June 8, 2026
KEY takeways
For decades, Heads of Underwriting in Trade Credit, Surety and Trade Finance have lived with the same trade-off: analyse quickly or analyse well. AI is about to break the historical trade-off between speed and analytical quality in Trade Credit underwriting. That is the thesis Olivier Placca, CEO of Langano Technologies, defended at L'Année de l'Assurtech 2026. This article develops the thesis, illustrates it through the multi-agent architecture Langano Technologies has built, and draws out the consequences for risk carriers.
What a Head of Underwriting actually does for thirty minutes
In credit insurance, the risk is not the insured, it is the insured's client. At the point of underwriting, the underwriter is therefore not analysing one risk but potentially hundreds of counterparties, each with its own financial structure, payment history, sector exposure and legal risk profile.
Every file imposes the same manual sequence: open several systems, calculate financial ratios, examine payment trends, search for recent news, scan for legal signals, write up the synthesis and propose a limit. On a substantial file, an experienced underwriter spends on average 30 minutes on this work.
Olivier Placca puts the problem in one sentence:
"There is always this trade-off problem between speed and quality, and that is a problem."
Three consequences follow, and they are familiar to every Chief Risk Officer and Chief Underwriting Officer. The volume an underwriter can handle reaches a ceiling. Quality drops when pressure on deadlines rises. Between two annual reviews, the portfolio remains exposed to blind spots.
Previous generations of tools never lifted this friction, because it was not located in data collection but in the human digestion of that data. That is precisely the point AI shifts.
How AI breaks the trade-off: six agents, one orchestrator
The Langano Technologies technical answer is called the Virtual Underwriter. Its architecture fits in one line: Six agents. One orchestrator.
Six specialised agents, each one master of its domain. Company Intelligence reconstructs the corporate profile and health of the counterparty. Financial Data Retrieval pulls the financial statements and calculates the ratios. Document Processing extracts clauses and points of attention from the contracts and PDFs in the file. Web Research detects weak signals in the press, public legal databases and registries. Portfolio Risk measures the impact of the new exposure on portfolio concentration and correlations. Credit Risk Assessment consolidates the outputs of the five other agents and produces the final scoring with its limit recommendation.
At the centre, an orchestrator receives the file, selects the relevant agents, delegates the tasks in parallel, aggregates the results and delivers the argued synthesis. This orchestrator operates as a conductor, not as a router. It is the orchestrator that makes the agents dialogue, resolves contradictions, ranks signals, and traces the chain of reasoning that makes the decision auditable.
The official flow fits in one line: File → Orchestration → 6 agents in parallel → Scoring → Argued decision.
The synthesis that took 30 minutes manually is produced in under 3 minutes. An underwriter can therefore handle 10 times more files, at equivalent or higher quality. The speed-quality trade-off disappears: the reading-synthesis step, which was its critical bottleneck, is no longer human. Orchestrated parallelisation takes its place.
One important precision. The output produced remains a documented analysis, not a decision. The underwriter is still the decision-maker. They arrive on the file with a synthesis already built, on which they apply their judgement. They no longer start from raw data, they start from intelligence.
What the Virtual Underwriter delivers in under 3 minutes
For each file, the platform delivers in under three minutes:
- an analysis of the counterparty's financial strength,
- an evaluation of liquidity trends,
- a benchmark of sector risk,
- a mapping of litigation exposure,
- a reading of payment behaviour,
- detection of early warning signals,
- a measurement of the impact on portfolio concentration,
- a limit recommendation, accompanied by its rationale and confidence score.
Explainability: a non-negotiable requirement in underwriting
On this point, Olivier Placca is categorical:
"We are in a business where it is non-negotiable, it has to be explainable. AI cannot be a black box, and if we do not have AI that is explainable, we cannot use it, at least not in our domain."
This requirement operates at three levels. Solvency II and equivalent regimes impose auditability on decisions that engage capital. Internal risk committees must verify the coherence of recommendations and identify potential biases. The underwriter must be able to defend a limit before the broker, before the insured, or in the aftermath of a loss.
The Virtual Underwriter builds this requirement into its architecture. The orchestrator traces the chain of reasoning and produces the rationale attached to each recommendation. That rationale spells out the signals mobilised, the contribution of each agent, their relative weight and the degree of confidence attached to the conclusion.
The underwriter receives a structured line of reasoning, not a verdict. It is this explainability, agent by agent, that separates an AI-Native platform from an AI module bolted on to an existing system.
AI hallucinations: what the word really hides
Olivier Placca questions the common use of the term hallucination:
"When we get answers that do not correspond to what we expect, we tend to say it is hallucination. The reality is that when you dig a little, you realise there is a logic behind it, and that it is producing a decision that is not in our context. Our job, when we do context engineering, is to fix that and to reduce these hallucinations."
An answer that does not satisfy an experienced underwriter rarely points to a flaw in the model. It points to three upstream weaknesses, which Olivier Placca names in his intervention: poorly defined context, badly calibrated prompts, input data of insufficient quality.
The Langano Technologies response rests on the specialisation of agents. Each of the six agents of the Virtual Underwriter operates on a precise business perimeter, with its own sources, its own calibration and its own validation rules. It is this specialisation, rather than a single generalist model, that anchors decisions in the real context of Trade Credit, Surety and Trade Finance.
What changes for the underwriter: less data entry, more judgement
The productivity gains do not reduce the underwriter's role, they redefine it. Olivier Placca puts it directly:
"To be a good underwriter, you need four elements: judgement, a sense of synthesis, experience, and a bit of intuition and flair."
These four dimensions, human by nature, are the ones AI cannot replace. What AI takes over lies elsewhere: the repetitive and time-consuming tasks that fill most of the 30 minutes of the average file. Olivier Placca summarises:
"AI will remove all those somewhat tedious tasks that are low value-added, and focus the human on bringing the right judgement to make the right decision."
The result is a transformation of the role that Langano Technologies captures in one formula: from data processor to risk strategist. The underwriter does not disappear. They refocus on what justifies their seniority.
This transformation opens up an additional operating mode, until now ruled out by the time the manual analysis cost. The Portfolio Risk agent of the Virtual Underwriter continuously measures the impact of each exposure on portfolio concentration and correlations. This permanent surveillance enables what Langano Technologies calls continuous underwriting: adjustments to decisions and detection of early warning signals that no longer wait for the annual review window.
Speed is no longer an attribute, it is a survival variable
Olivier Placca closes his intervention with a direct formulation:
"AI is a big game we are all playing, and the name of the game is speed. In a competition where your competitors go 10 times faster than you, you have no chance of succeeding."
The formula translates a tangible commercial reality. For a Trade Credit Insurer, the speed of approval determines the level of service delivered to the broker and the satisfaction of the insured. For a Surety carrier, the same equation applies to the issuance of bonds on time-sensitive projects. For a Trade Finance institution, the speed of decision determines the ability to capture opportunities with short windows.
Carriers that do not embrace this transformation lose more than operational efficiency. They lose market competitiveness. Speed, however, only has value if it is precise and explainable.
Conclusion
The trade-off between speed and quality has never been a truth of the underwriter's profession. It was a constraint of the tools available to the industry. With explainable AI, that constraint lifts, and the profession returns to its original nature: exercising judgement on a risk.
The question now facing risk carriers is no longer whether they should integrate AI into their underwriting processes. It is how quickly they can do so, and with what level of requirement on explainability.
Going further
To go deeper into how Langano Technologies industrialises risk analysis for Trade Credit, Surety and Trade Finance, you can consult the detailed presentation of the Virtual Underwriter or speak with the team for a scoping conversation tailored to your portfolio.
