Can sector snapshots, market studies, freedom to operate searches, business plans, competition analysis etc. be automated and reports generated on demand?
Below is a healthcare sector snapshot that was auto-generated reports via code and using various AI techniques. Before glancing over it, here are some notes:
The report is a series of snapshots to illustrate the possibility and is not intended to be comprehensive.
Ignore the lack of aesthetic aspects as UX was not the focus of this report.
The focus was given on auto-generation and hence human editing has been avoided.
The only human intervention was in inserting takeaways to point out trends etc. (these could be automated too)
The data is focused on startups in India and for a small timeline of Dec 2017 to April 2018.
Patent analysis is extended to 5 years.
Slide 2: What are people following in the startup ecosystem?
Slide 3 & 4: Which sectors raised money? Where does healthcare stack-up? Within healthcare which startups raised funding?
Slide 5 & 6: Expand understanding of the startups that received funding and classify them
Slide 7: Organically come up with a new way to classify & cluster startups. This provides a great richness to understand startups and relevant competition
Slide 8: Conduct a technology deep dive on each startup. An example here — understand the patent landscape
Such a report can be generated on-demand literally at that moment. The report will represent the state of the sector at that moment and can be generated multiple times. Slide generation is automated using Python libraries like “python-pptx”. This can also be used for adding logos and improving UX visual appeal.
The above report demonstrates various aspects. For example, company analysis, VC money flow, patent analysis, etc. This is just the beginning. A lot more can be added like earning transcript analysis, investor decks, founder analysis, peer-reviewed publication analytics, company image analysis etc
The cost of building this is minimal. Upfront R&D is to figure out the data, cleaning it and writing algorithms. The operational cost is primarily storage costs of data and any 3rd party API call charges. At such low cost of report generation, new business models like micro-subscription models can be opened up thereby disrupting consulting, media, publishing & report generation industries.
The above report was achieved through a mash-up of technologies as can be seen in the picture below.
The opportunity is to structure various reports at different depths of detail. To illustrate this, find below 2 charts focused on AI and in specific Deep Learning. This is done by converting patents into a multi-dimensional vector space using AI and clustering them. One can go to greater depths with specific domain knowledge-based extraction like analyzing what neural network architectures are being used in the patents being filed using deep learning. Between the two charts, one can see there is a trend towards understanding deep learning neural models from the brain, applying on speech & image data-sets, and also analyzing anomalies in telecom network packets.
Slide 9 & 10: Understand upcoming technologies and architectural trends. In this case for Deep Learning
Accuracy can be improved with more training and better vocabulary building. Human editing can take it to great levels. Augmenting humans to write reports I believe is a promising area that I tend to call Augmented Reporting. But what is keeping me excited is the picture below. Adding an AI layer that can take contextual input and intelligently generate the schema of reports and produce them. Multiple layers of intertwined AI like our brain. That’s the future of AI and the one that will penetrate the higher end strategic tasks.