Shouldn't we Have Scaled AI by Now? What it Takes to Grow Big with the Mega-Accelerator

Angela Maragkopoulou, CIO B2B / Senior Vice President Business Solutions, Deutsche Telekom Ag

content-image

Apart from several brave, digital AI natives who are clearly on their game with AI (or ML, or Deep Learning, or Data Science or all the above if you wish), most of the rest are timidly wetting their feet and strangely, taking a step back. They try small MVPs, test projects, an investigative piece of the universe here and there. But then, despite the success, they stop. Not growing big. Not scaling. Not transforming towards the so far expected mega-acceleration. Not enjoying exponential change (and benefits, opportunities, growth). Contrary to our expectations, AI is not exploding. Yet.

The reasons for this are as much multiple as they are multifaceted. They have to do both with the natural maturity of organizations and of capabilities and with the alienation of a critical mass of people from the topic of AI. It is both the difficulties of the new as well as the complexity of this kind of new.

1. AI is Complex. And a bit elitist: Everything connected to it is too. AI has to do with tough, mind-bending math, difficult concepts, sometimes the inherent acceptance that the machine got there in its own means. It even accepts this non-information of the human and instead quality-tests the outcome of the machine effort through testing, while accepting the human will never reach the knowledge of how and why the machine decided what it did. This makes AI as an accelerator, far more distant in understanding than other areas of technology and science.

2. The Expert Market is Small and Immature: Data scientists, data engineers, the complete clan is as far from understanding the corporate environment, the enterprise management, the business need as it could possibly be. Either because the expertise area is developing through a pool of academics (and we all know academics are closer to science than making money) who need significant time to develop business thinking or because the pool is young and business inexperienced.

Truth be told, we run to the hated question: “What exactly do you want to achieve?” coming out from the mouth of an AI person all too often. Which means zero added value for this expert market at this given point in time.

3. The Lack of Any Kind of Data Strategy is Not Being Addressed: And no, data strategy does not mean putting all your data in one place, on one bet or in one lake. Data generation, data labelling, data storage are the key feed to AI. The un-siloing and choices can make or break business models based on AI. They can differentiate products, operating models or possibilities to revenue generation. For political or knowledge-gap reasons, data remains in large organizations mostly fragmented and siloed. Either everyone wants a piece of the pie or no direction is given for this resource. But without data, you are an AI zero.

For a technology or IT organization to explode in AI, there needs to be a shift from siloed work to interdisciplinary collaboration

The answer to how can we move in order to scale AI has three axes:

- Culture Relevant: For a technology or IT organization to explode in AI, there needs to be a shift from siloed work to interdisciplinary collaboration. We need business, delivery, operation, and analytics to work as one. It is then mandatory to forego an agile transformation before scaling or while you scale AI. And as organizations need to digitalize first, AI will scale only if we move from experience-led leadership to data-driven decision-making.

- Expertise Relevant: It is of no wonder that an organization needs experts to be able to scale the functions relevant to AI. But the direction is NOT pure AI skillset. The drastic need is for AI translators. Building AI capability from your own business experts or nurturing the business expertise into AI specialists is the way to go. Create a university which will build this exact time of business ownership from the rare animals data scientists are. Help them understand your commercial environment, your business model, your financial landscape. Team them with interdisciplinary teams, working side by side with the company process optimizers and assist them to make the step change. And above all: let them know that the most hated question is that of asking for specifications. We cannot allow AI experts to default to the IT world of the nineties. We already know too much.

- Operating Model Relevant: Give as much power and mandate as you possibly can to your data strategy and your data strategist. Data strategy for AI to succeed and explode needs to stand side by side with uncompromisable principles like security and customer privacy. Have the CEO back this up, create a role reporting directly to the Board, boost and uplift as much as possible. Data needs to be incorporated in every architecture in every area, in every aspect for data to be reachable and exploitable.

Apart from the above, finally and in the most compelling way possible, spend money, time and effort to educate everyone. All the time, continuously and meticulously. Make AI part of the everyday picture and create confidence to the employees, they know what it is, how it evolves and how they can innovate with it. Democratize it, give access to it, let people play with the newer more accessible versions of it.

Then, it will self-scale much like everything else AI self-does.

Weekly Brief