read against the plan
Each account's rated output is read against that subscriber's own plan terms, and the comparison is per record rather than a monthly sweep over a sample. What comes back is the accounts by name, which is what a re-rate needs, rather than an error rate, which is what a report needs.
-
The finding carries its evidence
Before anything goes to a model, the finding is checked: complete enough to read, carrying evidence, with at least one piece of that evidence readable, about the account it names rather than something next to it, and dated recently enough to say something about now. Evidence that only writes the claim out a second time ends the matter there. Each check that fires states its own reason in a sentence a person reads.
-
The rate is recomputed from counts
A finding that claims a rate is settled by arithmetic on the raw counts, recomputed rather than accepted. A count divided by nothing comes back as unable to judge.
-
A second credential forms the verdict
The judging runs on a different credential from the submitting, and which credential a message carries is settled by the door it arrived at rather than by a field the sender fills in. A verdict of unable to judge is an answer in its own right: it arrives with its reason written next to it and goes to a person, exactly as a disagreement does.
-
Revenue assurance approves the re-rate
The item waits, with the account named, the plan terms it was read against, and the records it came from. Approving is what sends anything, and who approved it, when, and what they changed stays on the record. The remedy is a configuration correction and a re-rate, made while a re-rate still fixes it.
-
-
Marketing
The audit tool knows the site is slow. The content tool does not, so it keeps writing for a page nobody stays on. Every campaign starts again from a blank page.
The marketing leadSalesIT and information security
How it is read today
The audit tool knows the site is slow; the tool that writes does not, so it keeps writing for a page nobody stays on. This one is Runink PULSE, not FACE.
How Runink reads it
Your site and the channels you publish on are read first, so the next brief starts from where you actually stand instead of from a blank page.
Where it shows up in your operating numbers
-
Work that starts again from a blank page
Where yours is
How long a brief takes to become something published, and how much of last quarter's work the next one reused.
-
before anything is written
Deciding what to say
What to write next depends on what the site is already doing, and those two answers live in different tools.
-
between brief and publication
Getting it out
Brief, draft, review, schedule — each step waits for the one before it, so whether you can answer a competitor's announcement at all is decided by your own turnaround.
-
when a lead arrives
Following up
The customer system knows a lead went quiet in week three; the scheduler does not.
A page nobody stays on, from the audit that finds it to an approved draft
-
The audit reads your site
You bring your website and the channels you publish on. Reading them and telling you where you stand is the first thing Runink PULSE does. What it finds comes back ranked, and it is applied from the same screen.
-
One diagnosis feeds every channel
That same diagnosis feeds the channel analysis, the content plan, the schedule and every draft. What goes out carries one argument rather than four versions of it.
-
The brief becomes a draft
The brief goes in and a draft comes back. From there it sits in a queue where you can see it, instead of going quiet between one step and the next.
-
Each piece carries one status
Draft, waiting for review, approved, rejected, published, archived. Those are the statuses PULSE keeps, so what is waiting on you and what actually went out are both visible.
-
A named person approves it
Every draft — post, whitepaper, cold email, call script — lands in a review queue with an approve and a reject. Approving is a step a named person takes rather than a formality the system performs for them.
-
Runink products you may have heard of
Three products and one joint paper. Each one is a long document that explains the mechanism rather than a brochure — what the software looks at, what it produces, who approves it, and where it runs.
-
Runink FACE
Fulfilment Autonomous Claims Engine
The one this page is about. Held entries, claims still inside their window, cold chain read after the write-off, demand that grew on the way up the chain.
Read the paper · 30 pages -
Runink PULSE
Prescriptive Unified Lead & Social Engine
A separate product, not a FACE feature. The audit, the research, the prospecting and the material a marketing team publishes, on one application the team operates directly.
Read the paper · 23 pages -
Runink CORE
The operations layer underneath both
Not something bought on its own. It is the answer to where your data is processed and who can see it, which is the question every other page here eventually arrives at.
Read the paper · 21 pages -
Runink CORE and Atlas
A joint architecture paper with Logical Leap
Continuous oversight with a second opinion on every finding. Written with the other company rather than about them, so both engineering teams describe the same shape.
Read the paper · 23 pages
Bring one lane, one claim, or one month of invoices.
A short conversation is usually enough to tell whether the losses you carry are the shape this addresses. If they are not, we will say so.
Book a consultation
Half an hour, with whoever owns the problem in the room. We will walk one real example of yours end to end.
Choose a timeOr write to us
Tell us what is costing you, in your own words. We reply within one working day.
Message received
Thank you. We will reply within one working day.
runink.org