• Sustainability
  • DE&I
  • Pandemic
  • Finance
  • Legal
  • Technology
  • Regulatory
  • Global
  • Pricing
  • Strategy
  • R&D/Clinical Trials
  • Opinion
  • Executive Roundtable
  • Sales & Marketing
  • Executive Profiles
  • Leadership
  • Market Access
  • Patient Engagement
  • Supply Chain
  • Industry Trends

Regulatory Process Innovation & The Dangers of Over-Promising with AI


AI has huge potential, but the key to ROI is in the application, writes Alan White.

Never previously at the forefront of technology innovation, the life sciences industry now accepts that it needs to transform its agility, responsiveness and cost-efficiency, and that advanced digital tools are likely to form an important part of that strategy. Yet more modest-sized pharma companies and smaller biotech firms, who need an edge every bit as much as their larger peers, need to take a guarded approach to overhyped, generic AI solutions and vendors and service providers who promise the earth as a result of their use of the latest technology. AI has huge potential, certainly, but the key to ROI is in the application. Alan White seeks out the practical-use cases.

By 2022, the pharma and biotech industries will have moved through three phases of evolution – first codifying and standardizing processes, then automating them, and ultimately deploying artificial intelligence and machine learning to further accelerate pace and productivity. 

These are among Deloitte’s predictions for life sciences, as it speculates about the future for drugs companies and their operations.1 Specifically, it claims automated writing of clinical study reports will be happening as standard, using natural language processing - industrializing the conversion of structured study data into text narratives. Meanwhile leading firms will have automated up to 95 per cent of regulatory filing, saving up to a year in their launch cycles.

Alan White

Such claims and forecasts are commonplace, and industry thought leadership and event seminar programmed are replete with speculation about the marvels and promises of AI. For cutting to the chase and getting better products to market faster and more cheaply, for instance - just as AI will hone early, complex diagnoses, enable more personalized treatment programmed, and improve patient outcomes. 

Without doubt, some of these predictions will come true in time - to a greater or lesser extent. But, for now, many of these promises are ambitious, outlandish and premature. Although the life sciences industry could certainly benefit from a digital overhaul of long-winded legacy processes, and greater dynamism in managing and honing product lifecycles, true shortcuts to tangible results are the exception rather than the rule today. 

Minding the hype cycle

Where manufacturers are outsourcing routine processes to external solutions and services companies, and being promised the earth in terms of smart automation, they must be clear about what exactly they are being promised and what measurable, visible impact any proposed new innovation will have on their operations. This is particularly the case for more mid-range pharma companies and smaller biotech firms, which may not benefit to the same degree from automated process efficiencies as Big Pharma, because their needs are not of the same scale.

The trouble with “futurologists” and visionaries – those invited to inspire audiences at conferences – is that they are very good at painting big pictures and firing people’s imaginations, but often guilty of setting unrealistic expectations of how quickly such improvements will materialize. It is all very well to give companies a dream to plan and budget towards, but procurement managers and department heads need to be able to demonstrate improved value and efficiency in the here and now from the solutions and services they buy into.

The good news is that technology-enabled transformation can happen on a much more modest and focused scale, and still have a big impact – today. Indeed, the more focused and specific the target use case and its mapping to a known “pain point,” the greater the chance at making a significant difference in a reasonable timeframe, and without major disruption to the status quo.

Accelerating adverse-event reporting

Take pharmacovigilance (PV) and the role drug companies’ sales people are supposed to play in reporting back any adverse reactions experienced by patients linked to any of their organization’s products (for instance, if such information is relayed during face-to-face or phone-based client meetings). If this reporting task is left to chance, happens manually and/or (because sales people are human) left until some later point, the quality and value of the sales agent’s input is likely to be relatively poor. They may scribble some notes for someone else to transcribe later, and/or forget to capture the fuller details that are needed for a complete PV report. They know they have a responsibility to pass on this feedback, but for a busy on-the-road sales rep this just isn’t a priority.

But what if a simple yet clever software tool could make light work of PV reporting for those frontline teams? What if they could simply input and dictate all the required details straight into a secure mobile app, then move on? This would alleviate pressure on the sales rep, who has other more pressing tasks to attend to. It would also save on painstaking follow-up work by PV/safety teams and contract service providers, who ordinarily would have to try to verify any ambiguity and fill any gaps after the event - and at a point when it might be difficult to track down the clinician or pharmacist with the original case notes.

The intelligence in a software solution like this could be in the smart workflow, prompts and auto-filling of information fields, and the ability to link voice notes to a file containing additional case data. It would also be in the tool’s ability to capture some of this data in a structured way as part of the recording process.

Small changes, big impact

It is in specific applications of advanced software tools that business process service providers can really add value for their life sciences clients, especially for smaller-scale operations whose internal resources are overstretched – that is, where teams can’t spare people’s time for manual form-filling and case follow-up.

When manufacturers are looking for support from service partners, and becoming enticed by talk of AI and process automation and the promised impact on efficiency, then, it is important that they are able to ground this in tangible everyday experiences - and are able to understand the specific ways any new innovation will make to their own workloads and resource use. If life sciences firms can have direct input into process improvements and where intelligent automation will be most useful, so much the better. Realistically, the most transformative advances through technology are going to be those that are a result of client-supplier collaboration, and proactive brainstorming about “better ways of doing things.”

As a final note of caution, companies need to be careful they are not carried away by overly ambitious expectations, especially where the proposed application of AI is on a grander scale. 

This can happen when firms fall under the spell of hyper-scale cloud-based analytics services, where they could potentially start to slice and dice of data in all sorts of new and exciting ways to identify patterns and insights they might never have spotted otherwise. But, without the right controls and quality/accuracy safeguards in place, firms might never really be able to fully rely on the findings, and deliver practical value from them. Alternatively, they may incur all sorts of additional work to get to that point, which ends up undermining the business case.

From a PV perspective, even tools which “automatically” read, extract and interpret text from documents, and put them into context - ostensibly to save a team of people from having to do it -could potentially create more work than it saves, or certainly at this still-early point along AI’s maturity curve. And of course, doing powerful things with data all starts from an assumption that the source data is definitive and of robust quality…

What may be a better use of budget would be to use the technology to ensure that the right data is collected in the first place, as in the example above about transforming the way drug companies’ sales reps provide adverse events information back to PV and safety teams. That is, by removing reliance on handwritten notes, or expecting administrative teams to be able to determine which cases need the most urgent follow-up, according to the seriousness of the adverse event which in turn should dictate the response time. Intelligent automation and a traffic light system could be the answer here, driving up performance.

Ultimately, AI is only valuable to organizations if it transforms painful processes for the better, and there isn’t a one-size-fits-all template for that. In which case, life sciences manufacturers should be wary of bold promises that may not bear out in practice when applied to their own operations.  Instead they need to look at the relative size of their requirement and find a suitable partner with a solution that will clearly make life easier, not potentially more complicated.

Alan White is CEO of Arriello.

1. The Future Awakens: Life Sciences and Health Care Predictions 2022 (Deloitte, November 2017).

Recent Videos
Related Content