Winning the Automation Game Harnessing Artificial Intelligence

The Current State of Business Processes Automation

According to a recent McKinsey survey, "although most respondents say it’s possible to automate at least one-quarter of their organizations’ tasks over the next five years, less than 20 percent say their organizations have already scaled automation technologies across multiple parts of the business". These early adopters of scaled automation technologies are not only reducing processing costs by making processes more efficient but are also improving customers' and employees' satisfaction by enhancing efficacy, eliminating irritants, and delivering frictionless journeys. In parallel, the disruptors are creating distinct competitive advantages, attractive product-market fit, and new revenue streams by leveraging the capabilities of innovative machine learning models, proprietary data sets, and the power of cloud computing.

Success Formula

Agnostic of the current state of your automation journey - infancy, emerging, or mature, here's the success formula to win the automation play.

1. Objective: Solidify your objective. Is your strategic goal to optimize the efficiency ratio and efficacy, augment talent, and/or create new AI-powered products and revenue streams?
2. Status Quo Assessment and Target Setting: Perform a status quo assessment. Create an inventory of the intelligent agents, listing their functional and correlated business capabilities across the organization, and the lower and prod environments. Establish a SMART target, and create a framework to define and measure business, operational and process success as a function of your objective, and the scope and type of automation use-cases - rules-based vs judgement-driven.
3. Governance & Talent: Establish governance. Choose an evolving Center of Excellence model, incubating with Centralized and maturing to Federated to support growth as your organization increases the level of maturity in automation. Recruit talent with specific skill sets aligned to your automation strategy - the skill-set requirement for RPA differs from the skill-set required to perform a features engineering and grid search with cross-validation.
4. Funding Model: Establish a funding model to support the program - quick wins, low-hanging fruits, and strategic projects. While the quick wins and low-hanging fruits yield an attractive payback period, strategic AI projects have a more material IRR, given appropriate resources and runway are necessary to capture feasible datasets, and build, evaluate and improve the deep learning models. The best practice is to adopt a traditional ROI-driven approach for a rules-based cluster of use-cases while employing a VC-based seed funding innovation approach on a machine learning portfolio of use-cases.
5. Use-Cases Pipeline: Establish a balanced cross-functional use-case pipeline. The use-cases should vary in complexity and impact, facilitating a continuous delivery model and organizational momentum to facilitate seamless change management. To accelerate value creation, focus on the happy path rather than the edge cases, and choose from the following low complexity, rules-based use-cases with proven success:

HR: Onboarding, Internal Mobility, and Payroll Processing

Finance & Accounting: Intercompany Reconciliation, Accrual Updates, Accounts Payable, and Management Reporting

Collections: Calculation, Letter/Email Creation & Distribution, and Profile Updates

To create a differentiated value, choose neural network-powered use-cases. For a roster of such industry-specific use-cases, feel free to reach out.

6. Technology Model: The technology landscape is highly dynamic. While the market leaders are well established, given the core capabilities of their intelligent agents and the deployment of new features frequently, the overall market is fragmented - new entrants rapidly enter the arena with unique product-market fits and point-specific propositions. Furthermore, the researchers are continuously extending the boundaries and are disseminating the advancements leveraging open-source platforms. As your program scales, it is valuable to follow an automation technology playbook, and frequently evaluate build vs buy vs partner options.

Feel free to reach out for my playbook, inclusive of technology stack radar and evaluation attributes, and build vs buy vs co-create evaluation model

7. Infrastructure: The arena you play in, the use-cases you select, and the approach you choose will determine the infrastructure requirements. For instance, a large BERT model will require 16 TPUs, given 24 transformer layers, 1024 hidden sizes, 16 attention heads and 340M parameters. Comparatively, an RPA intelligent agent will require a standard interactive client, runtime resource and access to a database. For the conciseness of this article, I have ostracized key details here - let's connect and discuss further infrastructure requirements factoring your context.

8. Delivery: The best practice is to leverage user-centered design and agile development principles. Align with the sponsor on a sprint plan, inclusive of the features, high-level scoped user-stories, sprint durations, business testing and feedback requirements, risk management and deployment framework from dev to test to stage to prod, including go/no-go cadence and sign-offs. Deployment to production checklist should include a holistic evaluation of technical, operational and organizational readiness.

9. Change Management: Organizations that realize 10x - not 10% - value from innovation programs have a robust change management discipline. These organizations will include change management resources from the infancy stage of the program so that the change management SMEs can help effectively create and socialize the need for change and strategic vision, and support in building a guiding coalition and champions, removing barriers and institutionalizing change.

10. Post-Deployment Operations: Post-deployment activities include monitoring the performance of the intelligent agents, managing the capacity of the intelligent agents' and infrastructure resources to meet evolving and cyclical business demands, continuously evaluating the throughput and re-training the machine learning models - without overfitting - to capitalize on dynamic consumer and market conditions, and managing incidents and changes in the upstream/downstream tech stack - critical for RPA applications.

11. Continuous Innovation: The trailing decade has been a golden decade for Artificial Intelligence. In this era, we have and continue to experience exponential value creation due to the convergence of cloud computing, big data and machine learning models. As a technologist who started building robots and autonomous vehicles over 15 years ago, have been spearheading innovation, automation and digital transformation programs in over ten industries, and have experience with industry-leading organizations that did not disrupt themselves get disrupted, I can confidently say that AI will have an even more profound impact over the next decade compared to the last one. This, my friend, requires that you continuously innovate to stay relevant.