Year 2019, yet another success for the Artificial Intelligence (AI) startups. Indeed there is increasing interest in AI acquisitions, and new annual high for money raised by AI startups (estimated at USD 18.5b+, with USD 3.7b being raised in just Q4 2019)… this is despite overall economics hurdle and slow down of the M&A activity globally.
The UAE is the first country in the world to establish a Ministry of Artificial Intelligence and one of the first to come up with national AI strategy. UAE is committed to attract investments into AI in multiple sectors including public sector, transportation, healthcare, renewable energy, and education.
According to Crunchbase, there are approximately 100 AI startups based in UAE (almost half of which are in Dubai), with top 10 accounting for the USD 20m of funds raised within the past few years. These represent deals that received some publicity, obviously figure would appear higher should we account for angel investors and private equity that kept low. The most successful AI startup in terms of fundraising is happened to be Crowd Analyzer, which secured USD 3.5m investment in Q4 2019.
So, what are the things potential investor or buyer may would want to know about AI startup?
Number #1. Obviously financial performance and market potential. I discuss, some of those in nontraditional (Digital) sense in one of my recent articles here.
Number #2. Technology aspects of the Artificial Intelligence.
AI startups predominantly leverage technology that can be attributed to domains of (1) data crunching (2) computer vision (3) natural language recognition.
For example, imagine startup that does sentiment analysis of the customers in social media. Such case would require to develop model trained on data. Data eventually would be collected from various sources (in different forms and shapes) and would require to be of certain quality. As a next step solution would leverage natural language recognition capability to derive insight (such as type of sentiment). Data collection and analysis require algorithms to be able to parse, normalize, categorize and contextualize the text before it can be analyzed for features and converted into insight.
It is a complex thing, but I invite you to look into the aspects of data and natural language recognition models that can provide some degree of certainty that investors or acquirers may decide to examine.
It might become apparent that success of the AI technology implementation is in the way data is threated, including aspects such as data ingestion, storage, and enrichment. And one may want to look at quality of the input data, data storage and ingestion pipeline. This is super critical as quality of data determines success of the solution.
Actual analysis of data (Natural language recognition in our case) is based on Machine Learning algorithms and models. Despite how models being developed there are some business metrics that can be applied to assess its performance. This may include comparison of the algorithms and models against top research in this area and possible baseline models.
One of the indicators of the overall performance might be defined as combination of accuracy and precision. Another important aspect might be ability of the solution to overcome statistical barrier. For example, degree of statistical bias and variance of the model may provide an indication for further development. One would ideally want to have a model with relatively low bias and variance level. This is where many ML projects go into iterative development and look for additional training data.
In my view, both sides of the AI (potential investors/buyers and startups) have to prepare for success and should assess aspects going beyond compelling business case and management commitments. Technology aspects appear to be often underestimated and should be considered by potential investors and buyers as they may have detrimental impact to the overall success of the business case.