When Offloading becomes a modernization enabler

TAS is focusing on an internal community to transform skills, processes, and operating models. From compliance to the knowledge base, through to application development, AI is becoming a widespread and structured workstream.

An internal community spanning across the organization is driving seven development workstreams in the field of artificial intelligence. The objective is twofold: to spread technological culture and skills within the organization and, at the same time, to rapidly bring concrete AI initiatives into execution. "The initiative represents the evolution of TAS Lab, which started in 2023 as an experimental space and has now been transformed into a more structured and widespread community model," says Fabrizio Annatelli, AI Program Lead, alongside Gabriele Paoleschi, from the TAS AI Community. "Adopting AI on a large scale is a journey that involves structure, processes, and above all, the organization. Therefore, we need to bring people on board, support them over time in understanding the technology's potential and risks, and develop new mindsets and skills."

Involving the company's workforce

Hence the explicit focus on reskilling, considered a key lever for managing the transformation. "Not riding the AI wave today means risking falling behind. But it’s not about replacing people: the scope of skills is shifting, as is the way they are utilized," underlines Annatelli. "TAS's response was to directly involve the company's workforce: the community, open to everyone, has seen widespread participation from employees across the company."

The TAS Community

The profiles involved are diverse: from developers to business unit and department managers. Everyone participates in the seven active AI workstreams, directly contributing to project definition and execution. "In this initial phase, we are focused on internal transformation and process efficiency," notes Annatelli. "We have structured pathways, organized into different sessions, to align the community on objectives and build a concrete execution plan. AI is not just an initiative confined to IT; it represents a cross-functional lever that requires organizational coordination and a clear structuring of processes."

The three key workstreams

Among the ongoing projects, three main directions emerge. The first concerns regulations and is among the most complex. "Working with banks and insurance companies," Annatelli premises, "a priority is to strengthen internal compliance safeguards and build a framework that clarifies what can and cannot be done with AI, in light of a continuously evolving regulatory landscape." Alongside compliance, the application development workstream leverages AI to support coding, in compliance with the principles of explainability and ethics. This is a more mature area, partly thanks to the widespread use of AI in software development, where the human-in-the-loop model remains central. However, the most strategic project is the corporate knowledge base. "AI models are generic by nature," highlights Annatelli. "Value is created when they are tailored to the specific context of the company. For this reason, we are building a structured knowledge base that includes regulatory, application, and business content. The knowledge base does not only serve to train models but enables a plurality of use cases: from proactive ticket management to generating analyses, all the way to supporting operational activities. AI agents specialize in specific tasks but do not make autonomous decisions: the responsibility always remains with humans."

From experimentation to industrialization

The community has already brought several workstreams into the execution phase and is preparing for an advanced training phase, particularly on the knowledge base. "We are structuring the system to ensure localized data governance for critical content. For the most sensitive use cases, a dedicated LLM infrastructure based on open-weight models is being trialed, with an initial summary expected by May," says Gabriele Paoleschi. "This allows us to securely manage large volumes of critical documentation. Another key element is the evolution of the operating model: no more rigid technological constraints, but adaptive processes, because AI evolves too quickly to be locked into static architectures. We need a dynamic approach capable of continuously updating itself, also in line with regulatory developments."

The first Use Cases: Between Electronic Payments and T+1 Settlement

Alongside the main workstreams, TAS is developing vertical applications, particularly in the electronic payments sector. Predictive artificial intelligence is already being used for transaction monitoring in fraud prevention using mature techniques. Generative AI enables a broader scope of applications: from the corporate knowledge base to decision support for T+1 settlement, from an advanced GUI design framework for banking applications to advanced reporting systems based on natural language. "A concrete case concerns the transition to the T+1 regulation, which reduces financial transaction settlement times to one day," explains Paoleschi. "We have developed proofs of concept to support matching and reconciliation activities, which are increasingly critical today due to shortened timelines, and here AI provides decision support by intercepting anomalies and suggesting possible corrective actions."

A look at the market

Finally, TAS continues to closely monitor market developments to integrate AI into its offerings in a way that aligns with customer needs. "We are in a phase of continuous scouting," concludes Annatelli. "AI is evolving rapidly and requires a pragmatic approach to be integrated into the business at scale. It is not a vertical technology, but a cross-functional capability that must be structured and governed."

English version by TAS

Original article from AziendaBanca by G.C. : https://www.aziendabanca.it/notizie/tecno/tas-nasce-la-community-per-scalare-ai

Within the Capital Markets industry, Intelligent Applications are rapidly evolving from a technology paradigm into a strategic lever for operational efficiency and value creation.

These applications represent the convergence point between traditional functionalities and advanced AI and Machine Learning capabilities, enabling new levels of automation, adaptability, and actionable insights.

This paradigm is built on five key design principles that guide both development and adoption.

Dynamic user experience enables applications to adapt to operational context and user profiles, delivering personalized interfaces and workflows. Embedded intelligence integrates AI directly into business processes, making it a native component of execution, control, and analytics activities. Autonomous orchestration enables end-to-end management of complex processes by combining automation, process mining, and adaptive workflows to reduce manual intervention and improve exception handling capabilities. A modular and reusable architecture, based on composable logic and API-first principles, enables the development of flexible and scalable solutions that can be rapidly integrated into existing ecosystems. Finally, unified and contextual operational data management helps overcome fragmentation and application silos, enabling advanced analytics and AI-driven functionalities.

While these principles define a clear architectural framework, the real challenge today is translating this vision into concrete use cases capable of generating measurable impact across core processes and the overall operating model.

BOSS: The Intelligent Application for Capital Markets Back Office

In this context, the evolution of our BOSS solution represents a key step forward. BOSS has been designed as an intelligent ecosystem supporting the entire front-to-back value chain within Capital Markets. The objective is clear: to bring the principles of Intelligent Applications into securities and derivatives back-office processes, historically characterized by high operational complexity and significant cost pressure.

BOSS will natively integrate AI capabilities — including LLMs, Machine Learning, and Generative AI — into operational workflows, alongside process mining, RPA, and agentic orchestration components. This approach not only automates repetitive tasks but also introduces predictive and advanced decision-support capabilities, particularly in processes such as settlement, reconciliation, and corporate actions management.

At the same time, the adoption of a modular, API-first architecture enables a composable approach in which capabilities can be progressively introduced, reducing time-to-value and simplifying integration with existing application ecosystems.

Reducing Cost to Serve: From Operational Efficiency to Scalability

The primary value driver behind this evolution is the reduction of Cost to Serve across back-office operations. In the sell-side environment, this cost is largely driven by human resources, which typically account for approximately 50–60% of the total.

This is precisely where Intelligent Applications deliver the most significant impact. Intelligent automation and process orchestration enable organizations to structurally reduce operational effort, while simultaneously improving process quality and resilience.

Analyses conducted across the main use cases show that operational effort reductions in the range of 20–30% are achievable, generating substantial economic benefits or, alternatively, enabling business volume growth without proportional increases in headcount.

Importantly, the value extends beyond direct cost savings. The real transformation lies in the ability to industrialize processes, reduce dependency on manual activities, and enable more scalable and data-driven operating models.

From Principles to Use Cases: Where Value Is Generated

The transition toward Intelligent Applications is already materializing through a range of use cases identified within BOSS, targeting the most operationally intensive areas of the back office.

Among the most relevant examples, intelligent automation of trade ingestion and validation significantly reduces manual activities and data-entry errors while improving upstream data quality. Downstream, anomaly detection models proactively identify inconsistencies and abnormal patterns, reducing operational risk and rework.

Another high-value area is settlement fail management, where Machine Learning capabilities enable true “Operations Copilot” functionalities. The application can automatically identify the root cause of issues, recommend corrective actions, and support operators during resolution, significantly reducing handling times.

In the areas of position and collateral management, predictive and optimization models improve both the quality of available information and the ability to support higher-value decisions, such as optimal collateral allocation, with direct impacts on funding costs and balance sheet efficiency.

Finally, the evolution toward operational “control tower” models, enhanced by anomaly detection and intelligent prioritization capabilities, enables proactive end-to-end process management, improving operational resilience and service quality. In this context, the introduction of natural language explanation layers and operational copilots represents a further evolutionary step, making data and insights immediately accessible even to non-specialist users.

The Opportunity

Intelligent Applications now represent a concrete opportunity to rethink the Capital Markets operating model. With BOSS, we are turning this vision into reality by developing targeted use cases that combine technological innovation with tangible business impact.

The journey is already underway: the next step will be scaling these capabilities and supporting our clients in adopting a more efficient, intelligent, and future-ready operating model.

Discover our Capital Markets offering

Over the past few years, the market for digital asset infrastructures dedicated to financial institutions has evolved rapidly. Specialized players have emerged focusing on execution, smart routing, custodian integration, and multi-venue access. At the same time, the European market is entering a new phase of maturity: MiCA, DORA, the DLT Pilot Regime, CASP authorizations, and developments related to stablecoins and CBDCs are progressively transforming digital assets from an experimental market into a potentially structural component of Capital Markets.

In this context, the challenge for banks is not simply gaining access to new instruments, but governing them within existing processes while preserving operational continuity, risk control, and compliance. This is where TAS’s value proposition is positioned: not to build a separate crypto value chain, but to help banks integrate digital assets, DLT infrastructures, traditional systems, and control processes into a single, coherent operating model.

From Crypto Trading to Operating Model Integration

Many of the platforms currently available on the market were designed with a predominantly “crypto-native” approach, primarily focused on enabling digital market access, execution, smart order routing, custody integration, and trading lifecycle automation.

From the perspective of European banks, however, the key challenge appears different: preventing digital assets from introducing a new operational silo disconnected from existing Capital Markets architectures.

Our vision is based on a clear assumption: in the short to medium term, digital assets are not expected to fully replace current infrastructures, but rather to progressively extend the bank’s existing operating model.

This does not mean underestimating the transformative potential of DLT infrastructures. Over the long term, features such as programmability, atomic settlement, and on-chain interoperability could drive profound changes in financial operating models. However, in the current phase, the prevailing model for European banks remains hybrid in nature: progressive integration between legacy systems, DLT, payment infrastructures, and existing control processes.

Market evolution confirms this direction. Increasingly, market participants are shifting their focus:
• from pure crypto trading
• toward tokenization of financial instruments
• regulated stablecoins
• digital bonds
• institutional custody services
• tokenized deposits and commercial bank money tokenization

A Truly Integrated Cross-Asset Platform

Our application proposition is built on the experience gained in managing mission-critical Capital Markets platforms for banks and financial infrastructures.

The objective is not to create a parallel “crypto stack”, but to extend the traditional Capital Markets value chain, enabling it to seamlessly interact with digital assets, tokens, stablecoins, wallets, and DLT infrastructures.

From this perspective:
• OMS, EMS, and execution components must evolve toward genuinely cross-asset capabilities;
• settlement must be capable of orchestrating both traditional and DLT/atomic settlement models;
• treasury and liquidity management must coordinate traditional accounts, wallets, collateral, stablecoins, and tokenized forms of money;
• compliance must integrate MiCA, DORA, and the new controls required by digital assets without multiplying separate operational and application frameworks;
• back-office functions must preserve reconciliation, control, accounting, and reporting capabilities even in the presence of heterogeneous assets and infrastructures.

The central point is therefore not adding a new application layer, but building an operating model capable of integrating traditional and digital assets across the entire transaction lifecycle.

Custody as the Core of the Operating Model

Within the institutional digital asset market, custody is progressively becoming the primary operational, regulatory, and commercial control point of the ecosystem.

The ability to govern:
• wallet orchestration,
• key management,
• policy engines,
• asset segregation,
• operational authorizations,
• integration with AML and compliance processes,

is becoming as strategically important as execution itself.

For this reason, custody — now assuming a central role — cannot be considered a separate or purely infrastructural layer. Instead, it must be natively integrated with treasury, liquidity management, settlement, collateral management, and operational controls.

Integration & Control: The Real Critical Layer

In the digital asset market, the challenge is not limited to “executing orders.” Complexity emerges when institutions must:
• coordinate legacy systems and DLT infrastructures,
• synchronize movements between accounts and wallets,
• manage funding and pre-funding,
• govern reconciliations,
• maintain auditability and regulatory controls,
• coordinate centralized venues, OTC channels, and decentralized infrastructures.

This is why the core component of our architecture is the Integration & Control layer.

This layer orchestrates distributed workflows, integrates legacy systems, coordinates financial flows, and governs operational events and exceptions, preventing the creation of technological silos.

It is this level that enables banks to progressively adopt digital assets within their operating model without compromising existing architectures or losing control over core processes.

Liquidity, Execution, and the New Market Structure

One of the most relevant characteristics of the digital asset market is the increasing fragmentation of liquidity.

Unlike traditional markets, liquidity is distributed across:
• regulated exchanges,
• OTC venues,
• specialized market makers,
• liquidity providers,
• decentralized platforms.

This environment makes capabilities such as:
• smart routing,
• venue selection,
• liquidity orchestration,
• execution policies,
• dynamic pre-funding management

increasingly critical.

Execution quality will depend less and less on simple connectivity to a venue and increasingly on the ability to orchestrate heterogeneous and distributed ecosystems.

Compliance and Regulation: A Framework Still Evolving

The European market is moving toward increasingly regulated and integrated models, but the regulatory framework cannot yet be considered fully stabilized.

MiCA represents a fundamental milestone for the issuance and management of regulated crypto-assets, while DORA strengthens the digital operational resilience framework. At the same time, the DLT Pilot Regime is creating the first regulated spaces for tokenized financial instruments.

However, the European framework is still evolving, particularly with regard to:
• security tokens,
• tokenized deposits,
• bank-issued stablecoins,
• on-chain settlement models,
• interoperability between tokenized money and traditional infrastructures.

In such a dynamic regulatory environment, architectural choices become decisive: institutions require modular, interoperable platforms capable of adapting to market evolution without forcing radical replacement of existing systems.

A Positioning Aligned with the European Banking Market

In this context, TAS positions itself as an independent technology vendor serving the bank’s operating model:
• it does not assume market risk,
• it does not custody assets,
• it does not operate as an exchange or intermediary,
• it enables banks to build their own digital asset operating model.

Furthermore, the economic value of the market is expected to progressively shift:
• away from pure execution,
• toward higher value-added infrastructure and operational services,
• liquidity orchestration,
• custody,
• collateral mobility,
• treasury integration,
• tokenization services.

The real challenge, therefore, is not creating a new crypto platform, but integrating digital assets, traditional systems, DLT infrastructures, and banking processes into a single coherent, governable, and scalable operating model.

This is where we believe the real competitive advantage of the coming years will emerge, and where TAS intends to deliver value: enabling the connection of different worlds without duplicating architectures, while preserving operational continuity and supporting the progressive evolution of Capital Markets toward new digital models.

Discover our Capital Markets offering

During the event “T+1: Implementation Has Begun”, Bancaforte spoke with David Mogini, Head of Capital Markets Business Unit at TAS, about the impact the new regulation will have on banks, market participants and the broader financial ecosystem, as well as the solutions that can turn compliance into a strategic opportunity.

Mr. Mogini, which processes will be most affected by the new regulation?

The transition to T+1 represents a significant shift in the post-trade operating model. It is not simply a matter of shortening the settlement cycle; it requires a fundamental rethink of the entire value chain.

The areas most affected are matching and confirmation, which must now be completed on the trade date (T), settlement, which requires more dynamic management of instructions, fails and partial settlements, and liquidity management, which becomes intraday and highly time-critical.

In addition, there is a growing need for integration with corporate actions, which are becoming increasingly interdependent with the settlement cycle. Interactions among these processes are now intraday and tightly interconnected: errors or delays upstream are immediately propagated across the entire processing chain.

How do you support clients in managing liquidity and the broader operational impacts?

Our approach is based on the understanding that T+1 is not simply about accelerating processes—it is about managing greater complexity.

Solutions must provide an integrated view of liquidity across both central bank and commercial bank money, granular monitoring of market- and counterparty-specific cut-off times, and intraday forecasting capabilities. These are complemented by auto-funding functionalities, automated matching and proactive exception management.

Many of these capabilities already exist, but they must be redesigned to operate in real time and in a coordinated manner. From an architectural perspective, we are evolving towards event-driven and streaming-based models that enable real-time data processing and move beyond traditional batch-oriented approaches.

At the core of this transformation is the introduction of end-to-end orchestration mechanisms, enabling continuous monitoring, intelligent exception handling and intervention prioritization. The objective is to make the post-trade value chain more seamless, integrated and responsive—not only to monitor operations, but also to support timely and effective decision-making.

What challenges are you encountering? Are banks ready?

The main challenges are not purely technological; they are also operational and organizational.

From a technology standpoint, system fragmentation remains a major issue. Current architectures were designed around separate domains and sequential processes, whereas T+1 requires real-time coordination across highly interdependent workflows spanning multiple business areas. Coordination does not necessarily mean that all processes must be managed through a single platform; rather, organizations need tools that provide a unified and integrated operational view.

A second challenge is exception management. What was traditionally handled overnight must now be addressed intraday, requiring higher levels of automation, alerting capabilities and robust Data Quality Management from the earliest stages of trade capture. This includes bringing data enrichment activities forward into the matching phase—for example, the early identification of the Place of Settlement.

Finally, there is the organizational dimension. T+1 demands closer collaboration between business and IT teams to design intelligent, automated processes, while also redefining the boundaries between middle-office and back-office functions.

The real question is not whether banks are ready, but whether their operating models are designed to function in real time. In many cases, the answer is still no. However, the entire ecosystem is moving rapidly in this direction.

In this context, we believe market participants should embrace the principle of “zero additional people”, leveraging automation, Straight-Through Processing (STP) and rule- and AI-driven decision-making models.

Institutions that can effectively orchestrate matching, settlement, corporate actions and liquidity processes will be able to turn T+1 into a competitive advantage, improving productivity while reducing the overall cost to serve.

English version by TAS

Original article from Bancaforte