We’re in an age where promise and fear for applied artificial intelligence (AI) are colliding. Healthcare continues to be at the forefront of innovation, when advances such as Google DeepMind’s yielding actionable acute kidney injury (AKI) diagnosis are equally lauded and debated for their ethical application. The promise of AI and healthcare rapidly takes us from leading edge to beyond expectation, providing an avenue to deliver healthcare to underserved communities and to uncover patterns at a scale that humans rarely access.
That promise takes so many forms, from personal assistant to prescriber and patient alike, to clinical trial accelerator, to driver of precision medicine. As Lauren Maffeo put it, “At first glance, healthcare should have an advantage: This industry has a huge amount of data. In fact, the volume of healthcare data will surpass finance, media, and manufacturing via compound annual growth of 36% through 2025. This is due in large part to increased use of emerging technologies in healthcare, such as medical imaging, chatbots, and big data analytics tools.” To an industry awash in data, our challenge is not whether to use it, but how best to use it to responsibly bring benefit to as many parties as possible.
As healthcare marketers, we are in a similar position: How do we leverage AI to better understand our audiences and align our messaging to their perspective? When we are able to partner with patients across their journey we can anticipate and react to forthcoming needs. With marketing leadership expressing interest in core AI capabilities (content personalization, predictive analytics, next best action), we’re seeing AI capabilities forming mainstream expectations for content engagement (https://bit.ly/2lxv0fg).
We’ve begun exploring how AI might be utilized to counter expectations, to delve into our own self-filtering behaviors, as well as how to scale regulated messaging. Additionally, we’ve explored how to downscale predictions, all in order to provide a valued experience to our content consumers.
Using AI to Engage (AKA Shatter “Filter Bubbles”)
AI advances have provided us with a carefully choreographed perspective on our world. The more valuable the recipient, the more layers that exist to filter information for value, whether a physician’s support staff, or an executive’s administrative assistant sifting through mail. Algorithms have been working to parse inboxes looking for errant or malicious messages (with varying success). A common, long-standing question then arises to the creators and distributors of marketing messages: How often does our marketing message actually reach its intended audience? Whereas previously we were relying on AI to shelter us from too much marketing/SPAM, now we seek an end to the barrage of content, the dreaded overload of information. Our AI filters have come to create a safe space where only the content we wish to engage is available.
With many calls to “burst” these bubbles, what happens when we utilize that same technology to carefully open the space to new ideas? Network analysis and the use of “next-best-action” predictions may allow for rapid investigation of the customer journey to examine referral paths to key content, and expose ways not to confront users with new perspectives, but to find ways to open doors, create relationships, and nurture perspectives.
The closer we look at content receptivity, the more the need to better understand our customers’ needs and align our messaging to those needs becomes clear. AI provides the potential to view interactions across dimensions to foster specific needs at scale. Advanced behavioral micro-segmentation allows for that better understanding of who our content consumer is, and which factor of that user’s record drives immediate action, which indicates receptivity, and which indicates needs for alternative approaches.
Using AI to Supplement Content
In content marketing, the currency is available content with which to drive awareness, evaluation/consideration, trial, and use. Already AI has been used to create content, whether supplementing news feeds, reacting to social media content, or summarizing vast amounts of information. Rather than using AI to simply generate more noise (which would then be filtered out!), the right use of AI content creation should leverage the micro-segments described above to create just the right tone in that conversation.
AI provides the ability to rapidly ideate and create content, especially digital, with opportunities for online engagement and opens the door for additional “earned” traffic, receiving inbound referrals from organic search, social sharing, and site linking. Additionally, with rapid content prototyping and variation testing, we have the ability to discover optimal configurations that may resonate to broader untargeted communications.
As healthcare communications is a highly regulated space, the ability to both introduce variations while implementing tight controls makes for an opportunity for which AI is uniquely positioned. Working with internal regulators to understand presentation and context, and having the reassurance that those rules drive content, can guide internal conversations and hopefully bring speed to market. Trusted content sources to train models exist organizationally, from support conversations, medical science liaisons, testimonial transcriptions, and sales training materials. AI content automation provides that ability to integrate and tune those messages for balance and resonance.
Utilizing Micro-predictions to Infuse Insight to Optimization
Marketing data are available across the majority of channels we utilize—and we hope to channel insights stemming from measurement to drive the optimizations that we take next. What happens when we are able to utilize predictive modeling to understand what will happen if I make this change before I make the investment? Employing predictions at a micro-subchannel level can empower organizations considering integrating data science methodologies to take that next-best action confidently.
If you’ve made the investment in the channel, you’re likely making the investment in measurement. This takes many forms, such as data extracts, analyst presentations, and on-demand dashboards. Likely you’re receiving data (and hopefully insights) back in response to the actions you are taking. You likely already have an idea of what you’re hoping to see reflected in the market, be it a raise in awareness, increased conversions/sales, higher velocity in switch, etc. Measurement strategy is critical, allowing for identification of key actions and key performance indicators, so we know what is and isn’t working.
As we move forward, we’re going to ask you to look at the data you receive across three areas: Audience Data, Channel Data, and Message Data. This arrangement will form the basis for where we go moving forward—and has formed how we’ve shaped our own data science practice. How you begin transitioning from insight to action varies by situation; some organizations are ready to make macro-level updates in budget and communication in near real-time, while others will need more planning in their approach. This is where we introduce predictive analytics to our mix, looking for ways we can begin to model what we might do—and what we expect to happen next. This scenario planning (running several what-if scenarios based on available data) ranges from broad scope investment questions to granular execution questions. Rather than introducing external/additional forces, our desire is to improve our performance—to game-plan with the pieces we already have in play.
High-value action-planning leads us to this combination of ongoing optimization-centric measurement and predictive scenario-planning with existing elements: The use of micro-predictions to not only deliver insight-based optimizations, but play out those optimizations using data science to right-size before investment. If our in-field reps concentrate only on efficacy, will anyone find our access messaging? If we increase our SEO sessions by 10%, what does that mean for display spend? If we move our savings card form to its own page, how will that affect activations?
Making the Shift in Engagement Approach
Utilizing micro-predictions to empower brand-level analysts will require some shifts in how we approach engagement today—in the form of process and tools:
(1) Initial modeling: Connect analysts with data science partners at the onset to create a “Day 0” model. This will have the up-to-date in-market plan and can be used as a starting point.
(2) Optimization scope: Look for opportunities for optimization within the existing channels/customers/messages. Introducing new elements can be done, but there will be limited historic data to utilize for predictions.
(3) Oversight: A clear accounting of data throughout the campaign is required to allocate value. Additionally, the right platforms are needed to enable rapid perspective.
(4) Data storage: Lake or warehouse, pivot table or CSV—there will need to be an ongoing record of interactions and associated data from which to make observations and model future outcomes.
(5) Data feeds: While data can be uploaded and modeled manually, whenever possible, utilize application program interface (API) feeds to increase scenario speed to market.
Our audience’s trust will always be precious. How we choose to engage with them must exchange that trust for the valued experiences our brands have to offer. The right application of AI, balanced with the right context and tone, has the potential to deliver an experience that benefits the healthcare marketer and healthcare consumer alike.
*Special thanks to Jacob Thomas and Drew Glenn from Syneos Health Data Science for their contributions to this article!
