This is the full surface of ClearGrid Command, the same system our own operations run on. No roadmap slides; everything below is live product.
Command opens on the portfolio, not on an empty search box. Every number here is a filter, so the question you just asked becomes the queue you work next.
Customers, valid debt, recovered and committed percentages and total payments — switchable between one lender and the whole book, with the change on each figure shown against the last period.
Teams compose the operational view they actually need without anyone rebuilding the system for them.
Promises to pay and settled deals plotted month by month across six months, the year to date or a full twelve, so you can see whether this month is a blip or a direction.
Allocated, attempted, right-party contact, promise to pay and broken promise, shown as one funnel with the conversion at each step. You can see exactly where the book is leaking.
Raised, awaiting bank approval, and approved-but-dishonoured — every state visible until the money actually clears.
Allocation and recovery broken out across nought to thirty days, thirty-one to sixty, sixty-one to ninety and ninety-one plus — with the amount and the customer count in each. This is the view that tells you whether you are working the winnable end of the book.
A monthly recovery goal, what has actually been recovered against it, and the percentage achieved — not assembled in a spreadsheet at month end.
What share of connected calls turned into a payment within twenty-four hours, forty-eight hours and seven days. The gap between those three numbers tells you whether your follow-up is working.
Promises made, payments actually received after them, and the conversion between the two — alongside a separate broken-promise count and rate that flags itself when it needs attention.
Recovery, engagement and days-past-due movement, by lender, product and period.
Control, evidence and trust — the numbers your committee asks for, and the trail your auditor asks for, from the same source.
Promises secured, right-party rate, deals settled and paid in full, and how many flagged calls are still waiting on review. Each one with its own definition on screen and its movement against the last period.
How many agents actually logged work as a share of the roster, calls made across every call type, total active time as a share of time tracked, and messages sent. Effort and result, side by side.
Every outcome split into reached and not reached, then broken down further — promises, settled, pending, contact, negative, no contact. Promises and settled deals are counted separately so the two milestones never blur into one number.
Re-run the aggregation or rebuild today’s figures when you need them current, then export the whole thing.
By lender, product and period, generated on demand with the columns you choose.
Delivery, engagement and response across every messaging channel.
Volumes, outcomes and dispositions, across both AI and human calls, on one set of definitions.
Every decision, interaction and state change, so any account can be reconstructed end to end rather than assembled from four systems.
Compliance and operations can download the proof, not just look at it on a screen.
Everything known about an account in one place — balances, contactability, history, propensity — kept current by the work itself rather than by a nightly batch.
Every call, message, promise, payment and change of status lands on one ordered history. Anyone who opens the account reads the same story in the same order.
The trail records exactly where a borrower’s conversation was left off. Any channel or any agent picks it up from that point instead of starting the whole conversation again.
Command estimates how likely a borrower is to pick up, how likely the person on the line is the right-party contact — the borrower themselves rather than a relative or a wrong number — and how likely they are to make a promise to pay. It also points to the window in the day when reaching them tends to work best.
Groups are defined by days past due, collections stage and behaviour, and then they keep themselves current as those facts change. Journeys and campaigns draw the people they contact straight from those groups.
How reachable someone is, which channel they answer on and whether they keep their promises are rolled up for each borrower. One look tells you how to approach this person and how much weight to give their word.
The same person can arrive twice under different lenders or in different files. Command flags the likely matches so they merge toward one borrower state rather than several competing versions.
Not just a number out of a hundred but a band and a plain-language call — whether a promise is likely this week. The agent gets a judgement, not a statistic to interpret.
Their own number, a spouse, an employer, an old friend — each labelled, each showing which channels it supports, whether it has been verified, and the last thing that happened when someone tried it.
Total exposure across every account they hold with you, what is still owed as a share of it, what is genuinely past grace, and when the lender’s own system last synced.
The full interaction history, filterable down to just the conversations, just the tickets, or just the journeys that touched this account.
When a borrower has gone quiet, tracing looks for fresh ways to reach them. Any new number or address feeds straight back into the account state for the next attempt.
Inbound is where recovery is won cheaply and lost quietly. Every inbound route lands attached to its account, in a queue somebody owns.
Email and ClearBot conversations land in one queue instead of separate tools. Each thread is attached to the account it came from, so whoever opens it can already see the full history.
Every inbound email becomes a ticket with a clear state: needs a reply, open, assigned, in progress, awaiting the borrower or on hold. One named person owns the deal at any moment, so nothing sits unanswered because everyone assumed a colleague had it.
When someone challenges a balance or a charge, the case opens as a dispute and moves through new, in progress, pending, resolved, assigned and closed. Each lender gets its own view with labels and days past due — how long the account has been unpaid — so the oldest cases are easy to spot.
While an account is in dispute, outbound contact stops automatically. Nobody has to remember to pause the calls and messages by hand.
Each published phone number — its DID, or direct inward dialling number — is pointed at the right dialler, agent group or AI overflow for the moments when every agent is busy. Live status and alerts show whether each route is working, so a silent number does not go unnoticed.
Manual calls, campaign calls, inbound calls and AI calls sit in the same log on the same definitions — which is the only way a comparison between them means anything.
Hand over a list of accounts and the platform works through the queue, showing you what is waiting, what is being processed right now and what is finished. You can watch it live and let AI campaigns run on their own once you trust them.
Every campaign with its contact count, how many are queued, how many completed, how many succeeded and how many failed — moving through queued, processing, processed and completed, and pausable or stoppable mid-flight.
Every call sits in one place, whether it was dialled by a person, placed by a campaign, answered inbound, handled by AI or passed from AI to a colleague. Filter by status, outcome, the disposition the agent chose at the end, how long the call ran, who hung up and whether it was abandoned before anyone answered.
Open the recording from the call row and play it back with a waveform, jump five seconds either way, change the speed, bookmark a moment or download the file. Your compliance team does not need a separate system to listen to a call.
By agent, campaign, lender and date range, then by call type, how the call ended, its status, its outcome, its disposition, whether it was abandoned and how long it lasted. Manual and AI calls sit in the same table.
Inbound numbers that do not match an account land in their own queue rather than disappearing.
A DID is direct inward dialling: one of the phone numbers your customers see and ring back. Each number carries its own rules, so you choose whether a dialler, a particular agent group or AI overflow picks up on it.
Pickup, connection and conversion are measured the same way for AI as for people. That is what lets you compare the two honestly instead of arguing about it.
The journey is the strategy: who enters, what happens, in what order, under which rules — built on a canvas instead of buried in a stored procedure. The composer that writes the messages lives here too.
Type the strategy you want in plain language and Journey GPT drafts the canvas for you. A person then reviews and edits every block before it goes anywhere near a borrower.
Every canvas is assembled from eight kinds of block: entry and triggers, channels and actions, conditions, flow controls, data and state, integrations, AI and optimisation, and exit and governance. Between them they cover what puts a borrower on the path, what happens along it and how they leave.
One place holds the rules every journey must obey: actions, triggers, segments, escalations, contact windows, frequency caps, do-not-contact (DNC) lists and offer eligibility. Change a rule once and every journey that relies on it follows the new version.
A journey stays a draft until someone deliberately publishes it, with locking, versioning and change notes recorded as it evolves. Nothing reaches a borrower because someone clicked the wrong thing.
Before a journey goes live it passes static checks, validation of every AI call and a test run on sample cases. You see exactly what the journey would do while no real borrower is contacted.
Folders and master journeys keep each lender’s treatment organised and route borrowers into the right stage: right-party contact (RPC), promise to pay (PTP), broken promise, hardship and skip tracing. Skip tracing is the stage for borrowers whose phone numbers and addresses no longer work.
Each journey carries its own report, so outcomes and performance stay attributable to the journey that governed them. When two approaches run alongside each other, you can tell which one did the work.
Follow-up messages drafted by AI with the whole account in view, then checked against the lender’s rules before anyone can send them.
Tone, purpose, language and wording rules set per lender, so a message on behalf of one bank never sounds like a message on behalf of another.
Approved SMS and email variants with controlled design and editable slots. The agent fills the gap; they do not rewrite the compliance line.
AI agents that call, verify, negotiate and log every word — in Arabic and English, inside the rules you set.
You decide how each agent behaves: the strategy it follows, the identity checks it runs before it will discuss an account and the disclosures your own policy requires it to read out. You can also see which agents are still in training and which are ready for live calls.
Agents speak both languages, so the customer hears the one they are comfortable in. This is stated plainly rather than left as an assumption.
When your people are on other calls or off shift, AI answers instead of leaving the customer listening to a ringing phone. You get an alert and can take the call back in one click, so control never leaves your side.
Your collectors do not open a spreadsheet and decide where to start. They open one screen that already knows what is overdue — and every hour they spend on it comes back as a number you can coach from.
Each agent’s day as four counts they cannot argue with: overdue, due today, coming up in the next seven days, and finished in the last seven. Every row carries the borrower, the lender, the amount, the due date and how many days late it already is.
One button that picks up exactly where the agent stopped. Nobody spends the first ten minutes of a shift working out what to do first.
One task at a time, on a single screen, with everything needed to make the call already on it. No tabs, no hunting, far fewer mistakes.
Outstanding balance, how many promises this borrower has made and how many they actually kept, and when the last one broke. The agent knows who they are calling before the line connects.
A primary line for what this call is actually about, and a fallback for when the borrower cannot pay in full — including what the agent is allowed to offer without asking anyone. A restructure inside the approved limit needs no approval call.
Select everything overdue and due today in one action, or pull every outstanding promise to pay, and work the set as a block.
By agent, lender, channel, task type and status — or sort into an order the agent has chosen and keeps.
Promises secured, right-party contact rate, calls made and time actually spent on the tools. Coaching starts from the same numbers everyone can see rather than from an opinion.
Hardship specialists, single-lender pods and language splits, fed automatically by the journeys and escalations rather than by someone reassigning rows by hand.
Rank the floor on promises secured, on calls made, or on time spent on the tools. High activity and high conversion are not the same skill and the same leaderboard cannot show you both.
Active time against idle time, promises secured, borrowers touched, repeat contacts, calls split between outbound and campaign, and which page each agent spent their day on. Sortable on every column.
Who is working right now, borrower-level insight rollups, and the team structures that journeys and escalations route into.
Most-viewed sections ranked, and where active time actually landed across the operation. If your team is living in skip trace rather than in conversations, this is where you see it.
A borrower who promised to pay and then did not is not a dead account. They engaged and they intended to pay, and something got in the way. Most collections operations never find out what — because nobody notices the promise broke until month end.
An amount and a date, held as a state on the account. Not free text in a call log that nobody ever queries.
The promised date passes without the payment landing and the account changes state by itself. No report to run, no month end, nobody having to remember.
Every borrower carries their whole promise history — how many they have made, how many they kept, the rate, and how many broke in the last thirty days. That is the difference between a first slip and a pattern, and it changes the conversation completely.
The upcoming promise, its amount and its date sit on the profile with an overdue marker the moment it passes. A repeat breaker is visible to whoever picks up the account next.
The break creates a rework task, assigned, dated and prioritised in an agent’s queue, with the days it has been sitting there shown in red.
How many times we have tried, how many were human calls and how many were AI, how many connected and how many were answered. Nobody re-dials a number that has failed four times.
If they cannot pay in full, the fallback is waiting — a restructure inside a limit the lender set. No hold music while somebody asks a manager.
Where a call is not the right answer, the broken promise hands the account to its own journey and re-engagement continues on whichever channel that borrower actually responds to.
If the borrower says they already paid, the dispute holds the amount and stops outbound contact rather than chasing someone who may be right.
Broken promises sit in the funnel next to promises made and payments received, with a broken count and a broken rate that flag themselves when they need attention. You can see whether your problem is getting commitments or keeping them.
Rules that live in a policy document get broken. Rules that live in the engine cannot be. Command enforces conduct at the moment of contact, and keeps the proof that it did.
Calling hours, message limits per day and per week, and rest periods between attempts. Set once per lender and per market, enforced by the platform on voice, SMS and email alike — not left to an agent’s judgement at 9pm.
When a borrower asks to stop being contacted, that is a do-not-contact instruction and it has to hold. Command records who applied the suppression, when it happened and why — then shows the evidence that every blocked message and call really was stopped.
Identity, the reason for the call and the borrower’s rights, stated on every contact in the language the borrower chose. The AI cannot skip them and a person cannot forget them.
What each borrower has agreed to be contacted on is held on the account and respected by every journey, so permission is a property of the borrower rather than a setting on a campaign.
Command watches spam thresholds, suppression rules and the health of each channel while outreach is still running. Trouble is raised mid-flight instead of being discovered once the damage is already done.
Every rule set carries a version. When a call is reviewed months later you can tell which policy was in force when it happened, rather than judging old work by today’s rules.
Suppressions applied, contacts blocked, disclosures made, windows respected — exportable for a regulator or an internal audit without anyone assembling it by hand.
Your licence and your brand ride on every conversation, whether a person made the call or the AI did. Unified QA scores both against one model, names exactly which check failed, and puts a human in front of every evaluation before it is filed.
AI calls and human calls are scored the same way, so “is the AI as good as my team?” becomes a question your own dashboard answers rather than an argument.
Every evaluated call lands as role model, needs improvement, below standard or fail. You can see the shape of your quality, not just an average that hides both ends.
Tracked for the period you choose and month to date, with the definition shown on screen so nobody has to ask how it was calculated.
Not a score with no explanation. The checks that failed most often in the window are listed and ranked, each with how many calls it affected and which category it belongs to — AI logic, call quality or how the call was closed.
Hallucination and getting a borrower’s gender wrong are tracked as named, counted failure modes, alongside pushing too hard for a payment date and referring someone to the wrong place. If our AI does it, it appears in your report.
A missed disclosure is not the same kind of problem as a clumsy closing line. Compliance failures are counted on their own so they cannot be averaged away.
Response time measured on every call and reported at the median, the 90th and the 95th percentile. A borrower who hears a pause assumes the line has dropped.
Jabi reads the call and drafts the full scorecard in seconds, every check carrying a confidence level and a quote from the call as its evidence. Nothing is filed until someone approves it.
Ask Jabi who is not doing well this week and you get a ranked list, the reason behind each name, and a link straight to the calls it read.
Any mention of a lawyer, the police or a complaint is pulled for human review on every call, AI or human. The escalation reaches you with a written reason, and we track that you acknowledged it.
The review backlog is bucketed by how long each item has been waiting, so a call from last week cannot sit unevaluated behind today’s.
Your rules are written in plain language and wired into how calls are judged, and every version is fingerprinted — so we can always show you which rules judged which call.
Agents ranked worst-first with their weakest area already named, so the conversation starts from what to fix.
Each portfolio runs in its own environment with its own identity, rules and files. Nothing crosses between them, and every lender can be shown exactly that.
Every lender gets its own sealed environment with its own web address, its own legal identity, its own settings and its own files. The register is the single place where all of that is set up and kept straight.
Policies, contact journeys, writing styles and workflows are set for one portfolio at a time rather than once for everybody. What one lender has agreed to never spills over into how another lender’s borrowers are treated.
Approvals, incoming file processing and daily reconciliation run on their own. Every step is logged, so anyone can see what happened and who signed it off.
Portfolio data and everyday operational documents stay inside the lender environment they belong to. Nothing drifts into a neighbouring portfolio by accident.
An agency can route accounts to the teams performing best, compare those teams on definitions that mean the same thing everywhere, and keep the borrower’s full conversation trail intact as work moves around the panel. Conduct rules are enforced by the platform rather than left to a policy document nobody reads.