BenevolentAI: Revolutionizing drug discovery using Artificial Intelligence

It takes over $2.5B and 10 years to produce one drug. Artificial Intelligence has the power to halve each of that. Enter, BenevolentAI.

Developing biopharmaceuticals (drugs) is expensive and time-consuming.  For example, developing a drug in 2010 took 10 years[i] and cost $2.6B[ii], 145% more than 10 years earlier.  This is because in the traditional drug discovery process, chemists sift through millions of compounds of which only a fraction are approved[iii] – for every 25,000 compounds that start in the laboratory, 25 are tested in humans, and only 5 make it to market.[iv]

Source: PhRMa’s 2015 “Biopharmaceutical Research & Development” PDF report

However, advances in deep learning and artificial intelligence (AI), availability of big data, and cheaper computing power are revolutionizing the drug discovery process.  They can make the cost of discovering drugs 50% cheaper[v] and reduce the time it takes from identifying a target disease to finding a molecule that acts against from 5.5 years to just a year.[vi]

A trailblazer in this space is BenevolentAI, valued at $1.7B within five years of launch.[vii] BenevolentAI uses AI to mine and analyze vast amounts already-published biomedical information, from clinical trials data to academic papers.  The company’s director, Jackie Hunter believes that “There’s a new paper out every 30 seconds. This is a huge amount of information that isn’t being used for discovery and development of new drugs.” [viii] Moreover, for researchers, assimilating this data and identifying meaningful hypotheses from it would be extremely challenging.  To solve that, an architectural innovation is necessary, and that is BenevolentAI’s value proposition – instead of trying to discover a new compound from scratch, BenevolentAI uses its AI system to try to find new drug candidates from existing information and new potential uses for existing drug candidates.

Thus, there are two ways in which BenevolentAI creates value.  One, it identifies molecules that have failed in clinical trials and predicts how those same compounds can instead be more efficient in targeting other diseases.  Second, it uses the predictive power of its AI algorithms to design new molecules, and extract new hypotheses based on over a billion relationships between genes, targets, diseases, proteins, and drugs.[ix]  Both these propositions make the drug discovery process more cost and time-efficient, creating significant value for the biopharmaceutical industry and the healthcare ecosystem.

Not only is BenevolentAI creating value, but it is also capturing some of it.  First, BenevolentAI has 24 drug candidates in just four years – significantly faster than traditional drug discovery.  Second, BenevolentAI signed an $800 million deal in 2014 to hand over two Alzheimer drug targets to an unnamed US company for development and will take a cut from the profits if its drugs are developed and sold.[x]  Finally, last year BenevolentAI reached a deal to license the right to develop, manufacture and commercialize a number of novel clinical stage drug candidates from J&J.[xi]

BenevolentAI’s ability to create and capture value, as described above, will continue to grow rapidly with better deep learning algorithms, robust bioinformatics databases, and higher computing power.  Consequently, it (and its peers) may altogether replace the drug discovery teams of traditional biopharma companies, as has been foreshadowed by several collaborations in 2016-17 between AI-based drug discovery startups such as Exscientia, Numerate, and Insilico Medicine, and traditional biopharma companies such as Pfizer, Genentech, and GSK.[xii]

That is what makes BenevolentAI a winner.



[i] Tufts CSDD, “Briefing: Cost of Developing a New Drug,” Nov_18,_2014..pdf, Nov 2014.

[ii] Tufts Center for the Study of Drug Development (CSDD), November 2014,, accessed 28 Jan 2018

[iii] Drug Discovery AI to Scour a Universe of Molecules for Wonder Drugs, Jason Dorrier, Nov 19, 2017,, Singularity Hub, accessed 28 Jan 2018

[iv] Drug development: the journey of a medicine from lab to shelf, Ingrid Torjesen, May 12, 2015,, The Pharmaceutical Journal, accessed 28 Jan 2018

[v] How AI Is Transforming Drug Creation, Daniela Hernandez, Jun 25, 2017,, Wall Street Journal, accessed 28 Jan 2018

[vi] Big pharma turns to AI to speed drug discovery, GSK signs deal, Reuters Staff, Jul 3, 2017,, Market News, accessed 28 Jan 2018

[vii] London-Based AI Startup on Hiring Spree Amid U.K. Boom, Giles Turner, Jun 14, 2017,, Bloomberg Technology, accessed 28 Jan 2018

[viii] This AI unicorn is disrupting the pharma industry in a big way, Joao Medeiros, 31 Aug 2017,, WIRED, accessed 27 Jan 2018

[ix] This AI unicorn is disrupting the pharma industry in a big way, Joao Medeiros, 31 Aug 2017,, WIRED, accessed 27 Jan 2018

[x] A British tech unicorn is trying to cure Alzheimer’s and ALS with artificial intelligence, Shona Ghosh, Apr 24, 2017,, Business Insider, accessed 28 Jan 2018

[xi] Machine Learning Drug Discovery Applications – Pfizer, Roche, GSK, and More, Jon Walker, Jan 11, 2018,, techemergence, accessed 28 Jan 2018

[xii] Machine Learning Drug Discovery Applications – Pfizer, Roche, GSK, and More, Jon Walker, Jan 11, 2018,, techemergence, accessed 28 Jan 2018



Amazon | To Go Or Not To Go


Fitbit: From Winner to Loser

Student comments on BenevolentAI: Revolutionizing drug discovery using Artificial Intelligence

  1. This is a really exciting development for the pharma industry and future drug discovery. I am curious to understand how BenevolentAI has a competitive advantage over the other drug discovery startups using AI mentioned at the end of your post. Is this something that many companies are developing and seeing the benefits of or is BenevolentAI using the tech more successfully than others?

    1. Thanks for your comment, Kat. BenevolentAI’s only source of competitive advantage as of now according to me is that they moved in early and they already have 24 drug candidates for testing, apart from partnerships with other pharma companies. Other companies like Insilico Medicine and Recursion also have drug candidates and pharma partnerships but BenevolentAI is the only unicorn with actual products (that are being tested). Who knows though, its peers may beat it eventually. But even if that happens, as consumers, we’d all still benefit!

  2. Thanks for this post, Saurav! This is a space that I didn’t know very much about and reading your post has made me quite hopeful about drug discovery and cures that might be found in the future using AI technology like this. In a similar vein to Kat’s question above, as I was reading this I found myself wondering in what ways it differed from IBM Watson and how the company is seeking to create value for consumers and to capture value for itself using a distinct approach from Watson’s as a start-up of five years is unlikely to be able to compete with the resources that IBM is placing behind Watson.

    I also found myself wondering about the human competitors’ reactions to this technology. Here I am wondering about the scientists in labs, researchers and doctors who are attempting to find the same cures and solutions but are not necessarily able to compute or discover things as quickly as BenevolentAI. Do they tend to focus on the Benevolent part of the company – that it is likely helping to make a difference in the world of medicine – or on the AI part – that if successful it might remove the need for their expertise?

    1. Interesting questions!

      So IBM Watson is brought up all the time as a comparison. While no one really knows whether it actually has better AI capabilities for drug discovery than Benevolent or others, what we do know is that IBM Watson’s main use cases have been big data analytics and image recognition in healthcare so far. Nothing about drug discovery just yet. I would imagine it could be trained to do the same work as Benevolent than others but perhaps the folks at IBM have other uses in mind.

      I think the answer lies in traditional scientists joining hands with computer scientists and doing the drug discovery work together. If each operates alone, faster drug discovery will not happen. Btw, this (joining hands) is indeed starting to happen at many large pharma companies now! Why else would big pharma be investing in these startups? 🙂

  3. Really interesting and great to hear about technology that has the potential to affect society in such a positive way. I’d be curious to know how all of these jumps in drug development affect regulators. Will the FDA be able to keep up as drugs are discovered and brought to market more quickly? Will the FDA itself be modified or will it become a bottleneck in innovation?

    1. Hi Rachel, thank you for bringing this up. Once the drugs are “discovered” by companies like BenevolentAI, they still need to be tested in clinical trials, after which the FDA will review them. So, the overall process will be shortened (and therefore, cheaper) but the FDA’s role will still remain the same.

  4. Great article Saurav.
    Along the same lines as other classmates’ comments, I think the million dollar question is as follows :
    Provided – as you described in your article – that competitive advantage lies in “better deep learning algorithms, robust bioinformatics databases, and higher computing power”, what is to say that deep-pocketed DeepMind (Google’s subsidiary focused on Machine Learning / AI that arguably posesses all three and that has recently taken a strong interest in the field of Heathcare and has already some firm ties with the industry in the UK for instance) will not simply overpower the small nimble startups such as BenevolentAI, thus submitting the drug discovery process to Big Tech’s supremacy (as other industries have before).

    1. Interesting point! So DeepMind will hopefully go on to do big things and really impact healthcare. If it does, it will be better for all of us. But if and when that happens, companies like BenevolentAI can still succeed side by side. There are many applications of AI / ML to healthcare and drug discovery is just one of them. I don’t think DeepMind has focused on drug discovery just yet but I hope it does because ultimately it will help the entire ecosystem, and eventually us consumers (patients). I would be very happy if Big Tech was able to disrupt healthcare and make it cheaper and better! But for now, BenevolentAI is focusing on a specific part of healthcare and has made great strides, whereas DeepMind is still searching for specific problems to solve.

      BTW, DeepMind was acquired for around 500-600M in 2016 and Benevolent AI is around $2B by now – so perhaps the latter is not so small and nimble compared to it 😉

  5. Thanks for sharing, Saurav! Given our discussion today regarding GE’s business model for capturing value via the Industrial Internet, I’m wondering if you have an opinion on how Benevolent.AI is monetizing its drug discovery algorithms. Clearly, value is being created, as evidenced by the $800M deal Benevolent.AI signed. I’m wondering whether or not the company is capturing this value in an optimal way. $800M for two drugs seems like a lot of money up front, potentially increasing the sales and marketing effort required to close deals. If Beneveolent.AI were to reduce the upfront price and increase the portion of income they receive from revenue sharing, it could reduce sales / marketing friction and project confidence in their own algorithms. I could not find news of any new deals closed that would indicate Benevolent.AI is experimenting with its business model, but it will be interesting to see how the model changes as the company grows.

    1. You’re right, Carl, but it is still early days. The 24 drug candidates that BenevolentAI has apparently developed will first need to be clinically tested and after the trials are done their efficacy will need to be checked. The jury is still out on whether it has established good economics. However, it has certainly created value and seems to be beginning to capture some of that through its partnerships with big pharma and the $800M deal etc. Time will tell as to whether it (and its peers) can capture more of the value that will hopefully be created in the field of drug discovery.

  6. This is a very interesting post! Thanks for sharing. I worked in J&J and Genentech before, and I understand how difficult and risky the bio-pharmaceuticals R&D is. Leveraging AI to accelerate the drug development will create a huge value for the whole society. Do you have any thoughts of the future of such AI-based startups? It seems that many biotech/pharma companies are cooperating with these startups to try the new technology. Do you think these startups will be eventually acquired by large biotech companies, or do you think some of them should leverage its core AI capability to sell drugs by themselves and capture all the value (becoming another Amgen or Genentech)?

    1. That’s the million dollar question, Ting. BSSE theories suggest that ideally these startups should someday be able to produce drugs themselves. Having said that, I personally believe that they will either give drug discovery services to big pharma / biotech or be bought by them. Either way, I hope the technology develops and better and cheaper drugs are produced!

  7. Interesting post – thanks Saurav! I also think BenevolentAI can be the winner here because it’s an early entrant and AI works better as it progresses (given that it builds on accumulated knowledge). However, it seems that BenevolentAI is a winner in incremental drug improvements but can it make more groundbreaking discoveries? Will the additional revenue from incremental drug improvements be enough to keep pharma companies competitive. If not, then biopharmaceutical companies might begin to see negative consequences from this investment if other companies are creating the “blockbuster” drugs.

    1. You raise an interesting point, Sairah. My assumption is that the indications for which these drug discovery companies will all be key areas for biopharma and biotech – cancer, cardiology, MS, ALS, Alzheimer’s etc – because these are the illnesses for which ‘blockbuster’ molecules have the highest probability of being created (market size, demand, lack of alternatives etc.). Therefore, the incentives are aligned. The only question is when will a drug get launched in the market, whose origins were in silico and not in vivo.

Leave a comment