HIGH VELOCITY AI - A GAME CHANGER FOR 2022
In our corner of the Artificial Intelligence (AI) / Machine Learning (ML) world, life sciences executives need the best possible answers to pressing sales and marketing business questions. Real-world questions they need answers to include:
How much are sales and marketing promotions contributing to total sales?
What is the ROI of my marketing efforts?
How can sales and marketing spend be better allocated, and what are the forecasted results of the reallocation scenarios?
What are my customers’ engagement preferences and what should their personalized engagement model look like?
Priorities are constantly changing and the answers to initial and follow-up questions are required in days, all reflecting current business conditions. Accurate answers delivered in 1-5 days are what we call High Velocity AI.
So what is needed to deliver High Velocity AI? AI activities are usually carried out in the following timeframes:
Business question definition - 1 to 2 weeks
Data identification and integration - 1 to 2 months
Model development and testing - -1 to 2 months
Insight determination and delivery - 2 to 3 weeks
The core activities don’t change for High Velocity AI, but the time frames are significantly reduced. In the list above, weeks become hours and months become days. This is only possible with the following in place:
Business question definition
Analytic team members need to be fluent in the business from which questions originate and fully understand the key drivers that could influence potential answers. In other words, they should already be thinking about the questions, and they should be able to translate executive requests into meaningful analytics in short order. An hour for the initial discussion with the stakeholder(s) and another hour to confirm what will be answered is sufficient.
Data identification and integration
The data has to already be there! If there is no repository (for the questions above) where promotional and sales data are updated both in discovery and analytic-ready form, High Velocity AI is not possible. The up-to-date “analytics data hub” is the most important starting condition. Most pharmaceutical and biotech companies reached the conclusion years ago that centralizing all of their promotional data in one place was a good idea, but the health, vitality, and integrity of the analytics hub has a direct impact on how quickly the data scientists can turn questions into insights.
Model development and testing
While there can be minor changes and additions in models for specific questions, much like the data, the modeling engines have to already be in place. Sentier has developed an AI platform that has the “on demand” models for attribution, optimization, and prediction that can be quickly applied to most sales and marketing promotional business questions.
Insight determination and delivery
Upon determining the core analytic questions, presuppositions about the potential insights (and how they can be acted on) should already be in the playbook. This does not imply introducing bias into the results; instead it translates into anticipation about what the data may reveal, and then gaming out what the responses and actions will be. It is a standard practice at Sentier that a draft deck is laid out right after the questions are finalized, and then refined as the results come in.
So why be prepared for High Velocity AI? Today, organizations are used to time-consuming AI efforts where essentially every effort is a one-off. They understand that answering these all-important questions is possible and having answers will have a significant impact on efficiency and competitiveness, but they are not aware that there is a better way.
Now imagine just how more effective companies will be if, for example, they can determine the optimal promotional mix for a customer group in one week, as opposed to having to wait 3-4 months for the same answer. AI is more than just a shiny new toy: for pharmaceutical and biotech companies it is becoming a path to deliver answers in vastly condensed timelines. Faster and better insights will ultimately be a positive thing for the companies who adopt High Velocity AI.
