Artian

The removal test leaves rigid process automation, which the company positions itself explicitly against. A planner reads an institution's runbooks, standard operating procedures and business process graphs and generates hundreds of interconnected agents from them, a natural language builder produces new agents from description, and a multi-agent orchestration layer coordinates models, tools, people and specialised agents across a workflow. Generating a working agent population from documented process artefacts is not achievable by configuration.

Oversight is treated as a design requirement rather than a disclaimer. The orchestration system is described as built for human oversight and governance, and notably it coordinates models, tools, people and specialised agents together, which places humans inside the workflow graph rather than outside it.

Customers configure workflows around their own rules, approvals and operating requirements, so the approval points are institutional rather than vendor defined, and execution is stated to be kept visible, explainable and easy to oversee. Against that, agents automatically resolve recurring incidents and are described as autonomous, and nothing states what class of action proceeds without approval.

The company names model risk governance as a built-in property of the platform rather than leaving it to the customer, which is unusual and directly relevant given that its buyers must satisfy supervisory expectations for any model influencing decisions.

Execution is described as visible and explainable, and the founders' backgrounds in deploying model driven trading systems at scale lend the claim credibility, since they have operated under exactly the validation regimes their customers apply. Held at B because governance is asserted as a feature without any described mechanism, and no accuracy, reliability or agent failure rate is published despite reliability being the company's central promise.

No customer is named and no deployment result is published, so the entire evidence base is team and backing. That base is unusually strong: the founders each bring over twenty five years at the intersection of artificial intelligence, financial markets and enterprise technology, one having been head of artificial intelligence and data at a major American bank with prior roles at a large technology company and another bank and a doctorate in machine learning, the other having run electronic trading and market making businesses at two Wall Street firms and a digital asset group. A former chief data and analytics officer of a global bank is associated with the company. Eight million dollars was raised, six of it in a seed led by an enterprise focused venture firm.

No boundary statement was located, and one roadmap item makes the question unusually concrete. The company describes an agent exchange enabling collaboration across use cases and, in its own words, eventually across institutions.

Agents generated from one bank's runbooks and procedures encode that institution's operating knowledge, so an exchange spanning institutions raises directly what is shared, who owns an agent built from proprietary process documentation, and whether a competitor could run something derived from it. The ambition is stated and the governance around it is not.

No data protection agreement, retention schedule, subprocessor list or deletion commitment was located. Data risk governance is named as a built-in platform property without being described. The reach is wide by design, since agents operating across trading operations, wealth relationship management and incident response touch client instructions, account information and internal procedure documentation, and nothing states how any of it is handled or segregated.

No attestation, certification, trust centre or enumerated framework was located, with security described as being at the core of the design. The company is targeting the most demanding supplier assessment environment there is, since large banks gate anything touching trading operations, and an assessed control set is the practical prerequisite for the first production deployment.

No regulator, statute or rule is named. The lead investor states the platform is designed to operate within regulatory constraints, which acknowledges the environment without identifying any part of it, and the workflows involved sit in supervised territory: pre-trade handling in capital markets sales, reconciliation in trading operations and adviser client relationships each carry their own conduct and record keeping obligations. For a platform whose differentiation rests on being production ready in regulated environments, naming those obligations would be the natural proof.

No individual is scored and the exposure is indirect but real, since agents operating in wealth relationship management sit between advisers and clients, and agents generated from existing runbooks inherit whatever the documented process already encoded, including any uneven treatment embedded in it. Automating a procedure faithfully reproduces its assumptions at greater scale and speed. Nothing describes review of generated agents for that effect, and no testing, outcome analysis or fairness statement was located.

No guarantee, indemnity or correction process was located. Visible and explainable execution gives the institution the means to reconstruct what an agent did, which is the foundation of recourse without being recourse itself. Nothing states what the vendor owes when a generated agent resolves an incident wrongly, mishandles a pre-trade request or acts on a misread procedure, how such failures surface, or what accountability attaches to agents the platform generated rather than a person wrote.

The platform is described as coordinating models in the plural, implying several underlying providers, and none is named, nor is any hosting arrangement or subprocessor list published. That gap sits awkwardly beside the company's own claim to built-in model risk governance, since model identity, version and change control are the first things any such governance framework must record, and a bank adopting the platform would need them before validation could begin.

The company identifies the problem correctly, naming workflow disconnects as the reason agents become isolated and fail to deliver end to end value, and its lead investor emphasises integration with legacy systems as a design requirement.

The distinctive integration is upward rather than sideways: the planner ingests runbooks, standard operating procedures and business process model graphs, which are the artefacts a large bank already maintains to document how work is done, so the platform connects to institutional process knowledge rather than only to software. Interfaces serve both developers and business users. No named system, platform or vendor appears.

No hosting provider, region selection, residency commitment or private deployment option was located. This matters more than usual for the intended buyer, since agents operating inside trading operations and client relationships at a large institution would normally require deployment within that institution's own environment, and nothing states whether that is supported.

No pricing, packaging or basis of charge was located. The platform spans a builder, an orchestration layer, prebuilt solutions and a planned exchange, which are different commercial objects, and nothing indicates whether charge falls per agent, per workflow, per seat or as an enterprise platform licence. For a product whose value proposition is generating hundreds of agents, per agent pricing would be a material consideration.

The buyer is one type, large financial institutions and specifically their technology leadership, in one market. Functional ambition is broader, covering capital markets sales and trading, wealth management and general enterprise operations, with named workflows spanning pre-trade request handling, reconciliation break resolution, adviser relationship management and operations incident response.

Those are genuinely different domains and the platform is horizontal within the institution rather than within the market, so coverage is a stated scope rather than a demonstrated footprint at this stage.