Auditing

USE CASE

Problem Statement

  • A credit card clearing house has an oversensitive fraud detection algorithm
  • This leads to many false positives on the fraud system
  • Each false alarm requires expensive human intervention – phone calls and verification

SUE Solution

Sue was trained to look at the each fraud alert and rate it for the probability that it was real. The client then chooses what probability requires action and which to ignore.

Metrics gathered

Sue gathered all the characteristics of all transactions that have caused alarms before including a mix of both real and false positives.

Success and Improvement levels

Fully 35% of the false positives could be removed without affecting any of the true positives.

Implementation Timeline

2 months to collect the necessary data at the client followed by 2 weeks to develop the model

Similar possible Use Cases

In any situation where a human can look at a transaction and tell if it is problematic, Sue can look at ALL transactions and test them. Sue moves from sample auditing. Telecoms, Banking, Finance, Triage of Insurance Claims
SUE

The ability to rapidly create and deploy virtual experts

Sue is a technology that allows experts to transfer their knowledge to computers. This gives the organization the ability to replicate the decisioning performance of the top experts throughout the organization. Sue then allows the company to manage outcomes and adjust the model to its changing needs. Virtual experts have near limitless capacity and can be integrated and deployed anywhere and in any operational environment.

The ability to rapidly create and deploy virtual experts

A Sue virtual expert is built in 4 easy steps:

1. Specify what the decision is and what questions you would ask to make the decision.

2. Teach. Sue will ask you numerous configurations of your questions to find your underlying rules.

3. Calibrate. Optimize your bot to suit your needs.

4. Distribute the model throughout your organisation so everyone can benefit from your judgement.

Manage your model

1. Change old decisions to influence behaviour.

2. Add importance to neglected variables.

3. Remove human bias to align decisions to strategy

4. Learn the hidden rules that guide ‘expert opinion’ in your organisation.

5. All with an easy graphic interface

Make processes transparent

Sue has built in graphic presentations of your decision structure that allow you to:
1. Learn what the unconscious processes are in expert decision making

2. Make those decisions transparent to the board

3. Justify those decisions to regulators

4. Change the way they are made in accordance with new guidelines

Other benefits

  • Review past decisions rapidly and detect fraud.
  •  Receive warning emails when decisions are taken that do not match Sue.
  •  Use Sue behind the scenes for decision auditing.
MORE USE CASES

Introduction

Where the Sue AI bot is being implemented in the market. This list is not comprehensive as the number and variety of implementations our users are generating are without end.

Monitoring IOT/Telemetry

Sue can convert any telemetry from a reactive rule base to a predictive model that will avoid unplanned stoppages and nip problems before they become critical. Sue is the ideal AI for IOT. With the vast amounts of telemetry being generated by IOT, there are not enough experts and engineers to read it all. Sue can read every telemetry reading and email/autocall/refertohuman when action is needed. IOT, Mining, High Precision Farming, Telcoms, Vehicle Monitoring

See IOT Use Case
See Vehicle Telemetry Use Case

Call Centres Next Best Move
Sue can recommend the next best move based on what is put in on each screen, standardizing service and optimizing flows.

See Use Case

Data Enhancement
Sue can predict missing data points in a set using the other data.

See Use Case

Automatic Renewals
Sue can approve renewal of risks and bills risks within fuzzy thresholds. Underwriting, Utility Bills
Auditing
In any transactional environment, Sue can read every single transaction and flag outliers and exceptions for human review including fraud, money laundering or pricing. Sue is able to imitate an expert that reads a page of transactions and spots the fraudulent one. Except that Sue can read ALL the transactions in a large organization and never needs sleep. Telecoms, Banking, Finance, Triage of Insurance Claims

See Use Case

Medical
Sue can learn any individual decision in the medical field and replicate it
Onboarding Questionnaires
Most AI engines need 80 pieces of information to fire, leading to client fatigue. Sue can give an initial approval after as few as 3 parameters. Credit cards, Employee applications

See Use Case

Cyber Risk
Sue can act as a help desk able to run set scripts 24/7 anywhere in the world, detect data breaches and protect database by running scripts and smart approving every db/os action
Targeted Marketing
You don’t need Cambridge Analytica to target advertisements or profile clients. With a few pieces of information Sue can segment your whole population and send each person the right message. Investment, Retail Deals, Politics
Trading
Sue is being used to analyse specific metrics, economic indicators, weighted risk parameters and soft input to take shareholding decisions ending in Buy/Hold/Sell. Transaction Pricing, Shares, Currency Trading, BlockChain Smart Contracts, Actutarial Pricing Methodology, Fitment of Statistical curves, Reserve Forecasting

See BlockChain Use Case
See Share Trading Use Case

Application Selections
Unlike humans, Sue can review every single item offered and give an initial unbiased opinion Underwriting, Risk Selection, Recruitment Selection, Holiday Authorizations, Exploration Predictions
Staff Predictors
Sue can read the psychometric results of a candidate and give you the probability of success in a division based on the results and performance assessments of other candidates.

© 2019 ANALYCAT