The AI Operating Model

Seven layers. Everyone buys the same bottom. Nobody can buy the top.

Software gets installed. An operating model gets run. We deploy sovereign compute, an open base model, your industry's regulation and your company's records, then put an agent workforce inside your workflow — and instrument every decision it makes and every outcome that follows. The bottom four layers are procurement. The top three are why the operation gets cheaper and better every quarter it runs.

Rented

Compute and base model. A supplier relationship. Swapped when something better ships.

Configured

Industry corpus and your company data. Necessary, and available to anyone with the same budget.

Compounds

What was decided, what happened, and what that teaches the next deployment.

The layer stack

Seven layers, three bands.

Competitors sell into the bottom two. We deliver all seven, and the top three cannot be delivered by anyone who does not run the operation.

Deployed surfacesagents inside the workflow
Judgement and transferproven once, reused
Action and outcome datawhat we did, what happened
Company datapolicy, product, systems of record
Industry corpusregulation, taxonomy, conduct rules
Open base modelcommodity, swappable
Sovereign computein-country, on-premise
Compounds

What was decided, what was done, what happened, and what that teaches the next deployment. This layer cannot be procured. It only accumulates, and only if someone is instrumenting for it from day one.

The bottom four layers are a purchase order. The top three are a history.

Data as the backbone

The deposit under the operation.

Every operation sits on a deposit — interactions, transcripts, dispositions, exceptions, escalations, overrides, resolutions. In most enterprises it is retained for compliance, stored at cost, and read by nobody. It is ore, and ore is not gold.

Deposit
Your operation
Refinery
Inside your perimeter
Decision
What leaves

The deposit is yours. It stays yours. We build the refinery on your side of the wall, and what leaves is a decision — never a record.

01

It is action-conditioned

Not only what the customer said, but what was done in response and what happened next. State, action, outcome. Most enterprise data records the state and skips the other two.

02

Outcomes resolve fast

In collections, servicing, renewals, claims and disputes there are a bounded number of things you can do, and you learn within days or weeks whether it worked. That combination is what makes learning possible at all.

03

The counterfactual is built in

Running an operation at scale means the same situation gets handled different ways across thousands of instances. That is an experiment which has been running for eighteen years without anyone calling it one.

04

It cannot be acquired

Compute is available to anyone with capital. Base models are available to anyone with a download. A record of what an operation decided and what followed exists only where someone was running it and instrumenting at the time.

Action-conditioned data

Frontier labs read about the enterprise. We stand inside it.

A decision record has three parts: the state something was in, the action that was chosen, and the outcome that followed. The public internet is full of the first part, described after the fact by someone who was not in the room. The other two are almost never written down — and a model cannot learn a policy from outcomes it has never seen attached to the actions that caused them.

Action-conditioned data: what a frontier lab can reach versus what running an operation produces Two rows compare the same three-part record. The upper row, available to a frontier lab, contains only a second-hand account of the state, with the action chosen and the outcome that followed both absent. The lower row, produced by running the operation inside the client perimeter, contains all three parts, and a return arrow shows the loop closing so that each outcome raises the starting point for the next decision. AVAILABLE TO A FRONTIER LAB Public text, documents, code, published transcripts. Situations described from outside the room, after the decision was made. STATE partial, second-hand ACTION not recorded OUTCOME not attached PRODUCED BY RUNNING THE OPERATION Captured on-premise, inside the client perimeter, where no external vendor reaches — because we are the one running the function. STATE account, policy, moment ACTION what was chosen, and why OUTCOME what actually happened THE LOOP CLOSES — EACH OUTCOME RAISES THE PRIOR ON THE NEXT DECISION

The same situation gets handled different ways across thousands of instances, so the counterfactual is built in. Pricing an outcome requires knowing the counterfactual — which is an actuarial asset, not a software asset.

Why the gap does not close with scale

A larger model reads more of what was written down. It does not create records of decisions nobody recorded. Scaling the top row does not produce the bottom row.

Why it cannot be bought

Regulated data does not leave the building — which is precisely why an architecture that requires the data to travel is structurally excluded from producing this record.

Why eighteen years matters

The record accumulates at the speed the operation runs. There is no way to buy the back years, and the acquisition of an operator gets you the seats rather than the history.

Capital can buy compute. It cannot buy eighteen years of outcomes.

Sequencing

Instrument before you automate.

A learning loop needs a stable thing to attach an outcome to. If the same customer is three different records across your CRM, your billing system and your collections platform, the outcome attaches to nothing and the loop never closes. Which is why phase one of every deployment is instrumentation — and why we do it whether or not you proceed.

A record with no loop is a data warehouse

Beautifully modelled, queryable, and it never gets better. Fifteen years of enterprise data programmes ended here.

A loop with no record is A/B testing

You learn that B beat A, but not for which customer, under which policy, in which state. So the finding does not generalise, and it does not survive a regulator's question.

Remember. Learn. Reuse.

Remember

Every account, decision, policy and outcome lives as one connected record instead of five disconnected systems. One customer, one history.

Learn

Because outcomes attach to specific decisions in specific states, the system can tell which approach worked for whom, and under what conditions. Not on average — specifically.

Reuse

A pattern proven in one place raises the starting point in the next. The second deployment does not begin at zero.

100Deploy 1
[XX]Deploy 2
[XX]Deploy 3
[XX]Deploy 4

Cost per resolved task, indexed. [Illustrative — replace with measured figures before publishing.]

Transfer

Structure carries. Records don't.

A pattern learned in one deployment makes the next one cheaper. That is the economic argument, and it is also the sentence that makes a data protection officer sit forward. So here is the mechanism, precisely.

What carries is shape

That accounts have arrears states. That promises have cure rates. That certain sequences of contact perform differently under certain conditions. The industry corpus is the shared schema.

What never carries is instance

Your customers, your balances, your records, your anything. Company data is the instances, and instances stay behind the wall.

Rung 1 · Live

Department to department

Inside one firm, where a single controller decides. Your collections function and your servicing function share a learning curve.

Rung 2 · In build

Firm to firm, federated

Model updates move, raw records do not, and each firm consents explicitly to participate.

Rung 3 · Roadmap

Market-wide

Only where participants have agreed to pool at signup. A ladder we are climbing, not a moat we already hold.

This is also why the moat is vertical and the platform is horizontal. A collections pattern learned at one bank transfers to another bank because both have accounts, arrears, promises and cures with the same shape. It does not transfer to an airline.

Governance

Certify, intercept, correct.

An autonomous agent in a regulated process needs three things before a compliance officer signs: proof it was certified against policy before deployment, the ability to intercept a decision in flight, and a record of every correction made after. We run that as a control plane over every deployed agent.

The control plane is also the capture plane. Every interception is a labelled example of the boundary. Every correction is a domain expert saying exactly where the system was wrong, and why.

Compliance is normally a tax on AI deployment. Instrumented correctly, it is the input — the governance layer regulated clients require is the same layer that produces the highest-quality training signal in the stack.

Commoditisation

Frontier models get cheaper. The advantage gets deeper.

The most common question from technical buyers: what happens to your position when open models are good enough that anyone can do this? We are the buyer of that trend, not the seller.

Better models make the operation cheaper. The base model is a line item that falls every quarter and swaps out when something better ships. Our unit of measurement is cost per resolved task, not benchmark rank.

Cheaper operations win more contracts. When capability improves, margin on a contracted outcome rises. When multilingual coverage improves, [120+] languages get cheaper to serve.

More contracts deepen the layer nobody can buy. None of it substitutes for a record of what an operation decided and what followed.

Platform proof

Two layers change. Five stay constant.

That is what makes this a platform rather than five bespoke projects — and why a deployment in your sector starts from a running position rather than a blank page.

SectorIndustry corpusAction and outcome data
BankingPrudential and AML codes, dispute rulesCollections outcomes, recovery rates
TelecomConsumer codes, tariff structuresChurn saves, credit decisions
HealthcareICD-10, HL7/FHIR, care guidelinesPrior-auth results, pathway outcomes
FintechE-money licensing, KYC tiersFraud calls and their outcomes
TravelFare rules, disruption policyRebooking and waiver outcomes
Deployment

Sovereign by construction.

Sovereignty here means in-perimeter, not government-only. Every regulated enterprise needs it; not every regulated enterprise is a ministry.

On-premise or in-region

Your choice, written into the contract rather than set as a configuration option. The record is captured and stays inside your perimeter.

Model-agnostic

No dependency on a single frontier lab, and none on a lab being permitted to operate in your jurisdiction.

Auditable by construction

Every decision reconstructable with the policy version in force at the time, the state the account was in, and the reasoning trace.

Compute funded as a real asset

Where an in-country facility is needed it is capitalised deal by deal by a real-asset partner, with an infrastructure operator running it. We anchor as the operating tenant. It sits on neither your balance sheet nor ours.

One number we contract against

Cost per resolved task, published monthly against your baseline. Not benchmark scores, not model size, not tokens, not seats.

You keep the record

The instrumented record is built inside your perimeter and it is yours. If we part, it stays.

Additive, not a re-platforming programme

We deploy alongside your existing core, CRM and telephony rather than replacing them. Scope can start narrow, prove value, and expand function by function.

Compliance is architectural

Audit trail, data residency and regulatory guardrails are built into the stack rather than bolted on, so the evidence your examiners ask for is an output of the system.

Full-population monitoring, not a sample

Fraud, mis-selling, evasion and agent absence monitored across every interaction, with inference running at the edge inside your environment.

Where this fits

Where we win, and where we don't.

On-premise deployment used to be the moat. It is becoming table stakes, and we are glad — because it moves the contest to the thing nobody can buy.

CategoryWhere they are strongWhere the line falls
HyperscalersBundled, already inside the bank, unmatched infrastructureWe compete on depth of operation, not infrastructure — and increasingly we run on top of them
Well-funded AI-nativesGreenfield, digital-first accounts. Better capitalised than usThey do not win eighteen-year incumbency inside a Tier 1 bank's core operations
Regulated-vertical specialistsClosest to us on deployment postureNarrower on scope, and software vendors rather than operators
Global CX incumbentsInstalled telephony and workforce platformsStructurally slow, because their revenue depends on the seat model agents replace
Traditional BPOsScale, relationships, delivery footprintBuyers of AI rather than builders. Our channel, not our rivals

What we lose, said plainly: greenfield digital-native accounts, and accounts already under a global vendor or hyperscaler mandate. We win where operations are complex, regulated, multilingual and already running.

Moat 01

Installed position

[176] regulated enterprises, [7] countries, 120+ languages, [62] local channel partners. Years to replicate, not dollars.

Moat 02

Switching cost

The software runs the operation. Leaving means abandoning institutional memory and re-certifying from zero.

Moat 03

The decision record

Eighteen years of situation, action chosen and outcome, captured because we operate. The only one that cannot be bought at any price.

Sequence

How it lands.

Phase 1 · [Weeks 0–4]

Instrument

Unify the record — one account, one history. Baseline every metric before anything changes. You keep this whether or not you proceed.

Phase 2 · [Weeks 4–12]

Automate

Agents deployed on the highest-volume, lowest-judgment queue, running in parallel with your existing operation until the SLA threshold is met.

Phase 3 · Ongoing

Compound

Loops close, patterns accumulate, and the next function starts from a running position rather than zero.

Phase 1 is the one most vendors skip, because it does not demo well. It is the only phase that determines whether Phase 3 ever happens.

Technical diligence

What technical buyers ask.

Isn't the base model doing all the work?

The base model does the reasoning. It has no memory of your operation, no record of what worked, and it resets every session. Layers five through seven are the difference between a capable model and an operation that improves.

Why can't our incumbent vendor do this?

They can buy the same compute and download the same model. What they do not have is a record of decisions and outcomes, because they were not instrumenting for it. That takes years to accumulate, not quarters.

Couldn't a competitor just acquire a BPO?

They can buy the seats. The instrumented history is not on the balance sheet, and the acquisition itself takes years.

What happens to our data if we leave?

You keep the instrumented record. It was built inside your perimeter and it is yours. We keep learned structure, which contains none of your instances.

How do we explain an AI decision to our regulator?

Every decision is reconstructable with the policy version in force at the time, the state the account was in, and the reasoning trace behind it. That is the reason the control plane exists.

What if the model you use gets deprecated?

It will. Base models are a supplier relationship and we swap them. Layers three through seven survive the swap intact.

Get started

Start with the instrumentation.

Phase 1 stands alone. We baseline your operation, unify the record, and show you what your current process actually costs per resolved task. You keep the analysis and the instrumentation whether or not you proceed.

AI Center of Excellence

Three contracts, three vendors, three cost centres. One Center of Excellence.

A Center of Excellence is a single operating unit that runs front, middle and back office work under one SLA, inside your perimeter. It replaces the outsourcing contract, the managed-services contract and the internal shared service centre with one thing — and unlike all three, it gets cheaper as it learns.

Human-led

People set the policy, hold the mandate, and decide every exception the agents escalate.

Operated by AI

Agents do the work itself — the volume, the routine, the overnight, the multilingual.

One SLA

One counterparty accountable for the outcome across all three offices. Not three vendors pointing at each other.

The consolidation

What it replaces.

Most enterprises run all three of these at once, with different vendors, different contracts and no shared memory between them. Every boundary between them is a place where work waits, context is lost, and accountability disappears.

TodayWhat you buyWhat breaksIn a Center of Excellence
Business process outsourcingLabour hours in a delivery marketAttrition, wage inflation, quality that varies by site and by shiftAgents that do not turn over, in 120+ languages, with no night-shift surcharge
IT outsourcing and managed servicesDays of effort against a ticket queueEffort billed whether or not it resolved anythingResolution priced per outcome, with the service desk and the operation sharing one record
Shared services / GBSAn internal cost centre defending its budgetAutomation targets arrive without a platform to deliver themOne controller, so learning crosses finance, HR, procurement and servicing without a consent gate

The consolidation is the point. Three separate vendors cannot share a decision record with each other, so nothing any of them learns compounds. One unit can.

Scope

Front office to back office.

A CoE starts where volume is highest and judgment is lowest, then moves inward. The meters show how much of each band runs autonomously today, across live deployments. [Confirm against current deployment data before publishing.]

Front office

[70%] autonomous

Everything that touches the customer. High volume, fast resolution, and the first place a CoE proves itself — because the baseline is measured and the delta shows within weeks.

Inbound servicingCollectionsOutbound campaignsComplaints intakeRetention and savesAppointment and booking flowsDisruption handlingOnboarding

Middle office

[XX%] autonomous

The judgment-heavy work between the customer and the ledger. Regulated, evidence-hungry, and where most automation programmes stop because the audit requirement was never designed in.

eKYC and onboarding checksAML and sanctions screeningFraud triageClaims and FNOLDisputes and chargebacksCredit decisioning supportCompliance surveillanceException handling

Back office

[XX%] autonomous

The internal functions a shared service centre already runs. One controller, so a pattern proven in collections carries into servicing and then into procurement without a new consent conversation.

Accounts payable and receivableReconciliationOrder and PO processingHR administrationPayroll queriesProcurement supportIT service deskReporting and close support

Scope expands without a new procurement each time. That is the difference between a CoE and a stack of point tools bought department by department, none of which compound across the boundaries between them.

Composition

What's inside one.

Four components. A CoE that is missing any of them is a pilot, and pilots do not compound.

Component 01 · Studio

Where the operation is designed

Workflows mapped, agents configured, policies encoded, changes shipped and QA'd. Your operations leads run it after [X] weeks of training — the CoE is not a black box you rent access to.

  • Agent configuration and versioning
  • Policy encoding with an audit trail per change
  • Champion/challenger testing before anything goes live
  • Live dashboards against contracted SLA metrics
Component 02 · Agentic AI

The workforce that does the work

Voice, chat, document and orchestration agents across all three offices. 100+ roles from a pre-built library, live in [2–4] weeks rather than built from scratch per client.

  • Conversational and generative agents, 120+ languages
  • Back-office orchestration across systems of record
  • Control plane — certify before deployment, intercept in flight, correct after
  • Escalation to named humans on defined conditions
Component 03 · Vertical models

Models that know the domain

Tuned on your corpus and your industry's regulation, evaluated against your audited outcomes rather than public benchmarks. Deployable on-premise, and swappable underneath.

  • Domain and dialect tuning for the markets you serve
  • Evaluation suite built from your own outcome history
  • Base model treated as a supplier and replaced when a better one ships
Component 04 · Compute

Sovereign, in-region

On-premise or in-country. Data does not cross a border, and the record accrues where no external vendor reaches. Funded as a real asset rather than as capex on your balance sheet.

  • In-perimeter deployment, documented in contract
  • Edge inference for full-population compliance monitoring
  • Capitalised deal by deal, with an infrastructure operator running the facility
The human layer

Fully autonomous. Human led.

Autonomous does not mean unattended. A CoE has a named human layer, and it is smaller, more senior and more durable than the floor it replaces.

Operations lead

Owns the queue, the SLA and the scope. Decides what the CoE takes on next, and holds the mandate inside your organisation.

Agent supervisors

Watch the live floor, intervene on escalation, and handle the judgment tail the agents route out. The human in human-in-the-loop.

Domain experts

Correct the system where it is wrong, and each correction becomes a labelled example. This is the highest-value seat in the building.

Model evaluators

Test against audited outcomes before anything is promoted, and re-test after every base-model swap.

Compliance and risk

Certify agents against policy, review the record, and produce the evidence examiners ask for as an output of the system rather than a project.

Exception handlers

Everything the agents are explicitly not permitted to close. Vulnerable customers, hardship, disputes above a threshold, anything novel.

The seats that remain move up the value chain rather than out of the building — from answering the call to supervising, correcting and certifying the thing that answers it.

Two routes

Inside your enterprise, or in your market.

Route A · Captive

A CoE inside your enterprise

Stood up within your own perimeter, usually by converting an existing shared service centre. You are operator and controller at once, which makes this the cleanest environment in the market for cross-department transfer — learning moves from collections to servicing to procurement on a single signature.

  • Fastest path to value, because there is no second consent gate
  • The SSO stops defending its cost and starts owning the group's agentic layer
  • Scope expands department by department, no new procurement each time
Route B · In-market

A CoE built with an operator partner

Stood up in a market with a contact centre operator, integrator or shared-services provider who already holds the client relationships, the licences and the compliance posture. They bring the deposit and the distribution. We bring the four components.

  • Anchored by the partner's existing regulated client book
  • Commercial structures from gain-share through to joint venture
  • Offered to one operator per market at the deepest tier

See the partnership model →

How it is funded

Three balance sheets, one delivery hub.

A sovereign Center of Excellence needs a facility, and a facility needs capital. We do not fund it off our balance sheet, and you do not fund it off yours.

The three-party structure behind a sovereign Center of Excellence Three parties feed one delivery hub. A capital partner owns the real asset, an infrastructure partner owns and operates the ground facility, and we provide the AI operating layer as anchor tenant. All three converge on the Center of Excellence, which serves regulated enterprise clients in-region. CAPITAL PARTNER owns the real asset capitalised deal by deal INFRASTRUCTURE PARTNER owns and runs the ground fit-out, power, connectivity AI OPERATING LAYER owns the intelligence anchor operating tenant THE CENTER OF EXCELLENCE one delivery hub, in-region Regulated enterprise clients — data never leaves the jurisdiction

Capital owns the asset. Infrastructure owns the ground. We own the intelligence that makes both worth owning.

01

Sovereign by construction

The facility is in-region by design rather than by configuration. Data never leaves the jurisdiction.

02

Funded as a real asset

Capitalised deal by deal by a real-asset partner. Capacity scales with demand rather than with anyone's fundraising.

03

A known playbook

Modelled on how the region's largest real-asset managers already develop, operate and divest data centres. Not a first-of-its-kind bet.

04

Demand unlocks the capital

Signed enterprise operations are the anchor tenancy that makes the facility financeable. The AI layer turns the real estate into a return.

The network

One centre, then a network.

Each CoE is sovereign to its market. What connects them is the operating layer, not the data.

Phase 1

Anchor

Stand up the first centre against contracted enterprise demand, in one market, with one capital and one infrastructure partner. [Location TBC.]

Phase 2

Replicate

Roll the same structure into adjacent markets with local real-asset and infrastructure partners. The template is the product.

Phase 3

Network

Operate a network of sovereign centres across APAC, sharing the operating layer while every record stays in-region.

Target markets: [Singapore as hub · Malaysia · Indonesia · Philippines · Vietnam · Thailand · Korea · Japan · India · Australia]. Sequencing follows regulated demand and partner availability rather than market size.

What crosses a border is learned structure. What never crosses one is a record.

Sequence

How one gets built.

StageTimingWhat happens
Assess[Weeks 0–2]Scope the offices in play, baseline the current cost per resolved task, resolve the data-rights position, and size the facility requirement if one is needed
Instrument[Weeks 2–6]Unify the record across systems. One account, one history. Nothing is automated until an outcome has something to attach to
Deploy the first queue[Weeks 6–10]Front-office, highest volume, lowest judgment. Runs in parallel with the existing operation until the SLA threshold is met
Hand over the Studio[Weeks 10–16]Your operations leads take control of configuration, policy and release. Training is part of the deployment, not an upsell
Extend inward[Ongoing]Middle office, then back office, each starting from the record the previous scope built rather than from zero

The facility, where one is required, is capitalised in parallel — it is not on the critical path for the first queue.

Get started

Build a Center of Excellence in your market.

Tell us the market, the regulated workloads, and whether the capital and ground partners are already in the room. We map the stack, the ownership structure, and a phased path to standing it up.

A Center of Excellence isn't bought. It's built.

Partnership · AI Center of Excellence

You have the deposit. We have the refinery.

An AI Center of Excellence takes four things: a delivery studio, an agentic AI workforce, vertical models tuned to regulated work, and sovereign compute to run them in-region. We bring all four. You bring the client relationships, the licences, the operating history and the corpus nobody else can assemble.

2008
Operating since
176
Enterprise customers
[7]
ASEAN countries
120+
Languages
125.6%
Net seat retention
[US$68M]
FY25 revenue
Who this is for

The operators who move first stop being the target.

Elsewhere on this site we make a hard argument: the labour-arbitrage model is finished, and the enterprises buying it know. We stand behind that.

The argument is about a cost structure, not about the people who built the industry. An operator with eighteen years of regulated delivery history, licences across several jurisdictions and a dialect-rich corpus nobody can buy holds assets we cannot manufacture and would not try to. What that operator lacks is the refining method, and the method is the only thing we sell.

Partner now and you are not displaced. You become the layer everything else runs through.

Your position

An inventory nobody can buy.

Relationships inside regulated perimeters

Banks, insurers, telcos, lenders, government bodies. Trust that took years and cannot be shortcut by a software vendor cold-calling a CIO.

Contractual access to regulated data

Under DPA, under licence, under audit. The access itself is the scarce thing.

A dialect-rich corpus

Hundreds of thousands of hours in languages and registers where frontier models are measurably weak.

Action-conditioned history

Not only what customers said, but what your people did in response and what followed. The data class that cannot be scraped or synthesised.

SLA credibility

You have signed contracts with penalties and met them. Enterprises buy that history, not a demo.

A client book with an expiry date

They will demand AI within [12–24] months regardless of what you do. That is an asset today and a liability the moment someone else answers it.

Three kinds of operator

One squeeze, three shapes.

BPOs sell hours. Integrators sell days. Shared services defend a budget. All three are being deflated by the same technology, and each needs a different way across.

Outsourced · BPO

Selling intelligence, not headcount

Your billing unit is the seat, and every automation win shrinks it. The way out is a second revenue line that grows as the first one compresses — products sold into the same client book, at a margin the seat never carried.

  • Processor under DPA — consent is a two-party question
  • Data fans out across many external controllers
  • All four layers apply
Captive · SSO / GBS

From cost centre to capability centre

Automation targets arrive from the board without a platform to deliver them, and every budget cycle reopens the question of outsourcing the function entirely. The answer is positional: stop defending the cost and start owning the group's agentic layer.

  • One controller, many departments — a single legal entity
  • Learning crosses finance, HR, procurement and servicing without a consent gate
  • The cleanest environment in the market for cross-department transfer
Advisory · SI

A practice to build, not a product to resell

Same squeeze on a different billing unit — you sell days and AI compresses days. Reselling a global platform leaves thin margin and nothing to differentiate on, because every rival in the bid quotes the same stack.

  • Certified implementation, verticalisation, managed operations
  • Sovereign on-premise delivery a cloud-only vendor cannot match
  • Annuity beside project fees — delivery margin expands rather than compresses

A shared services centre is the only place in the enterprise where intelligence can legally cross every department.

What we bring

Four components. We bring all four.

Studio

The delivery layer

Where operations are designed, agents configured, workflows QA'd and changes shipped. Your ops leads run it after [X] weeks. Not a black box you rent.

Agentic AI

The agent workforce

Voice, chat, back office, orchestration. 100+ agent roles, live in [2–4] weeks against pre-built libraries.

Vertical models

Tuned to regulated work

Trained on your corpus, evaluated against your audited outcomes rather than public benchmarks. Deployable on-premise.

Compute

Sovereign, in-region

Data does not cross a border. Capitalised deal by deal by a real-asset partner, with an infrastructure operator running it. Not on your balance sheet.

What we build together

Stop selling labour. Start selling intelligence.

Four layers. Each adds revenue before it touches a seat. You can stop at any rung and the economics still work.

Contact centre. Agent copilot with real-time next-best-action, live compliance guardrails and sentiment nudges. 100% QA coverage replacing [2–5%] sampling. Dialect-aware voicebots on high-volume, low-judgment queues. Automated after-call work and disposition. Live supervisor assist while the call is still open.

  • Optimal contact-time modelling on outbound
  • Verified-caller identity registration [carrier and market dependent]
  • Campaign sequencing and channel selection
  • Photo-to-structured-data for field and retail visits
  • Voice-note-to-report, in dialect
  • Route and visit-plan optimisation
  • Auto-scoring of mystery-shopper reports against the client rubric
  • Visit-integrity detection — GPS, timestamp, photo metadata
  • Level 1 hiring screens, fully automated
  • Level 2 assessment, AI-assisted, human decides

By design: no facial-emotion scoring and no affect inference on candidates.

Commercial — gain-share on verified savings. [XX%] of the delta against an agreed baseline, [24–36] months.

Products built on the CoE and delivered under your name, into the book you already serve.

  • AI collections — contact timing, promise-to-pay prediction, hardship detection
  • eKYC, onboarding and AML — documents, liveness, screening, adverse media
  • Fraud and scam-call detection
  • In-language content moderation
  • Compliance surveillance across 100% of calls
  • Claims triage and first notice of loss
  • Dispute and chargeback handling
  • Retail-execution analytics, priced as software
  • Regulated-footprint auditing — branch, pharmacy, franchise standards
  • Agentic follow-up loops after any visit or interaction
Commercial — revenue-share on new product revenue. [XX/XX] split. You invoice the client.

Your relationships are a distribution channel that has only ever carried one product. Package the products above and sell them as software to the banks, telcos and insurers you already serve. You are the trusted counterparty with the SLA and the human-in-the-loop guarantee. We are the engine underneath.

You charge like a marketplace. The AI is the margin.

  • Co-branded or white-labelled product set
  • You hold the client contract and the SLA
  • We hold the engine, the evaluation and model governance
  • Pricing moves from cost-per-seat to per-outcome or per-licence
Commercial — channel margin or marketplace take-rate. [XX%] to you on gross product revenue.

Not every partner reaches this rung, and we do not offer it twice in the same market.

  • Speech stack — ASR and TTS for the market's primary language and major dialects
  • Evaluation and benchmark suite — whoever publishes the eval sets the bar
  • Consented, de-identified data licensing to model builders
  • Sovereign compute as an asset rather than a cost centre
  • Regulator alignment — co-authoring the conduct standard for AI servicing
  • GLC and SOE demand — the state buys sovereign before it buys foreign
Commercial — equity, joint venture or royalty. Structured in the room, not on a website.
The seat question

The part nobody says out loud.

Every AI win in a contact centre shrinks billable seats. Anyone evaluating this partnership is doing that arithmetic in the first minute, and a proposal that does not name it is not credible.

We are compensated on the value created, not the headcount removed. We win when your multiple re-rates, not when your floor empties.

And the workforce is an asset in the AI economy rather than a liability to it. A few thousand domain-fluent, in-language people are exactly what frontier labs and every enterprise deploying agents are short of.

Data labelling and RLHF

Annotation and preference collection in scarce languages, sold to your own CoE and to global labs directly.

Agent supervision

The human in human-in-the-loop, at scale. Exception handling and the judgment tail that stays human.

Evaluation and red-teaming

Adversarial testing and quality assurance of AI output — a growing job, not a shrinking one.

How we work together

Four ways money moves.

ModelWhere it appliesStructure
Gain-share on savingsLayer 1[XX%] of verified cost reduction against an agreed baseline, [24–36] months
Revenue-share on new productsLayer 2[XX/XX] on product revenue. You invoice, we settle.
Channel marginLayer 3[XX%] to you on gross product revenue sold into your book
Outcome pricing with SLA penaltiesAny layerPer resolution, per collection, per verified claim. We carry the downside.

We do not take a management fee for showing up. Every model above pays us out of value that did not exist before the partnership.

Why we can price an outcome

Pricing an outcome requires knowing the counterfactual — what a normal recovery rate looks like by portfolio vintage, by delinquency bucket, by jurisdiction. That is an actuarial asset rather than a software asset, and it is why a software vendor quotes per seat.

Why your clients will accept it

Contingency pricing has been the norm in collections for nearly two decades. Outcome-priced AI is an extension of a commercial model your buyers already run, not a new experiment you have to sell them.

Equity, joint venture, territory licence and minority participation structures exist and we will walk through them — in the room, once there is something concrete to structure.

De-risking

Our side of the risk.

No seat reduction in Phase 0 or Phase 1. Written into the agreement.

Baselines agreed and instrumented before any gain-share is calculated. Champion/challenger design, so the attribution survives a renewal argument.

Your team operates Studio independently by [week XX]. Training is part of the deployment, not an upsell.

Data stays in-jurisdiction. On-premise or in-region, your choice, documented.

If Phase 0 misses the threshold, we stop — and you keep the assessment and the instrumentation.

No exclusivity demanded from you in Phase 0. [Territory exclusivity available at Layer 3, against volume commitments.]

Sequence

The first 100 days.

PhaseDaysWhat happensWhat you get
Assess0–14Baseline KPIs, map workflows, quantify the opportunity, resolve the data-rights positionA line-item savings model and a data-rights opinion
Prove15–45Deploy [1–2] Layer 1 use cases against a live queue, in parallel with your existing operationMeasured delta against baseline
Commit46–70Gain-share contract signed on the proven use cases; Studio training beginsRevenue from savings
Extend71–100Scope the first Layer 2 product against a named client in your bookA product with a buyer identified

No capex. No platform fee. Nothing invoiced until the delta is measured.

The ask

Four decisions.

None of these requires capital, board approval, or a change to a client contract.

01

A named sponsor

An executive on your side with authority over one operating queue.

02

One live queue

Highest volume, lowest judgment. That is where Phase 1 runs.

03

A data-rights position

Your counsel's read on what your DPAs permit, against the three rungs above.

04

A decision date

[DD Month].

Start here

Tell us what you run. We'll show you what it could earn.

Fourteen days, no fee, no commitment to proceed. We baseline your operation, model the delta, and return with a savings model and a data-rights read. If the numbers do not work, you keep the analysis.

Confidential, and used only to prepare your assessment. We respond within [24] hours.
[Form is not wired to a backend — connect to your CRM or form handler before publishing.]

Request received.

We'll come back within [24] hours with a scope for the fourteen-day assessment and the questions we need answered before it starts.

Same clients. Same data. A different business.