CUSTOMER-INSPIRED AI- THE AWS WAY

---

Examples of Customer-Inspired AI from AWS

Over the past weeks, despite the terrible second wave of COVID-19, the AI practitioner’s world had the opportunity to ride some positive waves too, with significant initiatives from the major players. 

In April 2021, AWS India organized an extensive multi-day analysts’ workshop, AWS netWORK – India Analyst Week, with business updates as well as customer and tech deep-dive sessions. Analysts and thought leaders from all relevant firms actively participated in the interactions. Consequently, in terms of intensity, the virtual event felt no less than attending the real equivalents. Analysts and practitioners debated, discussed, shared doubts and dialogs, and took away new thoughts and insights to share with the broader world of end-users and tech industries alike. 

1-2-3 of the AI Story: Why It Must Pivot on Customer Journeys 

Most great digital tech initiatives capture and represent the supply-side waves. While they enthuse me as a builder and inventor, what “inspired” me most as a hands-on AI Practice Leader for our end-user clients is the unique and integrated digital and AI storytelling during AWS netWORK. The differentiating factors were clear:

1. The entire 360 story in the integrated AI-cloud-data space was told through the customers’ eyes, through the lenses of end users’ business metrics. These were not just promised or targeted outcomes but visible milestones achieved on ground, in the digital journey-maps of the banks and insurance companies, or in the manufacturing/ retail/ supply chain set-ups. 

2. Every piece of AI-ML and related technology stacks, be it the cloud compute/storage infrastructure for AI, the data pre-processing and pipeline management services, or the APIs and modules in SageMaker, Connect, DeepLens and Clarify, all started from specific “Customer Problems” and ended with applied “Customer Solutions.”

3. Each of the customer solutions qualified the DMR (Demonstrated-Measurable-Relatable) criteria of AISWITCH AI metrics best practices, beyond just creating AI-ML technology marvels: 

• Demonstrated (proven outcomes, impact metrics already achieved) 
• Measurable - in monetary value and/or P/L impact (e.g. cost savings in USD, beyond just as % value of any business/ ops process)
• Relatable (for industry peers, partners, providers and even competitors, also for benchmarking and AI-journey maturity assessments that a lot of telco’s, hi-tech and manufacturing & retail companies are doing with us)

What is Customer-Inspired AI?

From the first day first session, Puneet Chandok, Director and Head - India and SAARC, demonstrated this next practice of AI storytelling, with his emphatic use of the phrase “customer obsession”,  reinforcing AWS’s vision in the country – to empower builders and businesses to build a better India.  Subsequent speakers, such as Tarun and Ashish, kept up the pace, and took us through every stage of the amazing tech carousel in the Tomorrowland of AI and underlying digital technologies of cloud and data, through the journeys of a set of early-adopter customers. These end-user leaders didn’t “do digital/ AI/ cloud” just for the heck of it; their businesses actually had operational and strategic problems that couldn’t be solved without these technologies. 

For example, IDFC, a mid-size bank in India, synced their rapid growth story with the speed at which they migrated 60 mission-critical operational and customer-facing enterprise apps to AWS cloud. They took 40 days to migrate the first 40, and then just 8 days to put forth the last 10 of their target enterprise apps to the cloud. IDFC created a consolidated hybrid data lake and built their comprehensive collections platform on AWS. Each of these capabilities is generating efficiencies and business gains, both for the enterprises and for their end customers. The speed and cost of migration, for instance, translate into hard dollar (or Indian rupee) gains in the bank’s efficiencies, as the migrations happen with near-zero latency and guaranteed reliability. 

Case studies of RBL and HDFC Life, too, were shared to show how they reduced specific transaction processing and verification cycle times by rapid implementation of AI use cases that were quickly built and productionized on AWS platforms. The event also highlighted stories of the digital journeys of manufacturing giants such as Ashok Leyland, and how they have, in turn, contributed to the India success stories. For example, Ashok Leyland achieved a 300% improvement in data processing speed and 40% cost reduction with iAlert, which is a connected vehicle platform for 150K+ units. 

How to Internalize the Customer-Inspired AI Storytelling Practice

The user-journey-based technology storytelling that I experienced through those 2-3 intensely engaging days at the Amazon event, triggered multiple epochs of reinforcement learning in my own neural nets, on the targeted next practices for AI leverage. 

From strategic consulting-heavy organizations like McKinsey, BCG, to the consulting plus tech implementation entities like Accenture and Deloitte, the services leaders have been telling this right story of business-relevant, outcomes-focused AI for a long time. 

But in the early over-enthusiasm days of AI builders, awed by the close-to-fantasy model visualizations and algorithmic wonders, the AI tech supply-side almost always started the story from the wrong end- with “how wonderful the hammer is”. Then it went on looking for the nails – the so-called ‘business problems’ – that were often rather imagined than real, ‘invented’ and ‘force-fitted’, to ‘showcase AI adoption’ in an enterprise. 

This empirical problem of pure Geek-speak has been the biggest ‘hidden-in-plain-sight’ deterrent for AI to proliferate in the large business user community– the untapped, unspoken-to but most relevant target user-groups of AI-ML. This problem is rampantly visible in the AI tech-world- be it in the start-ups’ and unicorns’ space or even in case of some of the world’s largest TSP’s. 

The downside of this tech-first approach was obvious and predictable without using any ML algorithms. Eventually, everything started looking like a nail! 

In several cases, the poor hammer couldn’t deliver as per the wrong expectations that were initially set, hence it had to play the punching-bag role- digesting all the blames of wasted resources and efforts. Like the proverbial hammer, the algorithms still cannot retort and speak up for themselves, at least not yet, given we’re quite far from singularity.

For a course correction, these are the 5 next practices that we must inculcate as AI practitioners, as key takeaways from Amazon leaders and clients:

1. Before we, the tech-enthusiasts, push AI into one more oblivion, as we have done over the past several decades, we must SWITCH all conversations of tech-inspired AI to customer- and business-inspired AI. 

2. We must let the business DNA of AI dominate our own genetic make-up as AI builders, because AI is fundamentally a business-driven disruption, making it quite different from other technical innovations that stand out as pure-tech marvels. 

3. We must train ourselves to build AI strategic plans in complete sync with enterprise business plans, given that AI tech-stacks, without the business data context, is just artificial and NOT intelligent. AI, ML, data sciences are all maths and stats applied in real-world business problems and are customer-inspired since their inception, by definition and by design. The domain/business problems are the critical contexts for AI to work upon, based on the underlying data stories. 

4. Cloud, data, and AI (algorithms & techniques, APIs, and containerization) stories are not separate. Hence, they should not be told in a disjointed manner. Every AI customer story is built upon some dark-data story of the customer enterprise across the 4 Vs, and runs on some kind of hybrid multi-cloud infrastructure. This is imperative for AI, given the nature of AI workloads both during training/model building as well as inferencing stages. 

5. Speed is the new currency/oil in the AI world. Enterprise AI workloads in most production instances require rapid and near real-time processing of massive data and parallel execution of transactions/processing/decisions. This combined need for speed, scale, and elasticity often cannot be achieved purely through on-prem compute and storage with limited capacity. 

Ultimately, as Puneet pointed out, there is no start and endpoint in an organization’s digital and AI journey. As more and more data pours into the lakes, as new types of business opportunities and challenges rain down on the enterprises, the journey goes on. The learning ecosystems, therefore, must continue alongside. The AWS India Online Summit scheduled for May 18-19, 2021 will give us another window of opportunity to dig deeper into the stacks of AI-ML, DevOps, Security, Cloud, and DB migrations. As practitioners, we must walk the talk together, to build and share the digital and AI stories of every end-user organization worth its salt.