HOW NATURAL LANGUAGE PROCESSING (NLP) IS TRANSFORMING FINANCE · NLP/AI Platform Chatbot Mid/Back Office Ticket / RPA Escalation and hand-back Centralized AI logic and model for chatbot

Begin Content Area = 16,00 End Content Area = 16,00

HOW NATURAL LANGUAGE PROCESSING (NLP) IS TRANSFORMING FINANCE

DR. ISABELLE FLÜCKIGERSWISSQUOTE CONFERENCE

09 NOV 2018

2


WHAT ISCHANGING IN FINANCE

Begin Content Area = 16,00 End Content Area = 16,00Left Spacing = 0,50 Right Spacing = 0,50

Copyright © Accenture 2018. All rights reserved. 3

PLAN TRANSACT ACCOUNT CONTROL COMPLY REPORT ANALYZE ADVISE

• Strategic

planning

• Target setting

• Financial

planning

• Forecasting

• Tax/treasury

planning

• Supplier

payments

• Employee

payments

• Customer

receipts

• Cash

management

• Transaction

accounting

• Asset

accounting

• Tax accounting

• Accounting

close/

consolidation

• Account

reconciliation

• Error processing

• Internal audit

• Statutory

reporting

• Tax filing

• Statutory

compliance

• Policy

compliance

• Regulatory

compliance

• Enterprise

performance

reporting

• Financial

reporting

• Management

reporting

• Portfolio

analysis

• Performance

analysis

• Investment

analysis

• Business

advisor

• Strategy

execution

support

• M&A Support

• Board of

Directors

engagement

PLAN TRANSACT ACCOUNT CONTROL COMPLY REPORT ANALYZE ADVISE

Focus of professional

staff time

Largely digitized

(>80%)

TECHNOLOGY IMPACT ON FINANCET

od

ay

To

mo

rro

w



EXAMPLE: CONTROLLER WILL BECOME DATA MODEL AND MACHINE LEARNING EXPERTS

Accounting Agile ways of

working

Data Models &

Finance

Architecture

Automation

(RPA)

Corporate

Finance

Controlling Process

Expertise

Statistics &

Machine

Learning

New business

models & value

chains

Plan Transact Account Control Comply Report Analyze Advise

Gather & crunch data

Maintain spreadsheets

Generate reports

Support decision making

Drive strategic programs

Personal development

Illustrative: Typical

tasks & time

distribution of

Controllerstoday tomorrow



HUMAN + MACHINES COMPLEMENT EACH OTHER

Lead

Em

path

ize

Cre

ate

Ju

dg

e

Tra

in

Exp

lain

Su

sta

in

Am

plify

Inte

ract

Em

bo

dy

Tra

nsact

Itera

te

Pre

dic

t

Ad

ap

t

Human-only

activity

Human and machine hybrid activities

Humans

complement

machines

AI gives humans

superpowers

Machine-only

activity

H M

IMPORTANT: ACTIVITIES ARE BEING REPLACED NOT JOBS PER SE



HOW A FUTURE WORKFORCE COULD LOOK LIKE

ROLEHUMAN

TECHNOLOGY

(AI/Robots)

ACCOUNTANTS

BUDGET

ANALYSTS

FINANCIAL

ANALYSTS

TAX EXAMINERS &

PREPARERS

AUDITORS

Adaptive Workforce

TREASURERS

Fixed Workforce

Technology

80%

80%

30%

60%

80%

20%

20%

70%

40%

20%

40% 60%

RPA MANAGERS

SCENARIO MODELLERS

DATA SCIENTISTS

ADVANCED ANALYTICS

EXCEPTION HANDLERS

CASH FLOW OPTIMIZERS

FUTURE

EXPERTISEUP TO:



Fact: Finance in practice are not only

numbers, algorithms and models but an

immense amount of administration,

documents, reports, information gathering,

…

8


MARKET EXAMPLES



EXAMPLE: (E)-MAIL RESPONSE AUTOMATION

Use case name: Status Dunning

Customer name:

…

Information need…:

Action: RPA validate & execute

Use Case Template Y

Use case name: Payment reminder

Customer name: John Doe

Customer no.: n/a

Customer email:

[email protected]

PO no: 30087361

Invoice no: n/a

Invoice date: 20.05.2018

Action: RPA validate & execute

SOLUTION DEVELOPMENT

ENTITY/FEATURE EXTRACTIONFROM EMAILS

Email/Ticket

CLASSIFICATON ACCORDING TO DAIMLER CATEGORIES

Hello,

I have a problem with the latest invoice

for PO30087361. The invoice has been

sent on 20.05.2018 and was supposed to

be paid after 30 days. Can you call or

email me? Phone: +49 711 540 8767.

Email: [email protected]

Thanks!

CATEGORIES

Invoice | Payment |

Reminder

Invoice | Status |

Check

Invoice | Status |

Dunning

…

① ②

Apply entity extraction to email/ticket content to

identify key information. e.g. product, serial number①

Based on the identified contents classify

emails according to Daimler categories ② ③

FILLING EXTRACTED INFORMATION INTO THE RIGHT TEMPLATE (IF APPLICABLE) TO GENERATE A RESPONSE

Use Case Template X

Automatically generated structured template

for follow-up processing through RPA

③

COMPANIES ARESUFFERING FROMA WASTE AMOUNT OFE-MAILS FROMCUSTOMERS, COMPLIANCE REQUESTS, EMPLOYEES AND SO ON

mailto:[email protected]

mailto:[email protected]



SPEECH TO TEXT, TEXT TO SPEECH AND STRUCTURED DIALOGUES FOR SALES, CLAIMS, TRADING, RISK MANAGEMENT ETC.

EXAMPLE: THE (TOO) MANY COMMUNICATION CHANNELS

Call Messaging Web Chat Email Form Letter

Contact Center/ CRM

Backend

Classification of the requests & AI/NLP based solving. Depending on the channel with today’s methods 30 – 70% can be automated

Human agents

……

IntegrationFlow

IVRChat

Frontend

Chat Call

Digital Conversion

Facility

Voicebot

NLP/AI Platform

Chatbot

Mid/Back Office

Ticket / RPA

Escalation and hand-back

Centralized AI logic and model forchatbot and voice bot

Backends

Actions are either directly executed by the AI platform or created as ticket to be processed by RPA or other means e.g. trade bookings, compliance information

Escalation and hand-

back



MINING ANDINTERPRETATION OF SIGNALS IN THE MARKET LIKE SPEECHESOF THE FEDERAL RESERVE OFFICIALS

EXAMPLE: SIGNALS IN THE MARKET

Sentiment Analysis

Consistency analysis

with former speeches

Image analysis body

language

Consistency analysis

Conclusion about the future

monetary politics



FINANCE FUNCTIONS REQUIRE MORE AND MORESPECIAL SKILLS. LARGE COMPANIES OFTEN SUFFER TO KNOWWHO IS THE RIGHT PERSON FOR A CERTAIN INFORMATION ORTASK WITHIN AN ORGANISATION

EXAMPLE: FIND THE RIGHT PERSON AND EXPERTISE

① Understands skills

③ Understands relative location

② Understands time, level, languages



COMPANIES HAVEAN IMMENSE NUMBER OF CONTRACTS FOR TRADES. HOWTO ACT ON EMBEDDED FINANCIAL AND REAL OPTIONS E.G. INCHANGING MARKET ENVIRONMENTS?

EXAMPLE: POST-TRADE

▪ Replace SME pool

with an AI routing

engine that will assign

emails to particular

Analyst

▪ Aid analysts with AI based

adjudication application that will

read the assigned emails, identify

the actions and infer the required

data using Natural Language

Processing

▪ Operation desk

receives many post

trade settlement

requests via emails.

▪ The received emails

are mostly written in

English and contains

important information

like SSID etc.

▪ Emails can contain

the information in the

form of plain text or a

PDF or an Excel

sheet

Main Pricing Terms

Other Option Terms

Fixed / Indexation

Other Pricing Options

Volume Swing / Swap / Diversions

Location / Incoterm Options

Quality Specification

Volume Commitment

Embedded Call/Put Options

Tolerance and Take or Pay Penalties

Main Volumetric

Terms

Other Advanced Options

Additional Volumetric Terms

NLP AI algorithms

can review physical

contract terms and

identify embedded

financial and / or

real options.

Identification Of Embedded Options

Gloss

SME

Pool

Analyst

Read emails, identify request

type, infer required data, take

appropriate actions

Settlements

Trade Settlement

Requests

Documents

HOW NATURAL LANGUAGE PROCESSING (NLP) IS TRANSFORMING FINANCE · NLP/AI Platform Chatbot Mid/Back Office Ticket / RPA Escalation and hand-back Centralized AI logic and model for chatbot