AI, the endearing buzzword we’re all enduring, prevalent yet perturbing. Can we really go a day without hearing about Artificial Intelligence? Isn’t AI already democratized? So why are so many organizations still confused about what to do, which use cases to select, who are the right partners, where should they focus, when do they start and end, and how do you measure success? There are so many different platforms to choose, exceeding the selection at your local buffet. Much Ado About Nothing, comes to mind. Now what to nowhere?
Challenges
Interestingly, the obstacles we see before us are those we’ve created. The AI hammer, thrashing about and attempting to apply great torque for every use case insight. We’re compelled to find the business problems where AI is the panacea; instead of identifying incremental improvements, we long for the big bang. The Data Lake is built, Data and Analytics Teams are hired, supposedly we’ve defined KPI’s, and the technology partnerships have been established. So, it’s been 6 months and we’ve yet to see ROI. Another 6 months, and some folks are ramping down, riddled with what happened.
Pitfalls
The AI hype has influenced many and impacted few. Really? Pavlovian conditioning guided our decisions on bringing all our data into a centralized location, which we think is the data lake but may be the data swamp. So, now that we have all the data, the self-fulfilling prophecy of the Data Scientists has been enacted, the data science models are getting mass produced by the widget factory, and our vision of “if you build it they will come” has resulted in very few populating the field. The Cloud infrastructure hosting the data science models are metering greater than a taxi in Manhattan, and instead of white now grey clouds abound. What happened? We applied the people, process, and technology levers but there’s uncertainty around what we’ve achieved.
A different approach
So, what is the answer? Perhaps, AI is not sufficiently democratized, we may need more of it. Start small, make incremental improvements, uplift the enterprise in bite-sized use cases. Constrict scope so that thinking becomes clearer and easier to wrap your head around it, forcing you to find a solution. A kind of crowdsourced alternative to the elusive big bang, if you will.
An outside-in approach that facilitates introspection by looking at the enterprise in terms of multiple concentric layers that eventually lead into the core of the enterprise, its organization, and products. Let’s depict each layer as a dimension of the enterprise and evolve it into a Darwinian “Value Network” from a Value Chain that consists of Finance, HR, IT, Legal, and Marketing, where an organization’s context and prioritization are substantiated by data driven weights. Incremental improvements of the organization may be applied to the 5 Dimensions, using the Use Cases below and applying the Punch List. *Note: the 5 Dimensions are inter/intra-connected, where the central node is dynamic, weighted in your organizational context.
Some use cases that lend themselves to an outside-in thinking with compressed, manageable scope:
- Finance: Accounts receivable with OCR
- HR: Recruitment, recommendation & ranking of candidates
- IT: Project management, risks & WBS
- Legal: Contract review
- Marketing: Lead generation for Marketing
Applying the Punch List comes next:
- Business: Persona’s, CX, Value Network (formally known as Porter’s Value Chain), Business Model Canvas, & others
- Data: Lineage, quality & cleansing, labeling, transformation, versioning, and governance
- System: Feature engineering, machine learning (simple à complex), model versioning, Lift & performance evaluation, Model Explainability & Interpretability, Model Serving, and Data Visualization
- Technology: ADDS Ops (AIOps, DataOps, DevOps, SecOps), Containerization, Spot Instances in the Cloud, and API Gateway
To sum up
Here is the path to democratizing AI:
- Identify the dimensions most relevant to your business, perhaps using the 5 Dimensions above as your guide
- Apply Analytical and Enterprise Architecture frameworks such as CRISP-DM and TOGAF, respectively, to guide your decision making.
- Adhere to an MVP Agile methodology during your investigation, analysis, implementation, and deployment, while continuously calibrating your dimensions.
The democratization of your enterprise is a journey, replete with complexities, misconceptions, trials and tribulations, rewards, and enlightenment, but it won’t begin with just visualization.
To realize it you must experience it in layers through an outside-in approach that combines many small successes such that the sum of the parts becomes greater than the whole. Magically, the obstacles before us start to fade away! AI based enlightenment is finally at hand!
A white paper to follow on this topic. Stay tuned!
Author: Sanjay Bhakta, Vice President – Global Head Enterprise Solutions, Marlabs
With additional credits to: Raj Menon, Marketing Director, Marlabs
Leave a Comment