AI-native companies,
built block by block.

A company is a set of blocks.

Every business has a few pains that outlived the tools it bought to fix them - the account that cancels without warning, the invoice chased for ninety days, the real risk lost in a queue of false alerts. They persist because no single department's software owns them: each one crosses the seams - sales, service, delivery, finance, compliance.

Each is a use case we build as a block - a closed loop around one business object, running the same five verbs every time: sense, correlate, judge, govern, act. An ai-native company is assembled from these loops. That is what the line on our masthead means, made literal: built block by block.

§ 01 Where it's proven

Proven where the bar is highest - in banking, under regulators, auditors and risk committees.

Being wrong has consequences in every field. This is where we have proved it so far.

  1. In delivery
    Global retail bank · financial crime & compliance

    A detection system that finds criminal money-movement patterns the rules never caught - built, validated, and handed to the financial crime team to run.

    read more read less

    Installing a full detection pipeline end-to-end: graph construction over the transaction population, community detection to find natural groupings of accounts, and a sequence model trained to recognise the behavioural signatures of twelve criminal typologies - structuring, layered chains, mule networks, trade-based laundering, professional money laundering and others. Hard Concrete gates to control what the model attends to. Low-rank adaptation heads so the model can be steered to new typologies without full retraining. A predicate router that maps model output to investigator-ready case narratives. Everything built against the bank's own data, not a benchmark.

    On handover a detection capability the financial crime team operates directly, with explainable outputs an investigator can read and a regulator can examine. Period one covered nearly a hundred thousand accounts, surfaced over six thousand reportable entities, and produced a separation ratio that makes the prioritisation meaningful rather than nominal. The team owns the pipeline, the weights, and the architecture. No licence, no dependency, no ongoing seat.

    ten weeks · full pipeline build

  2. In delivery
    International wealth & retail bank · operations, risk & compliance

    A board-ready plan for practical AI across the back office - the highest-value opportunities, sense-checked for feasibility and risk, and sequenced for where to start.

    read more read less

    Installing a focused four-week assessment across the operational and control teams - risk, compliance and customer operations - reading the current processes and constraints, running a working session with the people who own them, and testing each opportunity for value, feasibility and risk.

    On handover a prioritised set of practical AI opportunities, with a clear recommendation on sequencing, governance and next steps - built to lift consistency and free capacity, not just cut cost, and grounded in the bank's own priorities.

    four weeks · assessment sprint

  3. In delivery
    Middle East retail bank · operations & processing

    Intelligent document processing and reconciliation, taken live end-to-end in five weeks - a pilot to prove the approach before building further.

    read more read less

    Installing a five-week pilot that takes one operations use case - intelligent document processing and reconciliation - end-to-end against the bank's real workflows and volumes, built to run on live cases, not demo on slides.

    On handover a working capability the operations team runs and owns, with outcomes measured against the value hypothesis and a clear, evidence-based read on whether to scale into a longer-term engagement.

    five weeks · pilot

  4. Completed
    Major European retail bank · everyday banking

    The bank now has a build-ready everyday-banking strategy - a prioritised set of ai-native propositions, sequenced by value and feasibility.

    read more read less

    Before the everyday-banking AI question sat as scattered point ideas, with no shared basis for deciding what to build, in what order, or how it fit an ai-native operating model.

    Now a prioritised portfolio of ai-native propositions - each defined, pressure-tested against strategy, regulation, and the competitive field, with a build case and sequencing - owned by the bank and actionable without us.

    strategy definition

§ 02 The use cases

What we build - one loop for each object a company runs on.

Each use case is a block: a closed loop around one class of business object, its sensors and actions distributed across the functions the object touches. We build them one at a time, and your team owns each at handover.

The account block

The object: The customer relationship, continuous from first signal to last renewal.

"We find out an account is at risk when the cancellation arrives. Growth means hiring more account managers."

How the account block works →

The cash block

The object: The money owed - quote, bill, collect, dispute, resolve.

"Profitable on paper, tight every quarter. We do the work, then chase the money for ninety days."

How the cash block works →

The reporting block

The object: Each management reporting cycle - the pack the exec and the board actually run on.

"The report takes weeks to build, and by the time it arrives the numbers have been shaped by everyone who touched it."

How the reporting block works →

The obligations block

The object: Each regulatory and contractual commitment, from the day it lands to the day it is audited.

"The regulator asked, and it took six weeks and four people to assemble the answer."

How the obligations block works →

The delivery block

The object: Each unit of promised work - matter, claim, order or mandate, from instruction to completion.

"The write-off surprises us at month end. Scope creeps in the inbox, never in the system."

How the delivery block works →

The financial crime block

The object: The counterparty's risk state, continuous from onboarding to exit.

"Analysts drown in alerts that go nowhere, and real risk hides in the queue."

How the financial crime block works →

The subscription block

The object: The recurring contract, cycle by cycle - from activation through each renewal to lapse or win-back.

"Churn arrives as a number at month end with no name on it."

How the subscription block works →

The capacity block

The object: Each person and role - load, capability, certification, dependency and attrition risk.

"Everything routes through the same three people, and I'll learn one of them is leaving when it's too late."

How the capacity block works →

The third-party block

The object: Each supplier, introducer or outsourcer - from onboarding through performance to exit.

"We discovered the dependency the day it broke."

How the third-party block works →

The product block

The object: Each product or service line - design, price, performance, suitability and retirement.

"Nobody can tell me which lines actually make money, or prove the product is fair when asked."

How the product block works →

The resilience block

The object: The important business service and its incidents - from normal operation through disruption to recovery.

"When something breaks the regulatory clock starts before we have understood what happened."

How the resilience block works →

And one view beneath them all - the operator's cockpit. Not a loop of its own: the live state of the business, read off the same record the blocks run on. No cycle, no assembly, no sign-off - you look, and it is current. It isn't bought separately; it ships with the first block you build, whichever that is. How the cockpit works →

§ 03 What a block is

Every block runs the same anatomy - the one described in our framework.

THE LOOP · SENSE, CORRELATE, JUDGE, GOVERN, ACT governed Receptors SENSE Nervous system CORRELATE Company Brain JUDGE Immune system GOVERN Muscles ACT
One loop, five organs. Every block is this shape.
  • Signals are sensed once, and owned by no block. An ageing invoice, a departed sponsor, a shift in ticket tone - each is captured once at the receptors, indexed to the objects it touches, and read by every loop that needs it. No block brings its own sensors. That would recreate the silos.
  • One brain across all blocks. Every loop judges from the same shared, continuously-learning memory. The cash loop and the account loop are reading the same customer.
  • No ungoverned action leaves a block. Every response passes the immune gate - authority, terms, tone, and a human gate wherever the rules demand one. Where the human sits is a setting you control, not an accident of the org chart.

A block is not a tool. A tool automates a task inside one function; a block is the loop itself - sense to act - closing across the functions a business object actually touches. That is the structural difference, and the reason a block reaches the pains a tool cannot.

§ 04 Blocks compound

Blocks compound

The ageing invoice is one signal. The cash loop reads it and sets the tone of the chase. The account loop reads the same signal and holds the renewal uplift back - this customer is owed a conversation before they are asked for more money. One signal, sensed once, made useful twice.

That is the argument for blocks over point solutions. A churn tool and a collections tool would each see half of this and act on neither. Because every block reads the same spine, each block you add makes the others sharper - the second block is better than the first from the day it goes live, and cheaper to add, because the anatomy is already there.

§ 05 Built around the systems you already run

Built around the systems you already run

Blocks read from the systems you already have: email and calendars, the CRM, practice, case or policy systems, billing and finance ledgers, document stores, payment providers, screening and regulatory feeds.

Nothing is replaced and nothing is migrated. Connectors are built or adapted per engagement, around the block being built - the named list, available and coming, is on our framework - and no eighteen-month data programme stands as a precondition.

§ 06 Where to start

Where to start

You don't pick a block off a list. It starts with a diagnostic that maps your firm's blocks: where the pain concentrates, which signals already exist in your systems, which loop to build first - and why.

You leave with your block map, and the first block scoped. From there, each block is a bounded engagement - blueprint, build, handover - owned by your team at the end.

Start with the diagnostic. Take the ten-minute diagnostic (opens in a new tab), or write to hello@somai.studio with the pain that made you read this far.