The power of AI is critical to revolutionizing many business processes and unlocking new value, but getting it right is really hard, few are doing it well.
By now, many of the opening data points in this article from Forbes are well known:
- "According to Venturebeat, an estimated 87% of data science projects never make it to the production stage."
- "TechRepublic claims that 56% of global CEOs expect it to take 3-5 years to see any real ROI on their AI investment."
Everyone has a favorite cautionary tale about a highly visible AI project failing.
And, our favorite quote:
Though this article is not specific to life sciences R&D, we believe many of the key points are industry agnostic. Read the full article when you have 5 min. In the meantime, here is a summary of key takeaways:
Questions to answer before engaging in an AI project. These nicely mirror questions you should be answering before any big tech investment:
- Define the business goals or challenges you are trying to solve
- Choose the right solution to the defined problem. AI is not the right solution to everything (though pop culture may make it seem otherwise right now)
- Assess culture, people skills, and change management and identify gaps that need to be filled
- Define KPIs that will indicate success
One item glaringly missing from the list above, though it is reflected in the featured quote, is fully considering current data business processes and putting together a data strategy plan. Without enough data, in the right format, easily accessible and regularly updated, even the most carefully planned AI project will not succeed. Data engineering is rapidly emerging as an imperative companion to data science and AI.
"The Pistoia Alliance released a survey in 2019 that showed 52% of respondents cited insufficient access to data as one of biggest barriers to the adoption of AI."
In terms of measuring ROI, there are some smart suggestions in the article. The most interesting dichotomy is the suggestion to focus on using AI to drive growth and expansion, but calculate the break even point based on when the initial investment equals cost savings like reduced headcount and higher efficiency to minimize risk.
Do you agree with this article? What else do you think is missing? Let us know!
Learn more about how we do the data engineering for you, automatically harmonizing and centralizing experimental data. Our data platform connects disparate silos to activate the flow of data across your R&D ecosystem, preparing it for exploration and analysis with AI and data science. www.TetraScience.com