AGEFI Luxembourg - septembre 2025

AGEFI Luxembourg 46 Septembre 2025 Informatique financière By Pascal HERNALSTEEN, Strategic consultant, GoJuSan A sset servicing is being reshaped by forces that are pulling the industry into uncharted territory. Retail al- ternatives are opening private markets to thousands of new investors. Evergreen funds are blur- ring the line between pri- vate and public capital, requiringcontinuousflows ratherthanpredictablecycles. Regulators are pushing for real-time reporting, while clients expect interactive dashboards rather than static PDFs delivered weeks after the fact. GP’s and their LP’s demand excellence from their service providers intermsofaccuracyandtimeliness.Clientchurnrates are rising, drivenmostly by quality issues that erode trust.Andatthesametime,clientsremainfocusedon price, looking for themost cost-effective service. This dual pressure with higher expectations for quality on one side, relentless cost discipline on the other, defines the industry’s dilemma . Traditional operating models are straining to cope. The complexity has only increased, with fragmented systems and production distributed across multiple onshore and offshore sites that demand constant co- ordination.Thisoperatingmodelhasnowreachedits limits. Technology is no longer a ‘nice to have’; it is the onlypath forward. FromGenerativeAI toAgenticAI GenerativeAI has already demonstrated its value. It canreadcontracts,extractpatternsfromunstructured documents, and draft reports in minutes instead of days. In many cases, it has improved both accuracy andtimelinesswhileloweringthecostofroutinetasks. But its deployment has often been piecemeal. GenAI accelerateswork;itmakesindividualprocessesfaster and cleaner, yet it does not, by itself, reinvent how work is organizedor delivered. AgenticAI executes These are systems capable of operating with a high degree of autonomy: they “understand” goals, plan the steps, and carry them out while adapting to changing conditions. They no longer wait for inputs; they act proactively. Insteadof hu- mansstitchingtogetherworkflows,softwareagents coordinatewitheachotherto deliveroutcomesend- to-end .Humans remain in the loop, but as super- visors and decision-makers for the exceptional cases.WithAgenticAI, the solutionowns the process and the outcome, in real time, not asaquarterlycatch-upexercise.Embedded intelligence replaces fragmentedmanual tasks.Inshort: GenAIassists,butAgentic AI executes. The First Battleground: Onboarding,AML, andKYC Few processes illustrate the stakes more clearly than investoronboarding.Itisboththeentrypointforbusi- nessandthefrontlineofcompliance.Gettingitwrong has consequences on both sides of the ledger.A false positive, where a legitimate investor is wrongly re- jected,frustratesmanagersandresultsinlostbusiness. Afalsenegative,whereariskyinvestorslipsthrough, isevenworse,carryingthepotentialforsanctions,rep- utational damage, or exposure tomoney laundering and terrorist financing. Tominimizetheserisks,mostfirmsrelyonexhaustive manual reviews. Compliance teams spend days, sometimes weeks, checking documents, cross-refer- encingdatabases,andwritingreports.Thelongerthis takes,thehighertheriskoflosinggoodinvestorswho simplyrefusetowait.Retailalternativefundsmagnify the challenge. Instead of a handful of institutional in- vestors,administratorsmustnowonboardhundreds, even thousands, of individuals in a continuous stream. Manual processes cannot scale to that level. AgenticAI changes the equation. Specializedagentscanextractandstructuredatafrom passports and contracts, check sanctions and politi- cally exposed person (PEP) lists in real time, analyze adverse media, and validate results through critical agents,monitorexpirationdatesofidentificationdoc- uments, and generate follow-up emails. Humans in- tervene only for edge cases. The result is faster onboarding, fewer false positives, fewer false nega- tives, anddramatically lower cost per investor. DataGovernance: TheQuiet Force Datagovernance is the silentweaknessundermining everything else. Asset servicers are drowning in un- structured inputs: Excel files, PDFs, andemails arriv- ing from multiple parties. The same information is sometimes duplicated across several systems, incon- sistent and incomplete, feeding downstream errors and reconciliation nightmares. Some providers have automated fragments of the process through OCR tools and workflow solutions, but the result some- times add more complexity to the global setup. A piecemeal approach cannot resolve the fundamental problem: without trusted data, every report and everydecision sits on shaky ground . AgenticAIprovidesasystemicsolution.Datapipeline agents ingest and normalize information from dis- parate sources. Validation agents detect anomalies, rerun workflows when necessary, and escalate only whenhumanoversight is required. Research agents enrich data with external feeds, en- suring it is complete andup todate. Insteadof skilled professionals wasting 90& of their time on collection and cleaning, they can focus on the 10%that requires judgment and analysis. The more these agents process,themoretheylearnandthemoretheyreduce thepercentageoftasksrequiringhumanintervention. The payoff is felt across the chain: more accurate re- ports, faster delivery, fewer restatements, and higher client satisfactionwhile reducing structural cost. Yesterday’s FutureMay WellAlreadyBeObsolete Thelessonforassetservicersisclear:technologyadop- tioncannot be reduced toa series of isolatedautoma- tion projects. The industry does not need another patchworkofquickfixes.Whatitneedsisa client-cen- tered methodology that begins with the end in mind: delivering accuracy, timeliness, and cost effi- ciency simultaneously . Themethodology is straightforward Start by mapping every process in the client value chain.Foreach,assignthreeratings:thefrequencyand impact of errors (accuracy), the consequences of de- lays (timeliness), and the share of total production costs(cost).Theseratingsmusttranslateintomeasur- able objectives so that progress can be tracked over time,notonlyintermsofcostsavingsbutalsointerms of service quality anddelivery speed. The next step is to look beyond individual fixes. Processes rarely exist in isolation; they interact with one another in ways that magnify weaknesses. A delay or error inone area often cascades andamplify across the chain. That is why leaders must think in terms of whole systems, not fragments, assessing in- terdependenciesanddesigningsolutionsthataddress thebroader operatingmodel rather thanone taskat a time. Once priorities are set, each process should be evaluated against different options. Offshoring, partial automation, a complete redesign, or even leaving it unchanged, and the impact of each choice measured against the three axes of accuracy, timeliness, and cost. Crucially, progress must be tracked against the objectives defined at the outset , ensuring that transformation is anchored in results rather than rhetoric. Thisapproachshiftstheconversationawayfromtech- nology for its own sake. It ensures that AgenticAI is deployed not because it is fashionable, but because it directly improves the client experience, strengthens theprovider’soperatingmodel,andenhancesitscom- petitive edge and right to win in an increasingly crowdedmarket. Theurgencyisheightenedbythefactthatnewplayers areenteringthefield.Unlikeincumbents,theyarenot weigheddownbydecadesofcomplexlegacysystems and entrenched operational setups. They can start from an empty page, designing processes natively arounddata, automation, and intelligence. For them, AgenticAIisnotatransformation,itisthefoundation. Incumbents, bycontrast, face theharder challengeof reinvention. They must unpick outdated processes, rethinkworkflows,andovercomeorganizationalin- ertia. It is more complicated, but it is also unavoid- able. Because in asset servicing, the future will not wait for thosewho hesitate. Yesterday’s futuremay already be obsolete and the only right to win be- longstothosewhoreinventthemselvesnow…and in the future . Yesterday’s future belongs to the past Opinion - By Prof. Dr. Bruno COLMANT, Member of the RoyalAcademy of Belgium I amnot an entrepreneur, for as Paul Valery (1871-1945) said, in- tellectuals struggle to accom- modate the disorder of events and do not possess the audacity that tran- scends their careful equations and thoughts. However, I would like to share some convictions re- garding the upheaval of arti- ficial intelligence and how wemust approach it. The in- spiration for this text stems froma bookmymother gave me upon graduating fromSolvay, more than forty years ago, which posited that the quality of executives lay in their general knowledge. I didn't read the book; the back cover was enough for me, but I've drawn some striking lessons from it that I'd like to share. Artificial intelligence is no longer a distant whisper, a futuristic hypothesis, or a topic for academic con- ferences relegated to the margins of our immediate concerns. No. The stark andundeniable truth is that we are on the threshold, indeed already at the heart, of an anthropological and thus economic transfor- mationof unprecedentedscale. Forwhat is looming, and is already unfolding before our eyes, is nothing less than a twilight of employment as we have knownit,aradicalredefinitionofthevalueofhuman labor, precipitated by the inevitable advent of the quantumbreakthrough. We are no longer speaking of an industrial revolu- tion replacing physical force with machines, nor of automationrelegatingrepetitivetasks.Artificialintel- ligence, in its current and future versions, is attack- ing, not without a certain irony of fate, the very attributes we thought were the inviolable sanctuary of our intelligence. Expert professions—whether legal, medical, financial, or consulting—those based onknowledgeaccumulation, complexdataanalysis, andpatternrecognition,arealready,notmerely"will be," disrupted. Intuition, erudition, the teaching capacity based on factual transmission: all of this is nowwithin reach, andoften surpassed in speed and volume, by neural architectures capable of ingesting and synthesizing colossal data cor- puses in a fraction of a second. The machine no longer merely calculates: it reasons, learns, generates, creates, and diag- noses. The Cartesian "cogito, ergo sum" seems to metamorphose into a "computo, ergo sum" for the algorithm, leaving us in a certain existential disarray. André Malraux (1901-1976) said that accurate intelli- gence is the deconstruction of the comedy. And so it is now that we must be intelligent and lucid in light of what we observe. KarlMarx (1818-1883) hadperfectlygrasped this evolution with a staggering prophecy. In Das Kapital, hehadwritten that theworker's skillwould become meager before prodigious science. And in "A General Introduction to the Critique of Political Economy," he anticipated the advent of a capitalism where the productive force would no longer be labor, but cognition, corresponding to contempo- rary digitization and robotization. It is here that the sublime paradox of artificial intelligence resides. By stripping us of our replicable tasks, of our compil- able knowledge, it returns us, with a radical imper- ative, to the very essence of our humanity. What, then, is to be done in the face of this rising tide that threatens to submerge the foundations of our employment system? The answer does not lie in a frantic race for techno- logicalmastery—for themachinewill always exceed man in computational power—but in a reclaiming ofwhatmakesusintrinsicallyirreplaceable.Wemust invest, first and foremost, in ourselves. Investing in oneself begins bypursuing studies to structure one's thinking. We must not listen to those who advocate abandoningeducation;onthecontrary,wemust.We must therefore learn, and better yet, learn how to learn in the logicofCicero, ceaselesslyandat all ages. The proliferation of information media allows us to inform ourselves and develop ourselves more than at any other time in human history. Investing in oneself—that is the categorical impera- tive of our time. But what does that mean, beyond the generic phrase? It is a multidimensional, pro- found, almost Socratic approach: Firstly, the constant exercise of critical and synthesiz- ing intelligence. This is not about memorization or proceduralintelligence,whichmachinesexcelatmim- icking and surpassing, but about the intelligence that discerns, contextualizes, questions, and formulates ethicalandstrategicjudgments.Itistheabilitytomake connections between seemingly disparate domains, tonavigateuncertainty,andtotransforminformation intowisdom. It is the spirit of discernment,where the machine excels in the spirit of geometry. It means developing our lateral thinking, our capacity for abstraction, and our propensity for solving complex problems that require muchmore than simple algo- rithmic optimization. It is the human logos in its noblest and most demanding sense, the ability not only to process data but to extract meaning, signifi- cance, andpurpose fromit. Secondly, immersion in general knowledge. General knowledge is not an academic luxury or a social adornment: it is the fertile ground for adaptability. It givesmeaning to facts, depth to events, andperspec- tive toupheavals. Thirdly, the refinement of our human predisposi- tions.What are thesegifts?Theyreside ineverything that the machine, however sophisticated, cannot truly emulate or replace.Aptness: it is the art of con- textual relevance, subtlenuance, humanresponsive- nessinthefaceoftheunforeseen.Itisknowingwhen to say what, how to say it, and to whom. It is situa- tionalintelligence,tact,anddiplomacy.Itistheability to sense the zeitgeist, to adjust one's discourse or action at the precisemoment, with an acuity that no algorithm can truly grasp. Trust and human rela- tionships: algorithms can simulate a conversation, but they cannot inspire trust, that rare and precious commodity. It is the sparkof ingenuity, the audacity of theartist, thevisionof the strategistwho reinvents the rules of the game. It is the ability to err magnifi- cently to find an unexpected path, an impulse that data alone cannot predict. Finally, fourthly, continuous learning and perpetual reinvention. We are no longer in a world where one learns aprofession for life. Skill obsolescence is rapid. To learn is to adapt. But this trainingmust transcend the acquisition of mere "hard skills" to focus on "soft skills," which are in reality "human skills." It is a con- tinuous process of self-assessment and self-improve- ment, a permanent quest for our intrinsic "gifts," our uniquepredispositions inexchange andcontribution to the community. It is acceptable to become a "per- petuallearner,"notoutofconstraint,butoutofneces- sity and a desire for fulfillment. It is understood that ourgreateststrengthliesinourintellectualmalleability andour inextinguishable curiosity. This period, while bringing its share of disarray and potential hardship for thosewhowill not knowhow or be unable to adapt, also holds an extraordinary, almost messianic, opportunity. Artificial intelligence, by relievingus of low-value-added tasks, offersus an unprecedentedopportunitytorethinkourprofession- al lives, andbeyond that, our lives ingeneral. It forces us to take stock, toexamineour conscience regarding what constitutes our uniqueness and our profound utility. It pushes us towards a return to the essential, to the rediscovery of our human capital in its noblest andmost profound sense. It is not a revolution prefiguring a utopian "great evening" that awaits us, but rather the twilight of a particularconceptionofwork,openingontothedawn ofanewera.Anerawherevaluewillnolongerreside incomputationalpowerorthecapacityfordataaccu- mulation, but in the subtletyof discernment, the rich- ness of general knowledge, the strength of human relationships,thedepthofempathy,andtheaudacity of creativity. Thechallengeisimmense,thetransformationradical. But humanity, throughout its millennial history, has always known how to adapt, to reinvent itself in the face of its upheavals. To serenely confront the onslaughts of this new world, to transform twilight into dawn, the most profitable, the safest, the most essential investment is the one we will make in our- selves, in that inalienable part of our intelligence and ourhumanity. It is there, andonly there, that our true comparative advantage lies. Time is pressing, the requirement is absolute, but the promise, for those who dare the path of introspection and effort, is that of a newhumandignity. And Iwill conclude this textwithaquote that should inspire us all. It emanates from Charles de Gaulle (1890-1970): difficulty attracts the man of character, for it is by embracing it that he realizes himself. And François Mauriac (1885-1970) responded, echoing him: a life isworthonlywhat it has cost in effort. Artificial Intelligence: the imperative to invest in oneself