converge shaping ai for southeast asia - deloitte united states · 2020-03-16 · converge shaping...

ConvergeShaping AI for Southeast Asia

July 2019

Cognitive Analytics Solutions Centre

02

Converge - Shaping AI for Southeast Asia

03


AboutConvergeConverge is a series of reports by Deloitte Southeast Asia Innovation that provides insights into the technology trends and startup ecosystem in Southeast Asia.

Each report provides a snapshot of the upcoming technology trends in a particular sector and introduces promising startups that are driving new ideas and taking on challenges that are unique to Southeast Asia.

Join us in discovering interesting and potentially disruptive startups and initiatives across a myriad of industries and feel the pulse of Southeast Asia’s thriving digital revolution!

Who the report is for:• Corporations exploring innovation capabilities and startup engagement• Individuals who are keen to understand technology and innovation trends driven

by the startup ecosystem in Southeast Asia

About Deloitte Southeast Asia InnovationDeloitte Southeast Asia (SEA) Innovation is a cross-function, cross-country innovation unit dedicated to driving the innovation agenda as a culture and value creator across the region.

Have feedback on Converge? Drop us a note at [email protected]!

About Deloitte Cognitive Analytics Solution Centre of ExcellenceDeloitte's Cognitive Analytics Solution Centre of Excellence (CASC) is a global initiative supported by the Singapore Economic Development Board, that uses cognitive and advanced analytics techniques to deliver risk sensing and predictions to global clients.

Limitations: While we try our best to ensure that this report is up-to-date and accurate in the presentation of information as of the date of publication, there are some limitations to this report. For example, some startups choose not to disclose their funding rounds and/ or funding amounts. As such, the data might not reflect the exact situation.

Sources: This report draws from startup databases such as Tracxn, Crunchbase, e27 and Tech in Asia; reported data from local, regional and global news sources; as well as respective company websites.


04

The Era of Artifical Intelligence

Artificial intelligence (AI) is everywhere these days. Human beings are building machines to perceive the world and make decisions like human beings do, to the extent that machines may outstrip the cognitive ability of the average person in the near future. As machines become smarter, a whole array of concerns appear, from the efficacy of the deployed AI to the ethics of letting machine-run efficiency affect our jobs and our way of life.

The pinnacle of an AI future sits at the intersection of machine learning brilliance and benefit to humans. In the business context, a professional supported by a “super human” machine can combine the best of what humans and machines have to offer. That said, checks and balances will need to be built into the human-machine interaction to safeguard against unintended consequences.

The deployment of AI entails its own chain of activities, from data collection and preparation to ‘train’ the AI, to investing into component technologies that enable the promise of AI to manifest. In the decades to come, startups, corporates and even governments will continue to pursue machine learning excellence in the many realms that AI will eventually enter, aided by the

continued improvements to supporting technologies such as processing power and cellular network technology.

Explorations into frontier technologies – including deep learning, industrial Internet of Things (IoT), image recognition, quantum computing and 5G networks – will always be expensive. Big technology companies like Google and Facebook, that have both the financial means and an immense amount of customer data, have a huge headstart on the AI race. That said, given the zeitgeist of an open innovation ecosystem, smaller players have been able to adopt some level of AI, which can still have long-lasting and widespread impact on businesses and citizens alike.

In this report, we will explore the various ways Southeast Asian startups have incorporated AI in their solutions to change the way business is done in their respective countries. Often, this is dependent on infrastructure provided by the big technology companies. The startups then add a layer of contextual insights and relevance in order for the AI to produce optimisation, recommendation, and personalisation suggestions that suit their market needs.


05

How to train your AI

Machine learning (ML) is a subset of AI that allows computers to learn and improve on its own without intervention or programming adjustments by humans. For an AI system to be effective, it is very important for it to be fed with relevant data so that it learns to recognise what it should and should not be processing. AI-training is especially important when the tasks on hand deal with unstructured data, like images and natural language, which generally stray from pre-defined models and categorisations.

While enterprise-facing industry solutions like programmatic advertising and cybersecurity are mostly able to run on the same core technology regardless of location, public-facing industries like retail and customer service require more customisation before deployment, to be able to adequately manage the linguistic, cultural and other differences that exist across countries.

Helping companies build AI that understands the local nuances of this region are startups like Supahands and Next Billion (see Table 1), which have emerged to provide or collect data points as required by clients. Relying on the gig economy, these startups are able to amass huge localised datasets efficiently through mobilising a crowdsourced workforce armed with affordable camera smartphones. With such data points, they are able to provide businesses with contextually-accurate insights.

Recognising Diversity

When Apple released the Face ID feature with the iPhone X in 2017, the idea of using facial recognition as a biometric security layer gained mainstream interest.

However, it also brought up the problem of AI being trained on a skewed dataset. A lady from China claimed that the iPhone X could be unlocked by her colleague with Face ID. This occurred even after getting the initial unit exchanged, prompting discussion about whether the AI software itself was the problem as opposed to the hardware1.

Other companies like Google, Hewlett Packard and Nikon have also experienced embarrassment previously, when users shared online about instances in which their facial recognition software failed to recognise non-white faces in particular.

Giving AI a Southeast Asian flavour

Table 1 - Startups digitalising data collection

Startup Description Country Founding Year Total Funding

Supahandshttp://www.supahands.com

Supahands provides training data for ML and AI, collected using proprietary technology and a crowdsourced workforce.

Malaysia 2014Series A - Undisclosed

Next Billionhttps://www.nextbillion.asia

Next Billion builds data platforms to provide insights into communities, consumers and retail trends.

Singapore 2013 Undisclosed

*Information and numbers sourced from Crunchbase, Tracxn and company websites

1. https://www.newsweek.com/iphone-x-racist-apple-refunds-device-cant-tell-chinese-people-apart-woman-751263


06

Can bots speak my language?

A low-hanging fruit for any customer-facing business looking to improve their operations with AI is the implementation of a chatbot, which makes for a cost-effective service agent that can work 24/7 when properly trained. One of the biggest challenges for chatbot adoption in Southeast Asia, however, is the lack of Natural Language Processing (NLP) ability.

Southeast Asia is home to a myriad of languages, dialects and vernacular, and effective NLP ability for this region needs to process everything from Thai to Teochew (a Chinese dialect), to the distinctions between Manglish (Malaysian English) and Chinglish (Chinese English). Indonesia alone speaks over 700 languages and variants2.

The lack of suitable NLP libraries to train AI in local languages has been noticed, and startups in the region have endeavoured to fill this gap by creating chatbot companies and building up NLP libraries (see Table 2). Beyond improving the linguistic ability of bots, these startups are also catering to their local markets by developing chatbot solutions on social media and communication platforms that have the highest adoption locally, and are embedding their solutions on platforms like Whatsapp, LINE, and Facebook Messenger instead of the usual websites. In addition, startups like Kata.ai and BJTech also provide developer suites, allowing local businesses to use their technology to build their own solutions, enabling more end-users to benefit from the ease of accessibility and instant responses that chatbots can provide.

When bots learn to be too human

In 2016, Microsoft created a Twitter bot named Tay that learnt to speak casually from interacting with users on Twitter. Microsoft had to shut it down within a day because it quickly devolved into a racist, sexist and hate-spewing internet troll3.

After the disastrous social AI chatbot experiment with Tay, Microsoft released a politically-correct version of Tay. Called Zo, it has the persona of a stereotypical teenage girl, who gives a wide berth to politics and any other potentially inflammatory topics4.

Machines mimicking human communication perfectly may be a worthy technical aspiration, but it may not always be a desirable one, if it is a facet of humanity that should not bebestowed with super-human capabilities.

2. https://www.techinasia.com/indonesia-foreign-startups-10-things-know-list

3. https://gizmodo.com/here-are-the-microsoft-twitter-bot-s-craziest-racist-ra-1766820160

4. https://qz.com/1340990/microsofts-politically-correct-chat-bot-is-even-worse-than-its-racist-one/

Table 2 – AI chatbot startups


Kata.aihttps://kata.ai

Conversational AI company, providing the ultimate toolset for developers to build sophisticated chatbots.

Indonesia 2015Series A - US$3.5M

BJTechhttps://bjtech.io

AI conversation platform designed for small and medium businesses that want to enhance their customer care experience.

Indonesia 2015 Seed - US$1.2M

Zwiz.AIhttps://www.zwiz.ai

AI chatbot and analytics platform for business.

Thailand 2017 Undisclosed

Expa.AIhttps://expa.ai

A unified solution for sales, marketing and customer support.

Myanmar 2017 Undisclosed

Source: Crunchbase, Tracxn, respective company websites.


07

From digitalising field operations…

While the push for digitalisation is not new, the growing promise of AI to bring about deeper insights and greater productivity is giving new impetus for businesses to transform their manual recording and field operations. Traditionally, roadshow promoters and retail auditors have been offline channels for collecting marketing insights. Increasingly, these roles are becoming the bridge between the offline and online worlds — field personnel are now able to digitally collect text and images easily, neatly and immediately from their on-the-ground interactions, when equipped with a customised phone or tablet application. Effectively, offsite employees can now feed real-time information into an online database, which the wider business can then act on more quickly as well (see Table 3).

...To achieving cost savings with smarter automation and hardware

Business automation, which can relieve the human workforce from repetitive standardised tasks, is an aspect of business improvement that is getting a boost from the development of AI as well.

Highly improved photographic, telephonic and sensing technology is allowing vast amounts of good quality information to be captured from scanning anything, from documents to calls in the work environment. In tandem with this development, the ability of AI to understand and process unstructured data is expanding the scope of processes that can feasibly be automated. AI-powered process automation is able to help businesses keep a constant watchful eye on systems, observing data and workflows in real-time, and acting on known issues that match identified patterns.

TAIGER is a Singapore-based startup that specialises in providing intelligent process automation. This can take the form of user-facing interfaces like chatbots and virtual assistants to handle clients and help staff, or systems running in the background that automatically recognises, extracts and processes information from uploaded data. TAIGER claims that its AI has been able to help its banking, insurance and government clients slash processing time by up to 90% and reduce costs by up to 80%, while maintaining about 90% accuracy5 (see Table 4).

In June 2019, TAIGER signed a global strategic alliance with image and photographic scanning equipment manufacturer Kodak Alaris. By combining TAIGER’s AI software with Kodak Alaris’ hardware, document scanners are transformed into intelligent productivity tools, with superior Optical Character Recognition (OCR) that can immediately capture scanned information for automated processing6.

5. https://ie.enterprisesg.gov.sg/media-centre/news/2018/3/ai-firm-taiger-gives-clients-more-bite-in-slashing-costs

6. https://www.alarisworld.com/en-gb/about-us/newsroom/2019/taiger-kodak-alaris-strategic-alliance


The AI Augmented Worker

Table 3 – Field operations productivity startups


Powatahttp://powata.com

Powata offers technology solutions for retail workforce management.


BetterTradeOffhttps://bettertradeoff.com

BetterTradeOff reinvents financial life planning with big data and analytics.

Singapore 2015 Seed - SGD$3.4M

Ekohttps://www.ekoapp.com

Eko is an internal communication and operations platform for businesses to bring together the right people, information, and tools to get work done.

Thailand 2012Series B – US$28.7M


08

As intelligent business process improvements continue to garner interest, startups are also finding niches to serve. Singapore-based Exact.ai, for instance, aims to transform back office operations for the media and advertising industry by providing sector-appropriate tools for payment and reconciliation, compliance control and auditing, and delivery tracking.

From monitoring to processing to optimising workstreams, the role of the human being in such functions will definitely be affected as smart machines take up more of the load.


Table 4 – Workforce and process automation startups


TAIGERhttp://taiger.com

TAIGER provides robotic process automation solutions primarily serving the banking, insurance, and government sectors

Singapore 2009Series B - US$31.3M

Exact.aihttps://www.exactai.com

Exact A.I. is an automated delivery reconciliation and payment platform for media. Using proprietary A.I., they support and improve back and middle office processes.


Will AI kill jobs?

This perennial question will never grow old. From automation to AI, technological progress has always caused human beings to worry about being displaced by robots.

A September 2018 report released by Cisco and economic consultancy Oxford Economics7 postulated that by 2028, 6.6 million jobs will be lost across the six largest economies within the Association of Southeast Asian Nations (ASEAN) grouping due to the adoption of AI and technology. Referring to Singapore, Malaysia, Thailand, Indonesia, the Philippines, and Vietnam, the research found that the biggest job losses based on the “displacement effect” would be in the agriculture & mining sector (see Fig.1). On the other hand, many other sectors are expected to see a net gain in job creation due to the “income effect”, or the productivity growth from deploying smarter machines.

If history has taught us anything, jobs will ultimately continue to exist, but the new jobs created from technological advancement will require a new set of skills. Those who fail to transition will need the social support of the state and other welfare providers to find their footing in this new economic order.

7. https://www.cisco.com/c/dam/global/en_sg/assets/csr/pdf/technology-and-the-future-of-asean-jobs.pdf

4.3-10.1

4.9

2.9

6.1

3.9

1.9

3.9

0.2

- 4.3

-2.0

- 4.4

- 2.9

- 1.2

- 3.0

% Change

5.5%

11.7%

11.5%

13.5%

11.5%

13.4%

13.6%

10.7%

% Change

- 12.9%

- 10.3%

- 9.2%

- 9.4%

- 8.2%

- 10.0%

- 8.4%

- 8.4%

Agriculture & Mining

Manufacturing

Utilities

Construction

Wholesale & Retail

Transport & Tourism

Business Services

Goverment &Community Services

- 0.1

Fig.1. Positive and negative impact of increased tehcnology adoption, by industry sector(ASEAN - 6, number of workers (axis), percentage of workforce (labels), 2018 - 2028)

Millions of full-time equivalent (FTE) workersSource: Oxford Economics, Cisco


09

The AI Augmented Industry

Behavourial patterns: a personalisation opportunity for businesses

With improvements to hardware and IoT sensors, data collected is getting increasingly personal. From individual preferences based on tracked frequency and intensity of usage, to one’s gait and mannerisms, so much can be uncovered about the uniqueness of a person from triangulating a few data points.

With AI becoming more powerful and accurate over time, mass personalisation is no longer an oxymoron. While the implications may be more ominous for individuals, this development of AI enables businesses to take a look at individual customer profiles, such that they can better provide individualised services and offerings, and also pick up on specific risk profiles.

As employee and customer transactions move online onto the increasingly intelligent smartphone, the opportunities for AI to promote topline growth and manage risks are available to almost every industry.


10

1. Security and identify verification

While two-factor authentication remains a norm, new technology is enabling other forms of biometric identification to become more effective as security layers.

Helping e-commerce players block fraudulent transactions is Singapore-based startup CashShield. Boasting real-time fraudulent pattern recognition even for new users without much historical data, coupled with passive behavioural biometrics, CashShield claims to apply principles of high frequency trading to make instant decisions on whether or not to block transactions (see Table 5).

Meanwhile, AI is also allowing voice to become a biometric identifier. China-based VoiceAI has patented solutions providing identity verification through the signal processing of a mere two or three seconds of user speech. Language-agnostic and relying instead on diction and other physiological indicators that can be assessed through speech, this technology has been deployed to facilitate the disbursement of pension funds for about 2.5 million retired civil servants in Indonesia..


Table 5 – Identity verification startups


CashShieldhttp://cashshield.com

CashShield helps companies manage their risks from fraudulent payments and hostile accounts.

Singapore 2008Series B - US$25.5M

VoiceAIhttp://www.voiceaitech.com

VoiceAI is the industry's leading provider of voiceprint recognition and intelligent voice solutions.

China 2016 Seed - CN¥10M


11

2. Personal recommendation and consultationFrom media content streaming platforms like Spotify and Youtube, to e-commerce platforms like Lazada, it is common for media and consumer platform applications to provide recommendations that keep users engaged with the repeated use of the platform. In order to generate these suggestions, these businesses track consumer use and preferences on both an individual level and aggregated user segments, sliced and diced from the overall dataset in various ways. Understanding the propensities of individual choices vis-à-vis other user profiles and segments allows businesses to better tailor their communication efforts and content for each user, which boosts engagement and sales. In the healthcare sector, personalising data-driven predictions and recommendations has more serious implications for one’s well-being.

Healint, a Singapore startup, is the developer of a migraine tracking software called Migraine Buddy which helps individuals keep track of their migraine patterns and pre-empt migraine attacks (see Table 6). Using the combined data of over a million registered users from all over the world, Healint generates real world evidence for patients, doctors, and researchers to understand migraines better and work towards improving treatment outcomes8 . With the data from Migraine Buddy, Healint has worked with pharmaceutical giant Novartis on several research pieces on migraines.

Is AI better than the doctor?

Did you know that bone fractures are most likely to be misdiagnosed between 8pm and 2am? A misdiagnosis prolongs pain and injury as patients and doctors do not pursue the necessary treatment.

Working on this problem is LogixLab from Malaysia (see Table 6), which has developed an AI-powered diagnostic imaging analytics solution. Aimed to help physicians rely less on their naked eye to pick out a single hairline fracture, the AI solution would deliver X-ray descriptions, identify specific lesions of interest, and suggest treatment. It would be able to do so consistently, regardless of the time of the day.

8. https://www.businesswire.com/news/home/20180819005021/en/Healint-Announces-Research-Showing-Increase-Anxiety-Depression


Table 6 – AI-powered healthcare startups


Healinthttps://www.healint.com

Healint is the developer of the migraine tracking and research platform, Migraine Buddy.

Singapore 2013Series Unknown - US$1.4M

LogixLabhttps://www.logixlab.tech

LogixLab aims to provide advanced and accurate AI-powered software for imaging analytics.

Malaysia 2017 Unknown

8. https://www.businesswire.com/news/home/20180819005021/en/Healint-Announces-Research-Showing-Increase-Anxiety-Depression


12

AI is also starting to help shed light on areas that have been more of an art than a science, one example being human resource management. Removing most of the guesswork, IBM announced early in 2019 that their AI can predict with 95% accuracy when employees are about to quit, allowing HR to plan for early intervention and coaching to avoid unnecessary attrition9 .

For jobs that require personnel to work with hardware and other physical resources and facilities, it is also possible for AI to tie the various elements together into a single dashboard view for continuous monitoring, complete with prompts and alerts based on AI-powered analysis.

By implementing various monitoring infrastructure - including wearables, telematics and computer vision - industries as wide-ranging as manufacturing, farming, trucking, facilities management, and construction can all be analysed in real-time and monitored for predictive maintenance and intervention. Now that planners have a better understanding of the status of every single element in motion, as well as environmental factors that may affect the output, businesses are able to deploy software, hardware, materials and human resources more optimally (see Table 7).

9. https://www.cnbc.com/2019/04/03/ibm-ai-can-predict-with-95-percent-accuracy-which-employees-will-quit.html

Looking at team sports through AI

Team sports can be said to be a microcosm of what happens at the workplace, condensed into an intense session of competition, in which individual talents have to work well together as a team in order to win.

In the world of e-sports, research organisation OpenAI built an AI that defeated a team of professionals and world champions at Dota 2, which is an immensely complicated multi-player team game with a huge number of characters and other game elements10.

The permutations for Dota 2 are endless, and many situations are happening at any given moment in the match. For OpenAI to be able to train its AI to win Dota 2 shows that current AI technology and machine learning techniques are able to tackle large-scale and complicated scenarios, upping the ante on the type of mainstream problems AI will be able to solve in the years to come.

10. https://www.theverge.com/2019/4/13/18309459/openai-five-dota-2-finals-ai-bot-competition-og-e-sports-the-international-champion

3. Resource management and monitoring

9. https://www.cnbc.com/2019/04/03/ibm-ai-can-predict-with-95-percent-accuracy-which-employees-will-quit.html

10. https://www.theverge.com/2019/4/13/18309459/openai-five-dota-2-finals-ai-bot-competition-og-e-sports-the-international-champion


13


Table 7 – Resource management and monitoring startups


Lauretta.iohttps://www.lauretta.io

Lauretta.io leverages on AI, ML and big data to provide human resource management solutions.


SmartAHChttps://www.smartahc.com

SmartAHC provides IoT-based data collection and analytics solutions for pig farming.

China 2015Series A - US$2.9M

DRVR https://www.drvr.co

DRVR provides fleet management and analytics.

Thailand 2014 Seed - US$522K

Striking the balanceEffective AI for businesses can only result from the concerted effort of those across the entire value chain, with a balanced consideration for both technological excellence and the humans interacting with the technology. The progress of AI is also dependent on the state of complementary technology to track and sense elements that may otherwise be imperceptible to the regular person.

The open innovation ecosystem remains an important bedrock for the development of AI and all its use cases, and new possibilities emerge when new data inputs are available. As AI practitioners seek new streams of highly localised data to aggregate, analyse and act upon, however, valid concerns about data privacy and information sharing will naturally continue to be raised. Already, such concerns have led to the emergence of standards like the General Data Protection Regulation (GDPR) in the European Union. It will be necessary to find the right balance between safeguarding privacy and promoting innovation as the AI industry continues to develop and grow.

While this report is focused on the enterprise uses of AI, AI can already be seen in many aspects of our day-to-day lives. The growing incorporation of AI into enterprise solutions only means that we will come to interact with it more and more, as an employee, a customer and a resident.

The interaction between humans and technological advancement will continue to enthrall and worry us all, but one thing is for sure: even as AI becomes more integrated into the workplace, the need for human labour will remain, although the concept of work and competitive advantage will change. It will be important for societies to adapt to AI augmented demands on a workforce and take care of those who may get left behind, in order for the promise of AI to prevail for the greater benefit of humanity.

14


About Deloitte Southeast Asia Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited (“DTTL”), its global network of member firms, and their related entities. DTTL (also referred to as “Deloitte Global”) and each of its member firms and their affiliated entities are legally separate and independent entities. DTTL does not provide services to clients. Please see www.deloitte.com/about to learn more.

Deloitte is a leading global provider of audit and assurance, consulting, financial advisory, risk advisory, tax and related services. Our network of member firms in more than 150 countries and territories serves four out of five Fortune Global 500® companies. Learn how Deloitte’s approximately 286,000 people make an impact that matters at www.deloitte.com.

Deloitte Asia Pacific Limited is a company limited by guarantee and a member firm of DTTL. Members of Deloitte Asia Pacific Limited and their related entities provide services in Australia, Brunei Darussalam, Cambodia, East Timor, Federated States of Micronesia, Guam, Indonesia, Japan, Laos, Malaysia, Mongolia, Myanmar, New Zealand, Palau, Papua New Guinea, Singapore, Thailand, The Marshall Islands, The Northern Mariana Islands, The People’s Republic of China (incl. Hong Kong SAR and Macau SAR), The Philippines and Vietnam. In each of these, operations are conducted by separate and independent legal entities.

This communication contains general information only, and none of Deloitte Touche Tohmatsu Limited, its member firms, or their related entities (collectively, the “Deloitte Network”) is, by means of this communication, rendering professional advice or services. Before making any decision or taking any action that may affect your finances or your business, you should consult a qualified professional adviser. No entity in the Deloitte Network shall be responsible for any loss whatsoever sustained by any person who relies on this communication.

© 2019 Deloitte Southeast Asia Ltd.

converge shaping ai for southeast asia - deloitte united states · 2020-03-16 · converge shaping...

Documents