Model selection for successful AI-based innovation adoption

, , ,

Build your own, buy off-the-shelf, or a hybrid version of both?

The life sciences industry is experiencing a wave of growth driven by innovative technologies and advanced research. COVID-19 has further cemented the strength of the industry by demonstrating to the world how technology and science open the door to a better future. This growth brings a surge of innovation to the industry that spans from drug development through manufacturing to brand management. Technologies are being introduced to improve processes and boost performance.

Specifically, in commercial operations, companies are moving from traditional business intelligence of periodical, hindsight analytics, to AI/ML based advanced technologies that enable predictive planning, science-based strategizing, and automated execution. The dynamic, extremely competitive market, created a need for agile, highly effective marketing. As a result, a plethora of AI-oriented technologies and products have sprouted, in attempt to address this need in various ways.

The question is, what type of solution would work best for your organization? Which strategy would give a timely response to your need and scale to continue and address the needs of your organization? Should you build your own, tailored AI solution from scratch? Or, will you be better off finding a proven, ready-made AI-based solution? And whether you build or buy, how will you attract talents to your team to ensure ongoing performance excellence?

Recently, we held a lively conversation in a roundtable session about this topic with executives from Novartis, SunPharma, Regeneron, and others. A growing number of commercial operations executives realize the importance of AI-driven innovation and are considering how to approach its’ adoption and implementation in a manner that will ensure success.


Which model fits best to your need?

There are three main approaches when it comes to selecting the method that leads to a successful implementation and adoption of AI:

> Building the technology in-house from scratch

> Buying a tool/solutions from a vendor

> Hybrid approach- buying an AI platform and building use cases on top of it


In our discussion we looked at the main attributes that need to be considered when choosing the right approach for your needs:


1. Agility / Flexibility– An organization that operates in an ever-changing market such as the pharmaceutical market, requires solutions that are agile and can be easily tuned and re-tuned to fit the constantly changing needs of the entire ecosystem. Other than adjusting the AI algorithms and ML models, the system needs to support new use cases, new types of data sources, new integrations with other tools in the environment, and more.


2. Scalability– Scalability is probably the most important factor that determines the long-term success of a solution. Without the ability to scale, the value of a solution is very limited, regardless of how good the models are. A robust solution allows you to scale and add more use cases to automate more business processes. Furthermore, life-sciences companies with a large portfolio of products that have relatively similar business process demands, greatly benefit from the ability to easily replicate a successful project from one brand to another, saving in time, effort, and resources.


3. Cost-effectiveness- how long did it take to implement the solution? How well is it utilized? How much value does it bring?. We strive for the shortest time and highest utilization and value.


4. Adoption– Adopting new technologies requires a change of mindset. As complicated as developing technology might be, getting people through change management and enabling adoption is much more challenging. Explainability helps adoption, by creating trust and engagement among sales consultants and field teams. All stakeholders ought to be able to see value in order to get their buy-in and adoption. You want to have them understand enough of the solution to feel confident and even proud in the data science behind it, yet the focus should be on the business challenge and the value delivered, rather than diving too deep into the technology.


5. Maintainability – The best AI models have to be constantly tweaked and improved. Companies need people that understand the business processes, can interact with the models, and continuously give feedback, to maintain quality models. The better your solution has been designed and implemented, the easier it is to maintain it and scale it by lower-skilled citizen data scientists.


Let’s review the pros and cons of each approach side-by-side:


Leading companies collaborate with best of bread partner

After reviewing the pros and cons of each approach we can learn that the recommended approach is the hybrid model: making the most of internal talent & leveraging the best available external talent.  When a solution is implemented on top of a platform, the challenge is no longer in building algorithms or developing technology. The challenges are designing the right business processes that translate needs into solutions, leveraging all available data, and ensuring adoption across the organization. A platform is comprised of a set of capabilities that organizations can harness to flexibly build their own solution. Doing it well will give them a headstart and clear advantage.


Early wins help drive adoption. Start small with one or two brands and let the results do the re-education job that will drive scale. Make sure to include the field reps, get their buy-in by showing them the actual results.  After all, the objective is to deliver value. The objective is not to incorporate technology. Technology by itself is not delivering business value. To implement a scalable AI/ML solution the best way is to assemble rather than build or buy.