How to Make Your Company Ready for Artificial Intelligence

Now that we are seeing many pharma executives appreciating what AI can do for them in their role within pharma, they are wanting to take advantage of AI. However, in many cases when they start, they then find many organizational roadblocks rear their head and hinder their efforts. Artificial Intelligence (machine learning, big data analysis, cognitive computing, deep learning, intelligent agents etc.) is essentially the ability of machines to utilize data and reasoning to achieve an objective. The amount of data that can be integrated is essentially infinite, and means that AI can churn through trillions of data points and analysis far more effectively and rapidly than a human can.


Most in pharma now understand that AI can integrate varying types of data – both structured (spreadsheet) and unstructured (e.g. text, video, social media etc.) and that this greatly enhances their understanding of both their markets and customers and uncovers new insights previously not identified. The use of this in many industries, including pharma, is starting to change business with impressive results. However, it is not magic and a lack of understanding and readiness can limit companies embarking down this route due to unrealistic expectations.

So, what can a company do to get their organization AI-ready?

Here are some steps to consider when embarking upon an AI journey.

1. Examine your strategic goals and objectives and what challenges you are currently facing in achieving these.
List which are the must win items in your plan, and which are the biggest challenges you are trying to overcome. We start all projects with a situational analysis to understand the clients’ situation and strategic objectives so we can determine where AI can assist best them achieve their goals and overcome their challenges. This starts with asking questions of the client to understand their strategic objectives and key challenges – sometimes this is relevant to a brand portfolio, sometimes to overall company growth. One client last year asked ‘Can’t the AI do this for us?’ The answer is no. Humans are required which leads us to step 2.

2. Understand what AI can and cannot do.
You will need to match your strategic objectives and challenges to be solved to what AI can actually do if you are planning how AI can help. Humans are a necessary part of the equation in many areas – building the IT infrastructure to ingest and process and warehouse the data, identifying what data is relevant and usable (containing raw individual level data with unique identifiers), cleaning the data, building the algorithms and training the model and then to iterate it. I saw a great quote of relevance here:

I initially thought that AI and machine learning would be great for augmenting the productivity of human quants. One of the things human quants do, that machine learning doesn’t do, is to understand what goes into a model and to make sense of it. That’s important for convincing managers to act on analytical insights. For example, an early analytics insight at a pharmacy uncovered that people who bought beer also bought diapers. But because this insight was counter-intuitive and discovered by a machine, they didn’t do anything with it. But now companies have needs for greater productivity than human quants can address or fathom. They have models with 50,000 variables. These systems are moving from augmenting humans to automating decisions.

3. Focus on which specific business objectives AI can solve and provide the most business value for you while meeting your strategic objectives.
Start small. AI works best when the challenge or objective is well defined and understood, and where the data is available to be analyzed to make the decisions required. If you ask a question such as ‘What message will retain this physician now?’ that is a straightforward machine learning problem to solve. However, if you ask ‘Find something’, then you may get some answers irrelevant to your business goals and strategic objectives. It may be something to look at later but is not a good place to start. If you have logical rules to apply, you may not need machine learning or AI.

However, if the rules are unclear and follow complex non-linear pathways, then AI will be powerful to assist. If the business objective is too unclear to define with respect to the data, create a sub-goals for the data to solve along the way to ensure the team are effective. The AI can assist in the decision making as you get to the main objective. Eularis have our Pharma strategy analysts on every project to guide our data scientists as to what the code should be looking for that tied in with strategic objectives, so where it should be looking and what we are trying to understand.

4. Know what data you have and how to access it and then identify what is relevant.
This appears to be the biggest challenge our clients have. Much data is in numerous third party vendor systems and the users of that data don’t know what is in the entire database, or how to access it. To ready up an organization for AI, solving your data issues is probably the first biggest hurdle you will face. If you would like access to a data inventory template that Eularis use with our clients, let me know and I would be happy to send it to you. That is the start.

Then accessing all that data and deciding which to include is the next step. This is a role for the data scientists as they need to look at each data source and run tests on the data to see if it is usable. There are many aspects to this. The first is a visual scan – if it is aggregate data then it is unusable. The second is looking at the structure and cleanliness of the data. So for example does it contain a lot of missing data?

This doesn’t mean it is unusable as data scientists can write algorithms to impute the data (fill in the blanks) but then if they do that they also have to test to ensure that the imputation was accurate. Cleaning and preparing data for use is a large part of a data scientists job. The other reason to sort your data in advance is that connecting data to a central platform (data ingestion) can become costly when being set up for different data sources. So rather than including all your data (some of which will be unusable due to being aggregate data rather than raw data etc.) which can cost a lot in data scientist time, and big data engineering time, don’t try and do everything at once.

AI sits on top of the big data platform (IT infrastructure) to provide analysis and actions from the data. Of course, if you have the budget to feed all the data in, do, but be aware this can add up in both time and cost. One pharma company doing this currently is spending close to $500 million to do this globally (and they are doing it in-house) and it is taking around 5 years. That is why it is often better to start small and then scale as you go.

5. Choosing and Building a big data platform.
There are a few decisions to make here. The first being in-house or cloud. We typically use cloud platforms as they are cost effective, easily scalable, and each project builds on the previous project data so as you get each win with data put into the system, you are still building the overall comprehensive ingestion and architecture. Just not all in one big go.

And they are secure as ours and the big ones all are kept up to date with all the latest security standards and audits. Regular independent verification of security, privacy and compliance controls is maintained and several independent third party audits are performed on a regular basis to provide this assurance. This means that an independent auditor has examined the controls present in the data centers, infrastructure and operations. The cloud has annual audits for the following standards: SSAE16 / ISAE 3402 Type II: SOC 2, SOC3, ISO27001, ISO27017, ISO27018, PCII DSS v3.1.

The third party audit approach is designed to be comprehensive in order to provide assurances of the level of information security with regard to confidentiality, integrity and availability. Customers may use these third party audits to assess how this cloud platform can meet their compliance and data-processing needs. Our Cloud Platform will also support HIPAA covered customers. So, as you see, security in the cloud is pretty tight.

After implementing these steps, you are ready to engage with creating your big data platform set-up and data ingestion and implementing AI. Eularis can help you through these but if you have completed the first part of step 4 (identifying your data and gaining access and samples), that makes the whole process smoother and a lot faster.