Business Intelligence & AI

Patrick Rotzetter
17 Nov 202353:32

Summary

TLDRLe script présente une nouvelle méthode d'analyse des entreprises, développée par Alis, qui utilise l'intelligence artificielle neuro-symbolique pour la prise de décision. Après cinq années de recherche et développement, cette approche permet de détecter des risques et des opportunités de manière automatisée, en identifiant des signaux d'alerte précurseurs. L'IA spot sept fois plus d'insights que l'exploration humaine, aidant ainsi les analystes à comprendre les causes profondes des problèmes et à élaborer des plans d'action correctifs. L'objectif principal est de repérer des opportunités même dans les unités commerciales en difficulté et des points d'amélioration chez les unités réussies. La solution d'Alis vise à résoudre les problèmes des outils de reporting actuels, en offrant une vision consolidée des données et en mettant en avant l'importance de l'analyse déconsolidée et de la dynamique des données pour les entreprises.

Takeaways

  • 🚀 Une nouvelle méthode d'analyse des entreprises est présentée, basée sur une intelligence artificielle neuro-symbolique pour la prise de décision.
  • 🔍 L'intelligence artificielle permet d'automatiser la détection des risques et des opportunités, ainsi que la découverte de signaux d'alerte précurseurs.
  • 🌟 L'objectif principal est de repérer des opportunités même dans les unités commerciales en difficulté et des points d'amélioration dans les unités réussies.
  • 📊 Les outils actuels de reporting sont critiqués pour ne pas offrir d'analyse multidimensionnelle ou de déconsolidation des données.
  • 🤖 L'intelligence artificielle générative est combinée avec l'IA symbolique et les graphes de connaissances (KGE) pour identifier la cause profonde des problèmes.
  • 🔑 L'algorithme de priorisation Mark of Shame est utilisé pour classer les problèmes business en fonction de leur impact et d'autres critères.
  • 🧠 Une solution basée sur l'IA est proposée pour aider les analystes à construire des plans d'action, plutôt que de simplement générer des rapports.
  • 📈 Des trackers spécialisés sont utilisés pour analyser les données financières d'entreprise de manière systématique et multidimensionnelle.
  • 🗣️ Une interface de speech-to-text est envisagée pour simplifier l'interaction avec l'application et aider à identifier les problèmes non résolus dans les rapports.
  • 🔄 L'IA est utilisée pour upskill les analystes et les aider à améliorer leur expertise en matière d'analyse financière.
  • 🎯 L'objectif de la solution est de devenir un outil d'analyse complémentaire aux méthodes de reporting traditionnelles, offrant une vision plus large et une aide à la prise de décision.

Q & A

  • Quel est le nouveau modèle d'analyse des entreprises présenté dans le script ?

    -Le nouveau modèle d'analyse des entreprises présenté dans le script est la méthode d'analyse des risques et des opportunités. Elle utilise une intelligence artificielle neuro-symbolique pour la prise de décision, permettant de découvrir automatiquement les signaux d'alerte précoce dans les données de l'entreprise.

  • Quelle est la durée de recherche et développement pour cette nouvelle méthode d'analyse ?

    -La durée de recherche et développement pour cette nouvelle méthode d'analyse est de cinq ans.

  • Quelles sont les antécédents de l'orateur en matière de science des données ?

    -L'orateur a travaillé dans le domaine de la science des données pendant cinq ans avant de rejoindre Alis. Avant cela, il était Directeur Financier (CFO) dans le secteur public en France. Il a également publié beaucoup de contenu lié à la science des données sur LinkedIn et Medium.

  • Comment l'intelligence artificielle peut-elle aider dans l'analyse des performances d'entreprise ?

    -L'intelligence artificielle permet aux organisations d'automatiser la détection et la découverte des risques et des opportunités, ce que l'on appelle les signaux d'alerte précoce. En utilisant cette méthode, on peut identifier sept fois plus d'informations pertinentes que par l'exploration humaine.

  • Quel est le rôle d'un analyste dans le cadre de cette nouvelle méthode d'analyse ?

    -Le rôle d'un analyste est de comprendre la cause fondamentale du problème. Grâce à l'intelligence artificielle, les analystes peuvent automatiquement détecter les anomalies dans les données de l'entreprise, puis trouver la cause racine et bâtir un plan d'action correctif.

  • Quelles sont les principales limitations des outils de reporting actuels ?

    -Les outils de reporting actuels sont principalement basés sur une analyse consolidée des données, ce qui masque les anomalies et les insights pertinents. Ils ne permettent pas une analyse multi-dimensionnelle systématique et ne répondent pas aux questions sur la conversion d'insights en plans d'action.

  • Pourquoi les entreprises ne détectent-elles pas proactiquement les anomalies avec les logiciels de business intelligence actuels ?

    -Les logiciels de business intelligence actuels ne sont pas conçus pour détecter proactiquement les anomalies. Ils laissent aux analystes la responsabilité de faire une analyse multi-dimensionnelle manuelle, ce qui est impossible pour l'homme lorsqu'il s'agit de données à plusieurs dimensions.

  • Quelle est la solution proposée par Alis pour améliorer la détection d'insights ?

    -Alis propose une solution qui utilise de l'intelligence artificielle, combinant l'IA générative et l'IA symbolique sous forme de graphes KGE, pour trouver la cause racine et bâtir un plan d'action correctif. Cette méthode permet de repérer jusqu'à sept fois plus d'insights qu'avec les méthodes traditionnelles de reporting manuel.

  • Comment la solution d'Alis aide-t-elle les entreprises à réagir plus rapidement que leurs concurrents ?

    -La solution d'Alis permet de détecter les anomalies et les risques plus tôt que les concurrents, grâce à une analyse multi-dimensionnelle systématique et automatisée. Elle offre également des insights sur les actions correctives à prendre, permettant aux entreprises de réagir rapidement et de s'adapter à la situation.

  • Quelle est la vision globale de la solution proposée par Alis pour l'analyse des entreprises ?

    -La vision globale de la solution proposée par Alis est de trouver des opportunités même dans les unités d'affaires en difficulté et des points d'amélioration dans les unités prospères. Elle vise à aider les entreprises à répondre aux questions critiques et à suivre les KPIs pertinents grâce à une analyse plus approfondie et à une prise de décision plus éclairée.

  • Quelle est la méthode proposée par Alis pour améliorer la qualité des données et la pertinence des insights ?

    -Alis utilise une approche de traitement par lots pour détecter les insights chaque mois en utilisant des données financières de qualité supérieure. Ils ont également développé un algorithme de priorisation pour gérer le flot d'alertes et pour identifier les insights les plus importants en fonction de la stratégie de l'entreprise et de la situation urgente.

Outlines

00:00

📝 Présentation de la nouvelle méthode d'analyse d'entreprise

Anthony, le responsable des opérations d'Alis, présente une nouvelle méthode d'analyse d'entreprise basée sur l'intelligence artificielle. Après cinq années de recherche et de développement, cette méthode, appelée "reporting des risques et des opportunités", permet de détecter automatiquement les risques et les opportunités, ou les signes d'alerte précoce, dans les données d'une entreprise. Cette approche permet de découvrir sept fois plus d'informations que l'exploration humaine. Le rôle de l'analyste est alors de comprendre la cause profonde des problèmes. L'intelligence artificielle, en utilisant des techniques génératives et symboliques, aide à identifier la cause racine et à élaborer un plan d'action correctif.

05:00

🔍 Problèmes avec les outils de reporting actuels

Le script souligne les problèmes avec les outils de reporting actuels, qui sont principalement basés sur une analyse consolidée de données. Cela masque les anomalies et les insights importants qui ne sont pas visibles dans l'analyse consolidée. Les outils actuels ne permettent pas une analyse multidimensionnelle systématique, ce qui est essentiel pour identifier les problèmes. Les outils de reporting ne fournissent pas non plus de KPIs sur le nombre d'insights identifiés et convertis en plans d'action. L'intelligence artificielle peut jouer un rôle crucial pour résoudre ces problèmes en détectant proactivement les anomalies et en aidant à répondre aux questions critiques.

10:00

🤖 Solution Alis pour la détection d'insights

Alis propose une solution qui utilise l'intelligence artificielle pour une analyse multidimensionnelle des données. Leur outil peut identifier jusqu'à 85% de plus d'insights que les méthodes traditionnelles de reporting manuel. Cependant, l'afflux d'insights peut être overwhelming, nécessitant une algorithme de priorisation pour aider les utilisateurs à identifier les insights les plus importants. Alis a développé un algorithme de priorisation qui prend en compte la stratégie de l'entreprise, l'urgence et la faisabilité pour aider les utilisateurs à agir efficacement.

15:03

💡 Amélioration du rôle de l'analyste avec l'IA

L'IA permet de libérer les analystes de la tâche de générer des rapports et de les aider à trouver des informations pertinentes. Les analystes peuvent ainsi se concentrer sur la résolution des problèmes et l'élaboration de plans d'action. Alis utilise une combinaison d'IA générative et d'IA symbolique pour aider les analystes à comprendre les causes profondes des problèmes et à construire des plans d'action. L'IA sert à recommander des actions et à poser des questions, aidant ainsi les analystes à améliorer leur expertise et à prendre des décisions plus éclairées.

20:04

🌐 Vision de l'IA dans l'analyse d'entreprise

La présentation d'Anthony met en évidence la vision d'Alis sur l'IA dans l'analyse d'entreprise. Il souligne que l'IA peut aider à détecter des problèmes de manière systématique et à recommander des actions, mais que les humains sont mieux placés pour prendre des décisions. L'IA peut également aider à améliorer l'expertise des analystes, en particulier dans les entreprises où il y a un taux élevé de rotation. Le but d'Alis est de créer une application simple et facile à utiliser qui automatise la génération de rapports et permet aux analystes de se concentrer sur la résolution des problèmes.

Mindmap

Keywords

💡Intelligence artificielle neuro-symbolique

L'intelligence artificielle neuro-symbolique combine l'apprentissage profond (neuro) et l'intelligence symbolique pour traiter et interpréter des données complexes. Dans le script, cette méthode est utilisée pour automatiser la détection des risques et des opportunités, améliorant ainsi l'analyse de performance des entreprises. L'approche permet de découvrir des signaux précurseurs (early warning signs) avec une efficacité sept fois supérieure à l'exploration humaine, illustrant son importance dans la prise de décision basée sur les données.

💡Détection d'anomalies

La détection d'anomalies est le processus d'identification des écarts significatifs dans les données d'une entreprise, qui peuvent indiquer des problèmes ou des opportunités cachées. Le script met en évidence l'utilisation de l'IA pour automatiser cette détection, permettant ainsi une identification plus rapide et plus précise des anomalies dans les entrepôts de données des entreprises, par rapport à l'exploration manuelle par des humains.

💡Graphes de connaissances (Knowledge Graphs)

Les graphes de connaissances organisent et intègrent des informations sous forme de réseaux d'entités et de leurs relations interdépendantes. Dans le contexte du script, ils sont utilisés pour informer l'IA générative et améliorer la compréhension des causes profondes des anomalies détectées, facilitant la création de plans d'action correctifs.

💡Analyse multidimensionnelle

L'analyse multidimensionnelle permet d'examiner les données à travers plusieurs dimensions ou angles simultanément. Le script souligne son importance dans l'identification des anomalies et des opportunités d'affaires, en mettant en évidence que l'IA peut traiter de manière systématique et automatique plusieurs dimensions, ce qui dépasse les capacités humaines limitées à des analyses tridimensionnelles sur des feuilles de calcul.

💡Priorisation des alertes

La priorisation des alertes est le processus de classement des anomalies détectées selon leur importance ou leur urgence. Le script décrit comment l'utilisation de l'IA permet d'organiser les alertes générées par la détection d'anomalies, en s'assurant que les utilisateurs se concentrent sur les problèmes les plus critiques, en fonction de la stratégie de l'entreprise et d'autres critères pertinents.

💡Consolidation des analyses

La consolidation des analyses fait référence à l'agrégation de données provenant de différentes sources en un aperçu unifié. Le script critique cette approche pour masquer des anomalies importantes au niveau granulaire à cause des phénomènes de compensation. Il plaide pour une approche de déconsolidation permettant une analyse plus fine et précise des données.

💡IA générative

L'IA générative est capable de créer du contenu nouveau et original à partir de données existantes. Dans le script, elle est utilisée conjointement avec des graphes de connaissances pour générer des explications et des plans d'action correctifs, illustrant son rôle dans la compréhension des causes profondes des anomalies et la facilitation de la prise de décision.

💡Détection proactive

La détection proactive implique l'identification des problèmes potentiels avant qu'ils ne deviennent critiques. Le script questionne pourquoi les outils de business intelligence actuels ne détectent pas proactivement les anomalies, mettant en lumière la capacité de l'IA à remplir ce rôle et à anticiper les problèmes, améliorant ainsi la réactivité des entreprises.

💡Rapports de risques et opportunités

Les rapports de risques et opportunités fournissent une analyse ciblée des potentiels défis et avantages pour une entreprise. Le script introduit un nouveau moyen, développé par Alis, d'automatiser ces rapports grâce à l'IA, mettant en évidence l'amélioration significative dans la détection des anomalies et la génération d'insights pour une prise de décision éclairée.

💡Analyse granulaire

L'analyse granulaire fait référence à l'examen détaillé des données au niveau le plus fin. Elle est cruciale pour identifier des insights spécifiques qui peuvent être masqués dans des vues consolidées. Le script met en avant cette approche pour montrer comment l'IA peut révéler des anomalies et des opportunités cachées par des effets de compensation dans des analyses moins détaillées.

Highlights

提出了一种新的商业分析方法,通过多年的研究和开发,专注于风险和机会报告。

使用神经符号人工智能进行决策,通过自动化检测公司数据仓库中的异常,发现风险和机会的早期警告信号。

与人类探索相比,该方法能够发现7倍多的洞察力,转变分析师的角色,更注重理解问题的根本原因。

利用生成型人工智能和符号型人工智能的结合,通过知识图谱(KGE图)来发现问题的根本原因。

该方法不仅适用于问题业务单元寻找机会,也适用于成功业务单元的改进点分析。

提出问题:为什么2023年的商业智能软件还不能主动检测异常?强调人工智能在大规模并行处理中的作用。

指出当前报告工具的问题,如合并分析导致的信息补偿和隐藏,以及缺乏动态技术分析。

强调需要系统性多维分析,以及人工智能在自动执行此类分析中的必要性。

讨论了当前报告世界的状况,指出传统报告工具只能提供全局视角,而无法发现被隐藏的洞察力。

提出了解决方案,使用机器学习和监督学习进行异常检测,并结合生成型人工智能和知识图谱。

解释了如何通过人工智能帮助用户构建行动计划,而不是完全自动化决策过程。

讨论了数据质量问题对报告的影响,以及如何通过工具解决这些问题。

强调了人工智能在增强分析师技能和专业知识方面的应用,特别是在人员流动率高的领域。

提出了一种新的人工智能应用方法,即通过生成型人工智能辅助分析师提出问题和建议行动。

讨论了人工智能在决策支持中的角色,强调人类在最终决策中的重要性。

提出了一种新的报告工具,旨在通过人工智能自动化和系统化地提供风险和机会的早期警告。

强调了人工智能在帮助用户理解和应对复杂业务问题中的潜力。

讨论了如何将人工智能与业务分析和项目管理相结合,以提高决策效率。

Transcripts

play00:00

started

play00:02

recing hey Anthony welcome welcome to

play00:05

this event and happy you could make it

play00:08

um thank you so um I'll let you start

play00:13

okay so today I would like to talk about

play00:16

um a new method uh to do business

play00:21

analysis uh at Alis we have been working

play00:24

on it uh for many years five years of

play00:27

research and development and uh it is a

play00:30

new way to do um we call it risk and

play00:34

opportunities

play00:36

reporting uh to to talk more about me so

play00:42

I am a chief IE officer at

play00:44

Alis uh I have been uh doing data

play00:48

science for five years uh before that I

play00:52

was a CFO so Chief Financial Officer in

play00:55

the public sector in France and I have

play00:58

been uh uh publishing many data science

play01:02

related content over LinkedIn and medium

play01:06

and uh since then I it has been two

play01:09

years that I I have joined Alis on this

play01:13

project on this

play01:15

adventure uh so uh what we are

play01:20

using we are using neuros symbolic

play01:23

artificial intelligence for decision

play01:25

making uh what we have understood

play01:30

is

play01:31

that uh for for to do a business

play01:35

performance analysis the artificial

play01:38

intelligence uh allow an organization to

play01:41

automate uh the uh detection the

play01:48

discovering of risk and opportunities

play01:51

what we call early warning signs and by

play01:55

doing it with method that we use we spot

play01:59

seven s times more insights than by

play02:02

human exploration so the question is no

play02:04

longer where to look

play02:07

at it

play02:09

is the why question the analyst and the

play02:13

artificial intelligence sport Insight

play02:16

needs to understand the root cause of

play02:18

the problem so it is a new role for them

play02:21

because because we use artificial

play02:23

intelligence to automatically detect

play02:26

anomalies within uh the companies data

play02:30

warehouse uh the the new role for the

play02:33

analyst is to find the root cause and

play02:35

for that what we are

play02:38

leveraging are artificial intelligence

play02:41

in form of generative EI combined with

play02:44

symbolic EI in form of kge graph uh so

play02:49

it is it is to find the root cause and

play02:51

on you you you spot the root cause uh

play02:55

all method allow also to help the

play02:59

analyst uh build a corrective action

play03:02

plan so it is the old question or to

play03:05

react quicker than your competitors and

play03:07

to to to to act on those uh root cause

play03:12

uh that uh our method allow to to detect

play03:17

so our

play03:19

approach its main goal is to find

play03:24

Opportunities even in a trouble business

play03:27

unit okay and

play03:30

Improvement points in success successful

play03:34

ones uh because even your best uh

play03:37

business units uh have Improvement

play03:40

points and I will show how how we are

play03:43

doing

play03:45

that okay what I ask our clients is

play03:51

this uh why in

play03:54

2023 no business intelligence software

play03:57

is able to proactively detect

play04:00

anomalies why is it is it the the end

play04:04

user the analyst

play04:08

to

play04:09

manually uh that he has to manually uh

play04:13

do a

play04:14

multi-dimensional to do an an

play04:17

analysis why we don't use artificial

play04:20

intelligence with massively uh parallel

play04:23

processing to detect proactively

play04:27

anomalies this is the first question and

play04:30

usually what I ask also to clients is

play04:33

this from your reporting uh we can take

play04:37

for example September do you know how

play04:39

many insights were identified and

play04:42

generally the answer is no because this

play04:45

is uh a kpi it is not a kpi that is uh

play04:49

um analyzed by

play04:51

companies and even imagine if we we we

play04:56

we we we respond to this question yes

play05:00

uh how many of those insights did you

play05:02

convert into corrective action plans and

play05:06

we can go even further uh from those

play05:09

corrective action action plans what is

play05:12

the trend curve of Correction rate of

play05:14

insight into action plans those those

play05:18

question should be the uh kpi of any

play05:23

Reporting System yet currently in many

play05:27

companies they are not uh answered they

play05:30

are not followed and it should be the

play05:33

the main goal of any reporting tool

play05:37

currently and we think at Al thisis that

play05:40

artificial intelligence uh is a parament

play05:43

is essential to answer those question

play05:47

and help a business answer those

play05:50

question and follow these uh main C

play05:53

Capi so what is the main problem of

play05:57

current reporting tools

play06:00

the main issue is

play06:03

Consolidated

play06:06

analysis all current

play06:08

reporting are based are built on

play06:13

Consolidated analysis of

play06:16

data why is it a problem I will show you

play06:20

a simple example right here we have two

play06:25

continents okay uh what you can see at

play06:29

the solidated view is that continent two

play06:32

it's in a worse situation than continent

play06:36

one but but if you DEC consolidate okay

play06:40

what you can see is that for continent

play06:42

one that seem better off there is an

play06:45

anomaly that is compensated by other

play06:50

countries for this continent this is the

play06:53

problem of consolidation because

play06:56

consolidation you have multiple uh

play06:58

mathematic IAL

play07:00

phenomenon like Simpson Paradox

play07:06

like mix when you mix there is a

play07:10

compensation between negative value and

play07:12

positive value but even in positive

play07:15

value you have a compensation uh if you

play07:17

have an extreme positive value and other

play07:20

value are smaller in comparison this is

play07:24

Extreme positive value will be hidden

play07:26

will be masked by Consolidated uh data

play07:29

by Consolidated value value this is this

play07:33

this seem a simple problem but it is not

play07:36

because right now every um reporting

play07:40

tool uh um uh consolidates data and

play07:44

don't have a a comprehensive and

play07:46

systematic way to

play07:49

do uh a deconsolidated

play07:54

analysis and it is also true in time de

play07:58

in time dimension

play08:00

right here I am showing you an example

play08:03

where um I am analyzing the current year

play08:08

and I compared it to the previous year

play08:11

so I am in 23 and I am comparing 22 I am

play08:16

studying what we call in b b business

play08:19

finance the year to date so I am

play08:21

analyzing the

play08:23

situation uh we are in October so until

play08:27

until October and I and I comparing the

play08:29

situation from the pre previous year so

play08:32

right here in my reporting what I

play08:34

observe is that I am better off in

play08:37

203 comparing to 22 but this

play08:41

Consolidated time analysis uh behind

play08:45

that you can have multiple

play08:49

situation

play08:51

okay it might be that the situ the

play08:54

current situation it explained by an

play08:56

anomaly a positive anomaly uh uh right

play09:00

here you can see an outlier a quarter

play09:03

outlier that is explaining the current

play09:05

better off situation of my current here

play09:07

to date but it can also be that uh my my

play09:12

my positive trend uh is degrading is is

play09:16

decreasing okay but you can Al also have

play09:19

a a tendency reversal so you wear better

play09:23

off you were improving during uh until

play09:27

the the Third quarter and after you are

play09:31

degrading those multiple and I I only

play09:33

took a three example but there is many

play09:36

many configuration uh all those example

play09:39

show that you need to

play09:41

deconsolidate even time analysis so you

play09:45

need to add Dynamic technical analysis

play09:48

within your reporting to analyze the

play09:51

trends uh and the dynamic of your data

play09:55

so and another issue that is uh I am

play10:00

stacking stacking up problems so for you

play10:03

to understand what is the current issue

play10:05

of

play10:07

reporting another issue is that to have

play10:10

a an Insight detection solution you need

play10:14

to do a systematic multi-dimensional

play10:18

analysis okay I will show you an example

play10:22

right here I am taking the client uh

play10:26

dimension in line I have product

play10:30

but every of those line are

play10:33

Consolidated okay so here the situation

play10:36

appear appear to be good okay all all

play10:39

are positive all is fine okay I am

play10:43

taking the now the product dimensions in

play10:46

every line of my client dimension for

play10:48

the product Dimension all are

play10:50

Consolidated and

play10:52

positive but what

play10:54

happen if I combine the two with those

play10:57

same data

play11:01

here with only two Dimension it is it is

play11:05

only to illustrate the necessity of

play11:07

multi-dimensional analysis with only two

play11:10

Dimension I can see I can spot problems

play11:13

and now imagine imagine to do that with

play11:16

a four four five six seven eighth

play11:21

Dimension uh right now every most of

play11:25

multinational companies big company have

play11:28

built uh olap cubes a multi-dimensional

play11:31

data warehouse so they are able to do

play11:35

those multi-dimensional analysis but it

play11:38

is impossible for human Okay human can

play11:41

can only

play11:43

do

play11:44

uh up to three dimension on a spre

play11:49

spreadsheet spreadsheet so so you need

play11:52

artificial intelligence to do

play11:55

this in a systematic and automatic

play12:00

manner

play12:02

yet not many many companies not many

play12:05

solution propose that okay and yet yet

play12:10

it is necessary to sport

play12:13

insights okay right now what is the

play12:16

situation of reporting of the reporting

play12:19

world what current reporting tool can

play12:22

see is this it is a global vision a

play12:25

reporting a Consolidated View Vision

play12:27

what we call a top

play12:30

approach so what they spot is what we

play12:32

call flush head flush head are rotten

play12:37

Branch they are problem that are so

play12:42

important that they are coming

play12:47

up the they are affecting the

play12:50

Consolidated

play12:52

view okay but when you spot them as a

play12:55

Consolidated level it is too late

play12:58

because what you you should have done is

play13:00

to spot

play13:02

them when they first appear before they

play13:06

they were so big that they were

play13:07

affecting the Consolidated View and what

play13:12

those reporting doing Consolidated

play13:15

analysis are not showing you what they

play13:18

can't show

play13:20

you are pocket of insights and golden

play13:24

nuggets what we call them it is meth

play13:27

methaphor metaphor but those are insight

play13:31

that are impossible to spot at a

play13:33

Consolidated view why because there is

play13:36

there is compens compensation effect

play13:40

they are they are concealed within your

play13:43

reporting you can't spot them as a

play13:45

Consolidated

play13:47

View and be because because you don't

play13:50

look at the at the most granular optimal

play13:53

level of data analysis you can't spot

play13:55

those insights and what we found with

play13:58

our solution with with our experimental

play14:00

uh uh work is that

play14:03

85% of insights are currently not found

play14:08

by traditional mean of reporting by

play14:10

manual mean of

play14:12

reporting when you use artificial

play14:14

intelligence to do that in a systematic

play14:16

way you find up to seven time more

play14:20

insights more anomalies within uh for

play14:23

example a business reporting uh for a

play14:27

large corporation

play14:29

so it is like we we are comparing it to

play14:32

a Iceberg of urance I don't know if you

play14:34

know the the management auor that told

play14:36

about that but it is like uh you are

play14:40

uring most of the problem that you have

play14:45

okay

play14:48

so how can you tell that the approach

play14:52

that we are proposing and and I think we

play14:54

are not the only one to propose um to uh

play14:58

to propose a solution of inside

play15:03

detection but how can you tell that our

play15:06

solution is better than others

play15:09

okay uh there there are three

play15:14

criteria like I said I think that you

play15:16

need a multi-dimensional analysis

play15:18

because no not no human and and no human

play15:23

are able to do it and you need

play15:24

artificial intelligence to do that in a

play15:26

systematic way because you need a

play15:29

multi-dimensional analysis you need to

play15:31

compare product geography a time uh

play15:35

region uh Etc to spot anomalies uh to to

play15:41

detect early warning sites to act on

play15:43

them quick and and and quicker than your

play15:48

competition but what happen when you do

play15:51

that is that you are overflowed by alert

play15:55

okay because you are sporting seven time

play15:58

more Insight the user will be

play16:01

overflowing by alerts and anomalies and

play16:04

what you need to

play16:07

introduce is some sort of priorization

play16:10

so you need to build up a priorization

play16:14

algorithm what what what what will be

play16:17

within this priorization

play16:19

algorithm uh are for example the

play16:22

strategy of the

play16:24

company uh some notion of urgency some

play16:28

notion of um feasability

play16:32

some notion of whatever we built we

play16:37

built with an University uh a

play16:40

priorization algorithm that we can

play16:41

personalize to the company Focus or

play16:44

strategy because you need to tell the

play16:48

user with this uh what are insight that

play16:51

are uh I high priority level what are

play16:56

the most impactful insights

play16:59

and what you do what you need also to do

play17:03

you need to to find the most optimal

play17:07

granular level of analysis and it is uh

play17:11

more simple to say it than to find it so

play17:14

you need a

play17:17

method you need to find the right level

play17:19

of analysis that you can

play17:22

compare across time okay because if you

play17:26

go too low to too um too below too below

play17:30

certain

play17:31

fesal you won't be able to compare it

play17:34

over time so you need to find a perfect

play17:38

granular level to do your

play17:43

analysis and artificial intelligence can

play17:47

help also with

play17:51

that

play17:54

so so those what I I just explained

play17:58

those are the problem of current

play18:00

reporting what we have building what we

play18:02

have built at Alesis is the solution to

play18:06

to solve those problem and to um to to

play18:11

give tools to a company to have an early

play18:15

warning side detection

play18:17

tool

play18:19

so this in this uh now I will talk about

play18:24

our data quality

play18:26

approach right now our solution is based

play18:30

on the use case for business finance so

play18:33

we use uh Finance

play18:36

data to detect

play18:39

anomalies so it is a a batch a batch

play18:42

processing uh uh tool right now for to

play18:46

detect Insight every month because

play18:48

Finance data is every month uh and why

play18:52

why we we chose this this this approach

play18:56

it's because Finance data uh generally

play18:59

is the mo is the most uh is um its

play19:03

quality is better uh because there there

play19:06

are lot of um regulation around Finance

play19:10

data because you got the taxpayer it

play19:13

wants it's money you want Auditors

play19:16

Auditors want certain uh certification

play19:20

around run Finance data shareholders

play19:22

also want a good quality of Finance data

play19:24

because they they want to know where to

play19:26

put their money okay

play19:29

so for the what for the first part of

play19:32

detec detecting inside we use Finance

play19:35

data and we found Insight what our tool

play19:39

is

play19:40

proposing is to find the causality of

play19:43

what Insight because Insight are just

play19:46

symptoms okay there will be anomalies

play19:49

within uh data warehouse uh so an

play19:54

anomalies of

play19:55

symptoms is not possibly a problem you

play20:00

need to find uh the causality the why

play20:04

for you to know if you need to act or no

play20:08

so right here we are helping the analyst

play20:12

in in his new role because this is a new

play20:15

role for analyst most analyst in large

play20:18

corporation are only doing the first

play20:21

part the what but the what like I just

play20:24

said we automate it with artificial

play20:26

intelligence so we need the user in in

play20:29

his in his new role and I we think that

play20:33

it it new role is is in more it has more

play20:37

value much more value than the first one

play20:40

the second one is much more complicated

play20:43

and and right here what we use is

play20:46

generative

play20:47

a but not just only generative a we have

play20:52

built rack system so retrial augmented

play20:55

generation that is a system that use use

play20:58

data to help the Genera VI generate its

play21:03

content and we where we get our data

play21:07

it's from operational data so the

play21:10

problem with operational

play21:11

data is that you you you got quality

play21:14

issues some talk about data swamp in

play21:19

many companies because you got e genous

play21:23

data quality inconsistency in accuracy

play21:27

uh you got because if you take also

play21:31

external data you you will got biased

play21:34

one-sided aliden data you are in a VCA

play21:37

World okay VCA it means a complex word

play21:41

in certain uh so generative VII will be

play21:47

here only to to to give you

play21:50

methodologies to give you uh what is

play21:53

currently

play21:54

available it will we are here in using

play21:57

artifici icial intelligence to to build

play21:59

a collective intelligence so it is to

play22:02

help the user build an action plan and

play22:05

not build for them the action plans

play22:07

because here it is too complicated to

play22:09

automate so artificial intelligence is

play22:12

only able to help the user find

play22:14

information and what we like to do is

play22:17

that here we are more in the data

play22:19

science size and here in this part we

play22:22

are more in a decision science s and

play22:25

decision science sites you got a

play22:28

psychology you got uh information that

play22:30

are not present in the data in the in

play22:33

the in the internet in the data way

play22:36

house in instructure data

play22:39

Etc one question are you going deeper

play22:42

into the algorithms so I can ask my

play22:45

question here uh I I will uh in the

play22:49

subse slide I I will go deeper yes so I

play22:52

would be

play22:54

patient I am doing uh in the step by

play22:57

step uh

play22:58

manner okay so what we use so I would

play23:03

like to to show you our stack of

play23:05

artificial

play23:06

intelligence so we use machine learning

play23:09

supervis and nonsupervised uh machine

play23:11

learning uh for the detection sit okay A

play23:15

Time series um we use Al EJ boost uh to

play23:20

predict values from uh

play23:24

the the detection in the detection part

play23:28

of our

play23:30

solution like I said we also use

play23:32

generative VI so what we do but I I will

play23:36

show

play23:37

you right here we combine generate VI

play23:41

with chge graph a chge graph that are

play23:44

built uh on operational data from our

play23:48

clients and we also use chge graph uh to

play23:53

inform uh the generate VI of some sort

play23:56

of methodologies

play23:58

uh for me it is an alternative of

play24:00

fine-tuning i i i i much more better

play24:03

build U an effective sge graph uh that

play24:07

gives uh that can uh when you have an an

play24:13

ontology or operational data than fine

play24:16

tuning a model because fine tune uh you

play24:20

need with time you need to to retrain

play24:23

your model Knowledge Graph is more

play24:25

easier to up toate to to prove over time

play24:28

and what we have also we have uh some

play24:31

parts of reinforcement learning the

play24:33

reinforcement learning part it's more

play24:35

for the super priorization the

play24:37

priorization algorithm that we use and I

play24:40

what I will show I will show the actual

play24:43

app just after this

play24:46

line This is the data science side okay

play24:50

and what we have on

play24:52

top are symbolic artificial

play24:56

intelligence

play24:59

because our prediction from machine

play25:01

learning supervis

play25:05

nonsupervised use in the detection sites

play25:09

a rule engineering expert system so we

play25:12

combined a connectionist with a symbolic

play25:16

artificial intelligence and I will show

play25:18

you just after this

play25:20

slide uh how how we do we we did that so

play25:25

and the second part is decision

play25:29

science we have a natural business

play25:32

analysis explainable intelligence that

play25:34

explain what is happening in orap but in

play25:38

why this insight has been prioritized

play25:41

Etc and we have what I have just said a

play25:44

virtual collective intelligence what we

play25:46

call a process intelligence that helps

play25:48

the user build his action plan so right

play25:52

here we have an escalation a hierarchy

play25:55

of

play25:56

explainability because those part of at

play25:58

the bottom are not explain are not that

play26:02

much

play26:03

explainable and more you you come up

play26:06

this uh line more they are more

play26:09

explainable so we use symbolic EI to uh

play26:13

to have an

play26:15

explainability within our our stack of

play26:18

artificial intelligence and I will show

play26:20

you how how it it happen in the

play26:24

solution so what are what what are the

play26:28

symbolic EI that we use we call them

play26:32

trackers okay trackers are specialized

play26:38

analysis so like I just explained

play26:41

earlier we have the maintain the M of

play26:44

data what we do is that we explod this m

play26:47

to have a a small granular small

play26:50

granular slides of data and for each

play26:54

slides of data we do a systematic

play26:57

multi-dimensional analysis and to do

play27:00

this systematic multi-dimensional an

play27:02

analysis we

play27:04

use some sort of rules because uh you

play27:07

need to know where to look at and

play27:10

currently our use case is uh is built

play27:13

around business finance so our our

play27:17

trackers are are here to simulate best

play27:20

practice SOA a state-of-the-art practice

play27:23

for business finance

play27:26

analysis and what happen is that we have

play27:29

two types of trackers we have trackers

play27:31

that are mathematical and business

play27:34

mathematical are here to spot a

play27:37

mathematical anomalies okay for example

play27:40

a gross revenue increasing

play27:43

significantly above his objective okay

play27:46

it is a threshold that we put that is

play27:48

best practice to analyze corporate data

play27:50

and it will it will spot this threshold

play27:53

automatically within granular data

play27:55

analysis but we get so business trackers

play27:59

business tracker are more uh uh um um

play28:03

specialized rule for to do corporate

play28:06

finance right here I I put an example of

play28:10

gross revenue increasing plus net

play28:12

revenue decreasing it is an an anomaly

play28:15

okay if your gross revenue is increasing

play28:18

and that your net revenue is decreasing

play28:20

is that you have a problem okay so we

play28:24

spot how much automatically and and we

play28:26

have 100 uh just here I just shown you

play28:29

two example but we have hundred of those

play28:33

trackers and we can with our client

play28:37

build more

play28:40

okay imagine that our client have some

play28:43

sort of strategy or some sort of

play28:46

specific problem we can build them with

play28:49

them those type of

play28:51

trackers so these are the symbolic uh

play28:56

part

play28:58

alongside Knowledge

play29:01

Graph

play29:03

so what I have just said until now seem

play29:06

very complicated okay but what we aim

play29:11

for is link

play29:13

reporting we want our app to be in one

play29:19

page

play29:20

okay with two

play29:23

parts we want a part that shows the risk

play29:28

and opportunities found by artificial

play29:32

intelligence right here in the first

play29:35

part you got overview so we we show the

play29:38

user what it it is used to see

play29:41

Consolidated

play29:42

data okay generally in pobi for example

play29:47

reports you got Consolidated data and

play29:50

the user can drilled on okay that's what

play29:53

happen in most reporting here we show

play29:56

the Consolidated data but below it we

play29:59

show where to act okay what are the top

play30:03

insights and right here you got a

play30:07

list so you got a

play30:10

list what are the most important

play30:13

important

play30:15

Insight individually and for each of

play30:18

those

play30:19

insights there there were there were a

play30:21

multi-dimensional analysis and and those

play30:24

are granular slice of data that were

play30:27

systematically analyze and compare and

play30:30

you got on the same line you got a train

play30:34

analysis okay so we we do a dynamic

play30:37

analysis we we show the user uh where

play30:42

this product for example is going in the

play30:44

next month we show a risk this is the

play30:49

risk of not attaining the budgets the

play30:51

budget Target okay we show the influence

play30:56

of outlier

play30:58

if this this uh result was explained by

play31:03

outlier and we show

play31:05

effects effects are uh the mix of price

play31:10

and volume okay uh something can be

play31:13

explain uh by an evolution in price but

play31:16

also by an evolution in volume so we

play31:18

show that in this visualization so this

play31:21

is the first part of our application the

play31:25

second part it is interpretability part

play31:30

what we use we

play31:33

use all fashion natural language

play31:37

processing combined with Genera VI to

play31:41

explain for each of those insights so I

play31:45

have just to click on it and I will get

play31:48

a full explanation of why this insight

play31:51

has been

play31:52

prioritized I will get also a

play31:54

comparision with other product

play31:57

etc etc for the analyst to to know why

play32:02

this insect has been prioritized and to

play32:05

be explainable because we are using

play32:07

artificial

play32:10

intelligence so what I am telling you is

play32:14

that our

play32:15

solution don't mean to erase the current

play32:18

weight to do standard

play32:21

reporting what we are building here is a

play32:24

boost is a a complex mentary is a more

play32:29

effective way to do traditional

play32:31

reporting and it is complimentary in in

play32:34

in many

play32:35

ways so if we compare the uh the ability

play32:40

to to give an overall picture A big

play32:43

picture our reporting is not able to

play32:45

give a big picture even even if we show

play32:49

the Consolidated view but standard

play32:51

reporting is more able to show uh an

play32:53

overall picture or big picture okay

play32:59

does the standard reporting do a

play33:01

complete analysis no I I've just just

play33:04

shown you uh standard reporting don't do

play33:07

a multi-dimensional analysis don't de

play33:09

consolidate so our solution is able to

play33:12

detect early oneing sight uh to detect

play33:15

uh problems when they happen and the

play33:18

fastest fastest way possible and they

play33:21

they do it it do it uh in a systematic

play33:24

way and with a priorization

play33:27

uh algorithm to do

play33:31

that is standard reporting a guidance on

play33:34

where to act no standard reporting in

play33:39

most company are

play33:42

Forest of reports and you don't know

play33:46

where to act you need to do manual

play33:49

digging it's like

play33:51

Fishing manual

play33:53

digging uh and human uh to do for

play33:58

example a drill down approach to

play34:00

deconsolidate manually it takes time and

play34:04

it is a guessing it is an arbitrary way

play34:09

of analyzing data what we do is that we

play34:12

automate and systematically with

play34:15

parallel uh processing algorithm with

play34:19

Cuda calculus we automate the systematic

play34:23

way of the aaging multi-dimension

play34:28

uh the

play34:30

analysis so in term of dimensional

play34:35

analyze with with standard reporting you

play34:38

can go to up to

play34:40

three

play34:43

dimension more you can't show it

play34:45

visually you can't show it in a in a

play34:48

pobi in whatever in a in in an

play34:52

orac you can't show more than three

play34:55

dimension with our solution this is why

play34:59

we don't show visualization in this way

play35:02

we we we do

play35:04

unlimited amount of Dimension at the

play35:06

same time in the

play35:08

meantime as many as the management

play35:11

account accounting for example for you

play35:13

are use case currently

play35:16

carries but what is useful with current

play35:20

reporting is that it give

play35:23

you where to

play35:25

demobilize because on you spot uh an

play35:28

Insight what high the high level

play35:32

management will will want to do is to

play35:35

move resources on those insights to dep

play35:38

prioritize maybe some other project that

play35:40

are not

play35:43

uh a PRI priority thanks to our analysis

play35:47

so this reporting can help you know

play35:50

where to act in this regard where to

play35:54

demobilize demobilize resources

play35:58

so this

play36:00

is the second part on you spot

play36:03

Insight what we are building is Genera

play36:06

VI to help the user build an action plan

play36:09

so I have my insight okay I the Deep

play36:12

detect we call it the first module

play36:15

detect an insight and I want to build an

play36:18

action plan with deep root this is the

play36:20

second module that we we got and what we

play36:24

we did is that we bu we built a

play36:27

methodology to analyze a

play36:30

problem okay what we are

play36:33

spotting we do in a in a multi multi

play36:36

multistep way a methodology to analyze

play36:40

this problem and we are combining behind

play36:42

the H A generate T VI with Knowledge

play36:45

Graph to do that so the artificial

play36:49

intelligence is here to to ask questions

play36:53

to spot information available but also

play36:56

to spot information that are necessary

play36:59

but but they are not available so what

play37:02

we what we would like to do with that is

play37:05

to induce some sort of gumba walk some

play37:10

sort of questioning some sort of inquiry

play37:14

to prepare the analyst to do a business

play37:17

plan and to show that at the actual

play37:20

meeting that will decide prioritize

play37:23

project so we are making a bridge

play37:26

between the we reporting world and the

play37:30

uh project management World in most

play37:33

companies those two words don't talk

play37:37

with each

play37:38

others so we are building a two a bridge

play37:41

between these two

play37:44

words and we are building also a process

play37:47

intelligence in the app uh you can uh uh

play37:51

you have a view uh of what artificial

play37:55

intelligence you use which which Step

play37:58

because we are simulating what what I

play38:00

didn't explain is that we are simulating

play38:02

sparing sparring partners so I am

play38:05

simulating a marketing director I am

play38:08

simulating a financial director so

play38:11

obviously each of those artificial

play38:13

intelligence will have access to the

play38:16

specific data of his domain and will be

play38:20

able to help the user uh build some sort

play38:24

of process intelligence Maybe in the

play38:27

risk opportunity scope you you talk with

play38:30

the the the

play38:32

marketing generative AI uh and after the

play38:36

context change when you analyze the con

play38:39

that has been changing uh with the um

play38:42

Supply uh uh Supply manager you want to

play38:47

go back to the risk opportunity scope

play38:49

and to talk with the uh Chief Financial

play38:51

Officer okay what we are building here

play38:54

is not to automate we are here to help

play38:56

the user find information and build

play39:00

something okay and I and we found that

play39:04

generate VII is not here to help user uh

play39:09

do for example text

play39:11

tosql uh help them ask question to Data

play39:15

Warehouse they are here to help to to to

play39:19

ask to to have methodology to help them

play39:22

get methodologies to help them find

play39:24

information uh in this way they are far

play39:28

more

play39:31

effective so that's was it my

play39:35

presentation I have lots of questions

play39:38

yes no problem because

play39:42

um how did you train the model yes did

play39:45

you train a model or are you

play39:48

using standard kind

play39:52

of for for the generative AI part okay

play39:56

uh for the model currently I am using um

play40:00

because I I have no choice because I am

play40:03

leveraging conage graph I am using uh

play40:06

three 3.5 five uh model of open ey I am

play40:12

I am uh it is a necessity because it is

play40:17

the only model that is is not that much

play40:21

costly and is able to quaring in um in

play40:25

Cipher

play40:29

but but what I I want uh and what I am

play40:32

looking for currently in term of

play40:34

development is to leverage for example

play40:36

zier zier

play40:39

7B or that sort of Open Source model to

play40:42

do that but currently I didn't see in in

play40:46

the research papers uh any evidence of

play40:49

such capabilities I didn't have time to

play40:51

find two an open source model of this oh

play40:55

okay but then for the inside you're

play40:57

using anomal detection model right

play41:00

mostly inside no because what we are

play41:04

using like like I said we are using uh a

play41:07

layer of symbolic artificial

play41:09

intelligence okay we are using

play41:12

effectively uh a Time series to uh

play41:15

calculate uh Trends and uh data

play41:18

scientist will tell me that I don't have

play41:20

enough data uh to do that but we have

play41:22

some technique of data augment

play41:25

augmentation mhm

play41:28

uh and our solution overall is an

play41:32

anomaly detection but we don't use uh an

play41:34

anomaly detection uh we have an an

play41:39

inhouse uh algorithm that is proprietary

play41:42

we are doing a patency pat patency yes

play41:46

claim we have a mark of Shain that is

play41:49

doing the priorization of

play41:51

insight so the priorization that you saw

play41:55

behind the Ood is it is a complex Mark

play41:57

of Shain to prioritize business problems

play42:01

okay and when you said we we using I

play42:04

mean this is interesting long long time

play42:07

ago when I started studying artificial

play42:10

intelligence it was still called expert

play42:14

system yes that's why that's why we use

play42:17

it yes and to see that expert system are

play42:22

coming back after probably 30 years

play42:26

there are I think right now uh recently

play42:30

a stud a study showed that generate tvi

play42:35

is more is more able to do text tosql

play42:39

thanks toled graph where thege graph

play42:42

have a database Shema in it so I think

play42:46

we we are seeing am and I I am writing a

play42:50

lot about it we are seeing uh some sort

play42:53

of neuros symbolic I call neuros

play42:57

symbolic system that are emerging

play42:59

because the problem with a pure deep

play43:02

learning approach is that deep learning

play43:04

is not

play43:06

interpretable uh or or it is very very

play43:09

much difficult to interpret yeah and I I

play43:12

believe that in most use case you want

play43:16

you

play43:16

need interpretability an

play43:21

interpretation because artificial

play43:23

intelligence its role is to if you want

play43:26

artificial intelligence to be

play43:28

prescriptive to prescribe uh

play43:31

action you need a collaboration between

play43:34

what the are the output of the

play43:36

artificial intelligence and the human

play43:38

and to have this collaboration you need

play43:41

some layer of

play43:43

interpretation

play43:45

so and

play43:49

uh but okay when you go to different

play43:52

clients right then you might have

play43:54

different rules business to you

play43:57

different data quality you have

play44:00

different semantics you have different

play44:03

context you have different everything

play44:05

right so yes so SCE you you very there

play44:10

are two question in your

play44:12

question for for data quality what

play44:14

happens is that our tool is so granular

play44:18

is so multi-dimensional that at

play44:21

first what you are getting are lot of

play44:24

data quality issues we are spotting

play44:27

anomalies that are not based on a

play44:30

business problem but anomalies that are

play44:33

based on uh on the quality issue and and

play44:37

we are we we help the client deal with

play44:42

that because we are expert my company

play44:45

has been working on on on this type of

play44:49

reporting tool for almost 20 years and

play44:52

we are expert in dealing with those

play44:54

types of data quality issues so we help

play44:58

them resolve their quality issues so it

play45:00

is the first St of this is the first it

play45:03

is the first tape and what you told me

play45:06

is

play45:07

that our current use case is Corporate

play45:11

Finance so we uh like I said with

play45:14

symbolic AI we built methodologies to

play45:18

analyze Corporate Finance but even even

play45:22

for this use case we can add specific

play45:26

trackers okay currently we are working

play45:28

with a large multidimens multinational

play45:32

corporation in France and we are adding

play45:35

uh specific trackers that are relevant

play45:39

of what what we are doing right now with

play45:42

their data so we can personalize our our

play45:45

tool because uh the Deep learn the

play45:49

machine learning part is is uh is

play45:53

instrument to the symbolic part

play45:56

the the main engine are symbolic the the

play45:59

machine learning part is instrumental so

play46:02

it doesn't it's not the main focus of

play46:07

symbolic layer at the top you show

play46:09

because our true Innovation if I I I I

play46:12

put it blankly it's not an artificial

play46:15

intelligence Innovation it is a

play46:18

methodology method of De consolidating

play46:21

multi-dimensional uh but it is only

play46:24

possible with methodology

play46:27

by uh parallel processing massively

play46:32

parallel processing and with artificial

play46:34

intelligence to do it in a systematic

play46:37

and automated

play46:38

way anyway this is a dream right to be

play46:41

able

play46:42

to um to get early warnings to get early

play46:47

insights to take actions to make

play46:49

decisions this has been a dream since

play46:51

I'm do since I'm doing technology by the

play46:54

way so it's not new

play46:56

it is a dream but we talk about um my my

play47:00

CEO talk about with high level figures

play47:04

in the big

play47:05

Corporation currently uh reporting tools

play47:09

don't deal with Consolidated problem

play47:12

because they don't have a

play47:15

solution they they simply don't have a

play47:17

solution or they just used uh

play47:21

statistical kind of methods right to

play47:24

find ins it doesn't yeah but then yeah

play47:27

then you still need to like you said to

play47:31

deconsolidate right now with a classical

play47:33

statistical method there there are nine

play47:37

nine methods to deconsolidate you can

play47:40

use PCA Dimension reduction etc etc the

play47:43

problem with those method is that uh you

play47:47

can't do it

play47:48

systematically and you need an a data

play47:51

scientist to to to ask himself how here

play47:56

there is maybe a problem and I I will do

play47:58

uh some sort of analysis but it is too

play48:01

late because when it SP the problem it's

play48:04

taking too much time yeah it's taking

play48:05

too much time yes so our our solution I

play48:09

won't reveal or we we did

play48:11

it obviously but but we we take a a

play48:15

surprising

play48:17

approach a Brute Force approach of

play48:20

course I approach interesting I don't

play48:23

know if there are more questions but

play48:26

um so it's quite interesting because we

play48:28

see the application of not only

play48:31

generative AI but also standard machine

play48:34

running uh you you spoke about xgb XG

play48:38

boost right so which is kind of standard

play48:41

machine

play48:42

learning for database for symb combining

play48:47

quite a lot of different yes and for one

play48:52

page because what we want

play48:56

is the Simplicity of it we want

play48:59

something

play49:00

U to be funny uh we are we are

play49:04

incorporating in the app a speech speech

play49:06

to interface but I can talk to the

play49:09

application to show me what I want to

play49:12

see yeah we want something that is

play49:14

simple to understand because we want to

play49:16

focus the Analyst job and they there

play49:19

might be hundreds of them in in our

play49:22

client to to focus on building a Sol

play49:25

solution to the problem that has been

play49:27

spot and not to build reporting because

play49:30

the situation

play49:31

currently most of the time of of of the

play49:36

job day of those

play49:38

analy are building

play49:42

reporting dealing with data quality and

play49:46

exact exactly we want to automate that

play49:49

and we think that artificial

play49:52

intelligence needs to be used to do that

play49:54

because it is possible to do that in a

play49:56

systematic way and we are not here to to

play50:01

to to allow the user to ask any question

play50:03

he wants right now all of lot of Builder

play50:07

are speaking about text tosql unlimited

play50:11

data on demand

play50:14

Etc we think that artificial

play50:16

intelligence is much better better than

play50:18

than us to find problems in a systematic

play50:21

way and to help making decisions based

play50:24

on the experience yes but but to make

play50:27

decision we think that the human it's

play50:29

it's is it's more better to find

play50:32

problems the EI is more better to make

play50:34

decision the human is more better but it

play50:37

can be helped with artificial intellig

play50:38

but you can recommending you can

play50:40

recommend actions

play50:43

exactly you can also ask questions for

play50:46

example you can have the we see it as

play50:50

questions right it is interesting

play50:52

because we see a generative as some sort

play50:54

of recommendation engine it is here to

play50:57

it is here to to recommend action but

play51:00

not to make action for the user yes a

play51:02

colleague of mine was showing me just

play51:04

this morning um just a quick uh quick

play51:08

and dirty app about reporting by the way

play51:12

and then somebody was writing a report

play51:14

and then the app was asking questions

play51:17

that report was not answering right so

play51:21

for example I don't know you have a

play51:22

project stus report and then you forget

play51:24

to men a Target date you forget to

play51:26

mention this that then the the board of

play51:30

the assistant will spot this and say

play51:32

okay what about when do you expect this

play51:34

to happen you need a Target blah blah

play51:36

blah so this kind of thing is really

play51:38

augmenting the intelligence yes because

play51:42

I think gener is is really good to to to

play51:46

to process a lot of data and to infly to

play51:51

to to next token predict what is missing

play51:55

uh what

play51:57

what's present and yes it's really good

play52:00

it's really good to do that like like

play52:03

like you just just said I use it a lot

play52:06

with u um with tabular with tabular data

play52:10

frame and I I I ask it I ask it what is

play52:13

missing from this

play52:15

uh yes yes it's very good and we are

play52:19

using it for that we are using it what

play52:21

we are doing is that you you are using

play52:23

it here in your solution for operational

play52:25

data but not only for unstructural for

play52:29

structured information but also for

play52:31

unstructured information we are putting

play52:33

in graph methodologies a way of doing

play52:37

things a way of doing a problem a

play52:39

particular problem Etc ET to help the

play52:42

user to to upskill the level of

play52:45

expertise of those analyst because uh

play52:48

for our use case we have a problem of

play52:51

turnover for those position and lots of

play52:54

analyst don't stay a lot of time within

play52:57

the company so we are helping them also

play53:00

with their for their

play53:04

expertise Mercy thank you very much

play53:06

thank you thank you it was very

play53:08

interesting good discussion hope

play53:10

everybody enjoyed so I will I will

play53:12

upload the video on YouTube and give you

play53:15

the links on LinkedIn and every

play53:17

everywhere possible thank you

play53:19

Patrick and enjoy nice to talk to you

play53:22

and enjoy the weekend

play53:30

speak

Rate This

5.0 / 5 (0 votes)

Related Tags
Analyse d'entrepriseIntelligence ArtificielleDécision stratégiqueDétection d'anomaliesPrise de décisionSystème d'informationDonnées financièresOptimisation des performancesGénérative AITraitement de données
Do you need a summary in English?