Who is driving your AI journey?

Artificial Intelligence (AI) opens a whole world of opportunities in every economic sector. However, those will be captured only by companies that craft an achievable roadmap and adoption strategy to become AI-enabled. In this process, a common question that companies may face is, who is the person or area responsible for making AI an integral part of the organization's strategy.

To the question, I would not be able to find a unique responsible for this endeavor, but a set of roles to drive the organization to develop core AI capabilities. For a company to be AI successful it is important to have different areas and key stakeholders actively involved, remove obstacles, and enable real adoption. In the long term, AI won’t be a one-team skill but a company capability. It is not a one-shot investment, it is a journey full of challenges, learnings, fun, and a very promising return on investment (RoI) if things are done properly. This post briefly shares some ideas and learnings about the critical roles in this process, their importance, contributions, and responsibilities.?

. . . . . . .

Data and AI leadership

Usually, it is a seasoned data expert or an experienced team responsible for guiding the company during the AI journey. Good AI leadership is about balancing multiple expertises, finding intersections between areas, and being able to jump into action with the AI/Data team whenever needed. AI/Data leadership must be able to speak with Business Units (BUs), identify valuable AI projects, and translate them to the right methodology and implementation framework. They are also responsible for keeping the organization realistic about AI capabilities beyond the hype and structuring feasible development plans. Finally, AI leadership must be able to work with the technical team (Data Engineers, Data Scientists, BI Engineers, and Software Developers) to ensure that efforts are properly directed and execution is correct. Data and AI's major role is to find the day-to-day, week by week, month by month, efforts that would transform small experiments and developments into consistent and robust company capabilities.?

Data team

Going beyond the hype and the academic State of the Art to create valuable AI developments on real data is challenging and requires very specific individuals and team composition. The data team must be crafted for the company's specific needs, ambitions, and direction. The Data/AI team is the front line to the data, they spend hours and hours crunching data for different purposes which gives them a unique intuition of what your company data is good and not good for. Tap into that intuition and you would be able to quickly discard or select promising directions to develop, ignore that intuition and you will easily end up going in circles without consistent AI capabilities in place. It is important for companies to push their data team to work close to the business problem, at the end of the day Data and AI must be about value creation not about independent prototypes, ETLs, models, jupyter notebooks, and dashboards. These are very useful tools but the major reward is about combining them to create value and efficiencies all around the company.?

Business Units?

Business teams play a major role when adopting AI, without their active contribution and buy-in it won't be possible to mature the developments, get real feedback to pivot the projects, collect data, or capture value from the developed systems. Despite this, it is important to keep in mind that BUs know well their business and how to operate the company successfully, but rarely have AI understanding. The latter must be fully provided by data leadership and the data team. On one hand, BUs must be an active part of the development, and their input considered for modeling strategies and decisions. On the other hand, BUs must be open to transforming the traditional way of doing things with alternative Data & AI-centric strategies. Successful AI projects have plenty of iterations, and every iteration must have input from business experts. It is OK to start AI projects with simple excel files and dashboards while uncertainty dissipates. Business Units are extremely good at using these initial developments and providing feedback about things to improve and things that work properly. These early iterations are important to mature the project definition, craft AI-centered processes, and start solving the integrations with the overall company product and roadmap.

Product team

Data teams must be obsessed with going live to transform "the business as usual". To make it happen companies must find the right spots to release their developments, this is the way to transition from a bunch of early prototypes, excel files, dashboards, and notebooks to real AI features and capabilities. During the early stages of the AI journey it may feel challenging to find these doors for going live, but once found, they are great to start iterating your AI capabilities and release better and more powerful versions of your developments.?

When talking with the product team I use smart assistants as an example of creating an extremely successful door to deploy AI. Back in 2011 smart assistants were dumb, to say the least. However, tech companies were just opening the door to send you better versions of their software on a constant basis. Today we all have a door to the most advanced Natural Language Processing (NLP) and search technology directly in our phones, speakers, headphones, and almost every other device. This same pattern can be found in almost every successful AI product nowadays (Search Engines, Autonomous Cars, Pricing Engines, Recommender systems, etc), they were good at creating doors for AI to mature.??

The product team is a great partner for this. A product strategy without AI functionality will be traditional software, and an AI strategy without go-live points risks staying on your data scientist's laptops forever. The product team must be open to learning and discovering what features can be enabled by AI, and the data team must be willing to partner in the main product direction to push smarter features with each release.

Tech and IT Organization

Tech/IT Org is a critical part of the AI strategy. They are responsible, among many other tech-related duties, for helping the data team serve and scale the AI features and integrate this into the company tools and products. Make sure tech and data get along very well, however, keep data/ai as close as possible to the business, their role is to transform the "business as usual" and the only way to achieve this is by deeply understanding the business. Make sure to have DevOps, and security working closely with your Data/AI team, you will need to build complex systems during all your AI journey, and better to have your information and models running safely from the beginning!?

C-Level

Every successful AI journey must have top-down support, but even more importantly clear expectations about the challenges this journey poses. The C-suite is very important in 2 parts of this process: i) making sure that AI teams are working on the right problems ii) Influencing the organization and making sure that changes happen when needed.

As an example, imagine shifting the credit underwriting process in the financial sector from human-based to real-time semi-automated decision-making. This AI application is proven valuable in the financial sector. Banks and Fintechs are heavily invested to make it happen, however, in practice very few manage to adjust their underwriting process accordingly, and the reason is not only technical but about changing processes, skills, and responsibilities.?

A common challenge that organizations have when developing AI capabilities is to assume that everything can be done with AI and end up having many developed in parallel, ending up with an extremely large and timely packed roadmap with very few resources. C-Level must work with the AI team to align expectations (what can be done, how long it would take, what are the main milestones to achieve it, what resources it would need) and select the fronts that would add more value to the company. At this level, it is important to focus on company benefits, not Business Units (BU) preferences or AI interesting problems. On one-hand BUs will bring several projects to the table to smooth the "traditional" process or remove tasks that they don’t want to do. On the other hand, the AI team would be always biased toward fancy projects, not the ones aligned with the company's best interest. C-Level's role is to keep the development focused on a few highly valuable shoots.?

. . . . . . .

Closing Thoughts

After all of this, who must lead the AI direction of the company? In my opinion and experience, it is not a single person role but a company capability to be created. Every team and person described above plays a major role in this journey. If you have a seasoned Data/AI leader in your org he/she must play a centralizing role, bringing together their technical know-how close to business knowledge and the top-down sponsorship and prioritization of AI initiatives. It is critical for them to educate the organization about AI and create accurate expectations about AI capabilities and restrictions. Even if the Data & AI team is in charge of driving the journey you will need a map to navigate, and the product team must be there to provide it making sure it has core AI capabilities to be developed and enough long-term AI doors (highways) to be successful. Finally, invest in your technical team, developing AI projects is hard and way beyond jupyter notebooks, craft your data team for success not for PoCs.?

. . . . . . .

If you are interested in discussing this or other AI topics, let me know in LinkedIn or in the comments!

Alejandro Betancourt

Linkedin | Twitter | Medium