Artificial Intelligence Is Not Easy - 5 Pitfalls To Avoid While Embarking On AI Programs
‘World hunger and poverty are now eradicated from the planet earth with the leverage of AI and technology’ is the news headlines I continue to live with big hope every day. However, it is still a far-out dream, and I am still hoping that as the human race, we can achieve it one day.
What upsets or puzzles me is that a lot of false hopes and promises are being delivered and marketed in the market that AI can, or individuals are on the path to achieving something way out which is too far-fetched. They are mainly selling gimmicks; the reality is something else. Don’t get me wrong, I am not here to talk about the perils or anti-AI adoption within the market. I am a big believer and work with clients, partners, and teams to drive and derive value from AI and exponential technology investments. In addition to this, AI and exponential technologies are becoming critical differentiation for many organizations. Increased investments and setting-up the organization structures continue to be ascending.
There are many success stories in the industry where researchers, organizations, and individuals have been able to have significant breakthroughs on leveraging and deploying AI and ML techniques to solve or enhance business/socio/economic/healthcare needs, which are making a huge impact. If you speak to all such bodies on their experience and paths that they adopted, which led to such achievement or irregularities. Scenarios, exceptions, and learnings (failures) are a natural path to seeing success in AI programs; however, it is inescapable not to think that the journey is going to be accessible and achievable.
In the last decade, working with different clients, teams and programs, I have penned-down what are the possible top reasons or areas organizations/academia/government institutions should keep in mind when embarking or driving their team/individuals/partners to deliver an AI program or outcome –
Don’t Sell the Dream: The most successful programs in AI and ML which have delivered exponential results have been narrowly focused. The best approache I have seen being adopted and recommend is that the AI program inputs, outcomes, and success measures be clearly defined at the start of the program. With a caveat that there are clear expectations that the program can be a failure as well without any outputs or wins.
In a design thinking approach, one can break-down the fields, identify proxy use cases or areas, and leverage the applications of the areas to align to business needs
Data as a Critical Element: Teams, executives, and many other individuals who are aspiring to drive the AI programs underestimate that the most crucial component of the program is the data. Data challenges are and have been the most vital variable in the success and launch of the programs. Data challenges are of different types and not limited to availability of data, gaining access to it, quality & quantity of data, the ability for source or other systems to ingest the process data back, data ownership and privacy concerns limiting the usage. AI can really work and fulfill its outcome if the classic Vs are met (Variety, Volume, Velocity, Value, and Veracity)
Experts are in Abundance: The second piece which is duly neglected while beginning on the AI journey is on who would do the work and what kind of skill sets would be needed to deliver the required results. A single team member or two might not be able to solve or get the desired results. It is inescapable that a cross-functional and also a variety of skilled professionals come together to drive the program outcomes. With the advancement of AI and a variety of methods available, reaching the same conclusion or an answer can be daunting. The approach, methodology, and application genuinely depend on the future of the program and that a single data scientist cannot solve for. In addition to this, other organizations are also battling to get individuals deployed within the organization. With the pool of skillsets being limited, it is essential that you have the right skills and roles mapped-out to ensure the success of the organization. If you are unsure of how to scale or don’t have a budget to do that, it is good to initiate the work with a partner’s help or leverage technology if it can be solved with the out-of-box technology solutions.
Don’t Take the Bull by the Horns: Organizations often make a big mistake to speed-up the deployment of the program, allocate a significant budget, and expect ROI quickly on it. In my humble opinion, if an organization is initiating the exercise, the budget should be tagged as a Research Expense in the balance sheet and for organizations. For the organizations where the journey has already started, then the expenses (people, technology and other process related expenses) should be heavily tagged as CAPEX instead of OPEX, as the outcome of AI would lead to competitive differentiation and create assets within the organization. Financial management, report-outs, and also demonstrating ROI is critical for the success of the AI program. When I talk about ROI, it is essential to note that it is not necessary to show savings or lifts, but can also be the learnings or findings which can be a treasure cove for organizations.
You are Not Alone: ‘Our industry, category, the problem is unique, and you would not be able to solve this’, or ‘please share examples for our category, product, focus area or specific to our company’. These are the questions I have received from executives or board members during the discussions of AI. No problem or industry or focus area is unique. In a design thinking approach, one can break-down the fields, identify proxy use cases or areas, and leverage the applications of the areas to align to business needs.