Mon. Jul 22nd, 2024

Implementing AI at large organizations

Originally published in Medium by the author

What’s the unique transformative technology that has hit us in the 21 st century? The answer to this from any business leader is usually AI or Machine Learning. It’s normal to see the business community raving about it.

“AI is one of the most important things humanity is working on. It is more profound than, I dunno, electricity or fire,” says Pichai

New business and services are getting launched that takes advantage of this technology. Here is a list of most promising Americas AI companies as rated by Forbes magazine in 2019.

The fascinating aspect of this article is that all the companies here are new. Missing from the list are the giants of corporate America.

Apple acquired Siri in 2010 to bring AI capabilities to the iPhone but has struggled to scale it till date. The article mentioned above brings an interesting question. Why are the resource-rich American corporations unable to leverage this transformative technology?

Large organizations traditionally buy technology via product route or outright buy companies like Apple did to acquire transformative technology. At this point, it’s safe to say there are structural issues that are blocking corporate America from adopting AI via ‘acquire and integrate’ route.

Let’s take a closer look at what are these structural issues.

Data data everywhere

A standard corporation is flush with data. Everything from customer to supply chain data is recorded for various purposes. Last 20 years of improvements in storage technology has made it possible to collect and store massive amounts of data. Still, only a few of the traditional corporations have managed to leverage this treasure trove of information.

Machine learning that underpins AI requires unbiased data to improve its efficiency. Bad data can quickly deteriorate the performance of your algorithms. American corporations see a lot of inorganic growth during its lifetime. As they acquire and integrate these smaller corporations, they are left with a spaghetti string of data systems that do not work well with each other.

The outcome of all the inorganic growth is a large number of isolated data lakes that’s unusable from a machine learning perspective.


Most corporations are rules-driven. These may be corporate regulations or policies put in place to comply with regulations, but rules are a significant barrier to implement machine learning successfully. AI paradigm expects that these dynamic rules are generated through the machine learning process.

The rules-based approach is always not wrong. Especially in the early stages, it’s essential to have rules to bootstrap the machine learning engine to understand the customer. A good example of this is a news feed app asking you to choose topics that you like.

In an internal organizations example, this may translate to determining access/permission based on standard rules. Once the algorithm understands the role of employee, it can provide tools and authorisation without a human micromanaging it.


This is probably is the toughest challenge. Most corporations are top-down. Even in flat organizations there are policies and procedures. It’s clear that a traditional organization would not be able to take advantage of AI until executives start operating as coaches providing resources and training. This is a significant shift from the conventional model where executives and product managers dictate the experience.

In the example above in a conventional company, executives would have decided what tools and permission are applicable for a specific employee instead of trusting the machine learning process.

Product managers and executives must resist the temptation to jump in, which is a fundamental mind-shift.


Implementation is the last hurdle in a corporate transformation. Any transformation that involves machine learning is time-consuming. Time is required to join rationalize and generate unbiased data streams. The development teams that are used to building rules into solutions has to learn how to leverage Machine learning algorithms. Another major challenge is managing machine learning failures. It will be tough to resist the guarantee that a rule-based paradigm provides especially during a significant crash.

Finally, the essential aspect of implementation is a buy-in from the board and C-Suite. Implementation requires restructuring of the organization and clear definition of roles of various internal teams. The transformation would not be feasible unless Owners(Shareholders/Stakeholders) understand the importance of change and the cost involved in terms of time and resources.


Traditional corporations are at a turning point. To ensure continued growth and success, they need to embrace an AI-first approach and embark on a transformative journey. This would require them to apply modern practices and re-engineer its operation model.

If the progress in services available to the consumer is any indication, then the organizational effort required is well worth it.

References & Notes

  1. https://www.forbes.com/sites/jilliandonfro/2019/09/17/ai-50-americas-most-promising-artificial-intelligence-companies/#b98fa59565c5
  2. https://appleinsider.com/articles/10/04/28/apple_acquires_siri_developer_of_personal_assistant_app_for_iphone

By NaveenSankarS

Founder@AmericanGarage( @aginc_us ). Passionate about #Healthcare, #Education & #Entrepreneurship. Medical tweets are not advice. Support: http://patreon.com/aginc

Related Post

Social media & sharing icons powered by UltimatelySocial

Enjoy this blog? Please spread the word :)