Picking Right Enterprise Data & AI Platform Strategy

Background

A modern enterprise data and AI platform has become a non-negotiable need for a company of a certain size.

The central idea behind such a platform is that it serves as a central repository where all the data can then be converted into knowledge that can then be used by business users to deliver value across various functions.

Therefore, it is imperative that we state clearly and unambiguously that data is a first class citizen in any enterprise and therefore needs a place to thrive and grow. An enterprise data and AI platform is therefore vital to the enterprise transformation journey.

Various data processes such as data engineering, business intelligence, analytics, machine learning further help enhance the value of this data by performing various operations variety of opertions right from cleaning and preparing data, unifying data from different data sources and creating new metrics/measures that help organisations who believe in power of measurement to streamline their operations, perform root cause analysis and understand drivers for a certain entity, create a holistic understanding of customers, employees, processes and measure and improve each and every aspect of enterprise process.

As one finds out the canvas of generating potential business value from data is huge. But enterprises have to traverse through the key decision of what kind of modern data enterprise platform suits their needs. There are various ways enterprises can build a modern data and analytics platform for today.

Build your own

** Should you build your data and AI from the ground up? **

Thats potentially a multi-million dollar question often running into 10s of million depending on the scope of such a data project endeveour.

A decision to build on your own should have a much more stronger business value case, since it is coupled with long term maintainence and support costs, keeping up with the technological improvements over time to upgrade and modernize a home grown platform. In addition, newer technical requirements emanating from new forms of use cases and plathora of design choices have significantly increased the complexity of building such data architectures, often making a project execution of a build approach itself risky.

Hence our view is “Build your own” data and AI platform needs to emanate from a clean longer term business strategy and goal that is articulated well and differentiation spelt out. This needs to be further supplemented with a well defined execution scope, time, human capital and cost allocated to such initiative.

At Cynepia, We beleive the answer to above question is almost an absolute ‘NO’ for 99.9% of the enterprises, If the larger objective is to use the data to derive operational analytics and decision analytics value by using the data and analytics infrastructure.

** Buy various point SAAS solutions and integrate. **

Here again there are potentially 2 main approaches:

• ** Would purchasing various Commercial SaaS solutions and integrating make a far more apt sense ? **

This has really been an approach on tear for past several years. Clubbed under a loosely defined term “Modern Data Stack”, A Lot has been written and discussed on this topic over last few years. A quick and apt summary of what is clubbed under such a terminology is discussed/reviewed by Approva Padhi from Foundation Capital hereModern Data Stack: Looking into the Crystal Ball.

The article also summarizes issues/areas of improvement for such a stack. At Cynepia, we beleive that there are many issues with the so called “modern data stack” approach. A few among them are as follows:

1. Lack of finished and consistent user experience and multiple end application interfaces leading to sub-optimal user centric design and productivity gaps.

2. Lack of thought to devolution of data and analytics skills leading to lessor scope for democratization.

3. Higher human cost to keep the modern data stack up and running over time handing integration issues and dealing with sub-optimal architectural choices made by the different SAAS tool vendors.

Often this would need adding further layers of software applications.

4. Cost conundrum due to piece meal approach and dependece on multiple SAAS vendors to deliver the unique SLA need of your organisation.

Leveraging Data APIs and apps from cloud vendors such as AWS, Azure or Google Cloud a better bet?

• Another way to build the data and analytics platform is to using many of the tools, point services and programmable interfaces provided by players such as aws, azure and google cloud platform. While the usual benefits of cloud such as flexibility and scalability are available here too. In addition to the above mentioned disadavantage, there are a few additional disadvantages:

1. Price variation: Often this services are priced by usage and cloud services bills tend to grow exponentially as the data grows. This is seen as a big irrant in services adoption on cloud beyond basic infrastructure services

2. Governance: Data Security & Governance of many of these cloud platforms, tools and applications was designed to be generic and therefore is quite combursome, but often may need specialist to be hired to keep your data secure flawlessly.

** Buy a modern unified data and AI development platform **

There is yet another class of vendors who provide a bundled data and AI Development Platform. These platforms are usally a subset of data and AI applications built on a common architecture and user experience theme unlike unbundled SAAS Applications.

However, compared to the hype, many of these unified platforms/vendors have often lost the ability to shape the market because of variety of reasons:

• Actual promise vs reality

Often incomplete or combined by acquiring different SAAS companies such as above and bundling the solution, completely ignoring the user centric design approach.

Often Costly and priced as per the needs of fortune 500/1000 companies making them prohibitive for the larger enterprise market.

• Missing Pieces of the puzzle further leaves adding further layers of software applications.

Is Modern No Code/Low Code UI Unified Platform the future?

** Unified Low Code/No Code Enterprise Data & AI Platform **

At Cynepia, we see another category which is not just modern and futuristic and ensures data and AI are democratized and has the potential to reach far more users and use-cases and larger enterprise market. For name sake, we call this category ”Low Code/No Code Enterprise Data and AI Platform”

So what are the advantages of a “Low Code/No Code Enterprise Data and AI Platform”

First and foremost to set our orientation right, we strongly believe in a unified data and AI development platform, because thats the best way to make analytics affordable and expand foot print beyond fortune 500/1000 companies. Ofcourse the baseline is actual promise should match reality.

It transfers technical debt of architecture, integration/support of multitude of SAAS applications to the vendor instead of you as the customer.

We are believers in No Code/Low Code Platforms potential to deliver better value. Why not for code? It is not to suggest that for code platforms is a bad idea. Every no-code platform rides on code, so that would be an oxymoron in a way, if one is proposing no-code.

Here’s are top reasons to go low code/no code:

It’s hard to hire best of bread/smartest of the Data Architects, Software Engineers, Frontend Engineers, Data Scientists and Project Managers at the price one is willing to pay. If one can, one could potentially generate a similar or potentially better outcome as a No Code/Low Code Platform. So that leaves you with next grade of skilled developers, who then need to be trained, upskilled and hopefully you have an code output that fits your aspiration. However, most likely case is you are left with more data debt than code that you can call asset.

Most Enterprises are not software companies by business. We have seen often keeping focus on building code over extremely longer term is hard, since quick business outcomes often supercede long term value of such code assets. Historical data evidences suggest, for code data pipeline shelf life is much lower than one beleives it is.

While multitude of best of breed SAAS applications still remains an alternate choice, but if an enterprise is focused on business value and ROI. There is no reason for it to technical debt of integration, support and stitching architecture is transferred to the customer than the vendor. There is also an overall ROI question on the whole data and analytics investment as discussed earlier.

How do you strategize?

Data technologies are evolving quickly, making traditional “build your own” may be risky, time consuming and inefficient. If you haven’t already made investment in any architecture, we would strongly recommend to invest in a user centric Low Code/No Code Enterprise Data and AI Platform.

If you have already invested in one of the above given architecture and accruing technical debt which feels unmanageable, you may want to restrategize and rethink of investing in a data and analytics architecture that suits not just your today’s need but also future.