FIO

LOGU

AI Adoption Roadmap to Pharma Companies

EXECUTIVE SUMMARY

Unlike rolling out Systems of Information/Thinking (SoT) (Traditional software applications, ERP systems, etc.), adopting a System of Intelligence (SoI) (Artificial Intelligence & related technologies) to your organization needs a whole new perspective. The following narrative talks about the perspective understanding required by the decision-makers and highlights a few areas where the pharma industry can gain from the advancements in Data Sciences.

BACKGROUND

Let us take a typical pharmaceutical firm that is into drug formulations. Its differentiation strategy is heavily dependent on innovating new drug formulations and streamlining FDA approvals to be ahead of its competition.

FEW AREAS THIS ORGANIZATION CAN BENEFIT FROM THE USE OF AI

Gain Cost Leadership

Organizations spend 20% of their revenues in research, and the sector collectively spent more than $70 billion in 2016 in the US alone. Typical organizations have hundreds of mostly disparate R&D cost centers with scientists discussing, designing and working on thousands of formulation experiments often in silos where learning exchange and thus research spend is suboptimal

Optimized Design of Experiments

Have a clearly defined data acquisition strategy to collect data at every stage. I would even recommend collecting meeting discussions, phone, chat conversations presentations, academic research on top of experimentation findings at various stages. Data science is evolving so fast, and having more data collection points is always better than lost opportunities.

With this collection of past experiments and their outcomes and by augmenting artificial intelligence with human reinforcements, the machines can quickly learn and taper down human interventions and thus constantly evolve on experimentation designs. Inherently machines are good at memorizing and processing large amounts of data and therefore can retain and gain from historic & diverse data, which may humanly not be possible. With all this processing and learning, AI will be able to recommend experiments that are optimized for time vs criticality, current market needs vs resource availability, low cost alternatives etc.

Research Process Automation

Machines are very good at storing huge amounts of data and processing the same; while they are still evolving in areas of creativity, intuition, and existential learnings. So this can be a powerful combination where scientist and computers can complement their expertise.

With advancements in gene and tissue simulations combined with the application of the simulation of the above-optimized experiments, AI systems can virtually simulate complete experiments and lay out a path for human experiments accelerating innovation. With advancements in computing power (Tensors & Quantum) and NLP, this can also offer as a voice-based consulting system for each scientist to learn from previous successful simulations and also command new simulations with ease.

Intelligent automation in the above two areas, combined with Zero attrition, 24×7 working bots, and being able to scale up the machine workforce at a click of a button, significantly optimizes the spend and thus helps the organization achieve cost advantage.

ACHIEVE DIFFERENTIATION

Compliance Automation:

This sector is heavily regulated and thus demands a lot of documentation and audit trails. The current state is to cater to this need as an explicit stage and to allocate resources and valuable time of a research scientist for documentation. By getting interpretability into supervised and unsupervised experimentation models, the steps are all recorded, and most of this stage can be automated. Also, having a uniform data warehousing strategy is key here.

With advancements in language understanding, natural language processing in multiple languages and mediums, AI systems will be able to automate the majority of reporting and compliances, which are more suitable and reliable for the regulatory authorities and approval boards. Directives from ethical AI are forcing research into identifying the learning mechanisms of AI. Thus we will be able to elaborately articulate most of the documentation and reporting requirements, thus reducing time to FDA and therefore, time to markets.

Market Intelligence

Currently, pharma organizations have no direct interface with caregivers or end consumers. Thus there is no feedback or input on how to make the products and services better. Currently, this is entirely handled by the distributor networks, and thus the manufacturers have no insight after the distributor hand-off.

A part of my vision is also to shift this organization from a customer mindset to a consumer mindset by gaining more touchpoints and interactions from the end consumers. I want to leverage big data analytics, social media, and IoT wearables to compliantly extract information from doctor-patient transcriptions, cross effects, anomalies, etc., to understand the causes and scenarios optimizing the effectiveness of the products and to keep evolving in this paradigm. This should let my organization tangent into various focus groups based on our learnings with these touchpoints.

Most of the doctor’s patient interactions are voice recorded in natural language by various doctors and are transcribed for documentation. With natural language processing, these recordings can be processed to understand and store the interactions.

And by introducing data and feedback collecting Anthropomorphizing robots at various healthcare and dispensing units, organization can increase its touch points with the end consumers and use this information to empathize, learn and customize care delivery.

Complaince Automation:

The industry as such is heavily guarded and confine their internal processes and do not share any learnings. There are no open-source system or knowledge sharing systems established, and a lot of duplication and redundancy exists in the sector.

AI can significantly help here by consolidating patents, past learnings. If organizations can open up learnings from failed experiments, without impacting their proprietary objective, it will significantly improve our combined advantage in this area. This can be further optimized at a policy level and by industry-wide collaboration.

The above three areas will enable you to be the differentiator in the market.

CRITERIA FOR SUCCESS

Rolling out AI across the organization surely induces a lot of churn & anxiety. Thus, all the key stakeholders, board members, industry consultants and subject matter experts must be involved right from the initial discussions to ensure an efficient rollout. Refer to our article on How Artificial Intelligence is Transforming Workplaces for our insights on this.

You should funnel out teams from various business processes and redraw the process maps with AI. They can be further optimized to achieve the strategic objectives like cost, differentiation & focus and once the business objectives are articulated, they should be brainstormed at a much detailed level to validate and be adopted by the existing IT strategy, and if needed, have an independent AI CoE.

Introducing AI is a significant operational change and more than that a psychological change to the organization. You should ensure the Organizational Change Management team is involved and equipped right from the kickoff and blueprint stage. The middle management and the shop floor team will be anxious and yes, there will certainly be staffing changes however this must be designed in a way that we can optimize the existing jobs and repurpose the redundancy to expand into other more interesting areas for the growth within the organization.

Change is always taken anxiously by any organization, however, to benefit from machine learning, natural language processing, and robotics it has to be embraced, adapted and healthily critiqued by all its participants. Thus, it must be conveyed and the expectation set that their jobs will only get better and more interesting with this dialogue. The redundant jobs will be repurposed into more interesting and engaging ones, healthy for the entire organization.

There will be a lot of areas where upskilling is required which will be a welcome change. However, the redundant job holders need to be either reskilled or let go. And yes there will be new roles created specifically in the AI area reporting to the board.

You should first seek key organizational stakeholder’s consent and setup a discovery sessions, engaging professors, consultants, startups and peers in this area. At this point, I would involve the directors and strategists to prioritize which departments will be the early adopters of the roll out and then create project implementation plans, resource plans and budget plans on the same. This stage will define the project charter and the roles and responsibilities of that department in onboarding the machine learning systems.

The program implementation plan will be designed to address the business strategy objectives for the management to adapt strategies to identify and prioritize objectives to be adopted for each department. This approach would reinforce the transition process planned for the particular department and contribute to the evolving IT Strategy.

The proposed initiative should address all the three Porter’s strategic objectives as explained below:

Cost Leadership: The cost leadership can be gained from numerous ways one of which is cited above, the procurement advantage we have when all the natural conversations are merged and processed in one NLP system. Also, since the past project information can be stored much easily with the NLP system the subsequent projects gain a lot from cost optimizations, very effective human resource placements & planning, while reducing the research project implantation costs which can be passed on to the market.

Differentiation: With the ease of NLP storing information the amount of knowledge built into the system is vast, such as information retrieved from the combined voice, email, call, document & transaction systems combined with the right processing techniques, the new drug research and experiments will be super evolved, very cost effective and highly productive. With time, the system with access to past vendors, past human resources and the amount of duplication in the new project charter will be able to recommend highly optimal implementation steps. This will lead to faster time to market and with the NLP compliance recording and documentation, the market differentiation can be achieved much quicker than the competition.

Focus: Also with the doctor-consumer/patient interactions becoming a part of the NLP system by way of voice recordings the NLP system will have firsthand information on drug behaviors and its interactions with various classes of consumers i.e., race, age, demographic, economic conditions etc., and thus will be able to make drugs tailored to individuals which will be more effective and thus more focused which indirectly compliments the market differentiation strategy as well.

Technically, I would want to be on a private cloud as the scaling up on processing power and storage space is quick and adaptive. Also, I’d prefer to leverage & gain from open source frameworks as this technology is rapidly evolving and anything proprietary will limit my growth to, they making it available. You should also pay dedicated attention to information security, data privacy and all compliances as this is a very regulated industry and very dependent on information protection and confidentiality.

REFERENCES

Porter, M. E., & strategy, C. (1998). The Michael E. Porter trilogy.

Andrew Ng – The State of Artificial Intelligence. (n.d.). Retrieved from https://www.youtube.com/watch?v=NKpuX_yzdYs

Pulusani, S. (2020, February 14). How Artificial Intelligence is Transforming Workplaces. Retrieved from https://www.fiolabs.ai/blog/how-artificial-intelligence-is-transforming-workplaces/

Bandi, P. (n.d.). https://www.csail.mit.edu/.

Leave a Reply

Your email address will not be published. Required fields are marked *