During the last decade, the explosion in enterprise data growth created a clear need for businesses to transition from Excel tables for data gathering, manipulation, and visualization to modern business intelligence (BI) systems. These systems fall under a category which we call “Visualization BI” or “VBI.” VBI systems are mostly focused on the same attributes as traditional BI systems, but in a much larger scale of data sources, computation, real-time presentation, and multi-tenant capabilities. We include both general purpose BI platforms and dedicated, vertically-focused SaaS offerings in this category.
The chart below provides a simple overview of a VBI system:
Chart 1: Typical Visualization BI (VBI) system
Currently, advanced VBI products can display the required manipulation of input data and summary charts. However, as data input and display requirements increase, so will the workload on analysts who are tasked with reading the VBI output for potential actions. This trend also increases the intellectual bar required from analysts and/or the potential areas for mistakes and inefficiencies. Sometimes, the insights created from the data and actions required to optimize business KPIs, are not physically possible for humans to process.
So – what’s new?
At Norwest, we are seeing the emergence of a new type of BI platform that we believe will augment and ultimately replace VBI. The motivation for this change is similar to the previous generation: exponential growth in data and inability to process with current visualization methodologies, coupled with human interpretation limitations.
As we all know, there is a growing capacity in harnessing artificial intelligence (AI) engineering know-how and deep learning models for practical applications. These technologies create a radical change which enables a new type of BI systems which we call “Insights and Actionable BI” or “ABI”. ABI systems include the same inputs as VBI systems use but they also include an AI module that either provide insights or make actions to optimize business objectives, or KPIs. Such systems match or supersede human analysts’ ability to understand data analytics, compare them with business objectives, and derive consistent and intelligent insights.
In different words, we are reaching an inflection point in which:
ABI >> VBI + Human Interpretation and Decision Making
Chart 2: Typical Actionable BI (ABI) system
We expect to see some of these new ABI systems tasked with providing insights, when business operators are still concerned about significant business risks from machine actionables. Otherwise, operators are already trusting such systems on high-volume, business-affecting decision making.
A simple analogy to this trend in BI, is the fast-evolving autonomous vehicle industry. For example, in the past (and present), cars gave drivers all the visualization tools required to make driving decisions. Today, advanced autonomous car systems provide road hazard insights. But, we all understand we are years away from full, level-5 autonomous cars in production, where deep learning will make all driving decisions based on a single KPI – fastest transit to a destination point.
These technology trends are not news to the venture world. The chart below shows the amount of venture capital directed towards artificial intelligence/deep learning. Of course, only part of this capital was directed to ABI.
Chart 3: Venture Capital investment in AI/ML in the US (Source: CB Insights)
Perhaps the best example of highly active ABI platforms in maturity is Algorithmic (Algo) trading. We have seen estimates that “fundamental discretionary traders” account for only about 10 to 15 percent of stock trading volume. This means that algorithms, optimized for different KPIs, are taking actionable decisions to buy and sell stocks. This phenomenon which was created in the last decade replaced the human broker who used VBI tools to determine stock purchase decisions.
The Future of ABI: Challenges and Opportunities
If one subscribes to the notion that VBI and analysts will be slowly replaced by deep learning powered ABI platforms, this trend creates new challenges and opportunities that should interest entrepreneurs, investors, and large corporations who can be affected by it:
ABI performance monitoring: Any computer program is subject to errors. When actionable insights are taken by analysts to improve business KPIs, they present why specific actions are required to their colleagues or managers. Most organizations will go through some chain of command approval process for such actions, depending on the potential impact of these actions to the business. Who will test the logic and decision making behind these actions when this human process will be replaced by an ABI platform? This is less of an issue in use cases where ABI is being used today, like Algo trading or online advertising, since business KPIs are typically one dimensional, easy to measure and simple to correct in real time. But what happens when such business KPIs span over days, weeks, or months and business KPIs are more complex to measure with multiple dimensions that affect each other? What technology can be used to measure the accuracy of the ABI system decisions?
Let’s consider an example – a future ABI platform is tasked with a business KPI to improve long term gross margins of a given hardware product. The ABI platform may determine that to achieve LONG TERM improvement in gross margins, the company needs to replace a few components that may degrade gross margins in the SHORT TERM. It may also require changes in the manufacturing line to accommodate these new components. Will humans allow such software to take the required actions, assuming they can be done electronically, or will they review the insights generated and use them to run a manual process? What if there are other methods to approach a similar goal, at a slightly lower performance but without a short term degradation and higher risk? How will operational risks factor into the ABI decision process? Finally – will someone build an independent system to test generic ABI system recommendations as a mean to hedge incorrect decisions risks?
Management excellence: Since the beginning of mankind, people with the most experience and management skills (among other traits) have become leaders. The same is true for business leadership. Management performance is measured based on shareholder value but the path to shareholder value is paved by operational excellence which is primarily acquired through years of experience and vision.
If operational efficiency will be conducted by ABI tools – what does it mean to future management teams? What function will they have in the presence of these tools that commoditize the many years of operational experience?
Let’s look at another example associated with the news publishing business. We are already aware of bots capable of capturing online news information and creating human-like articles. The main business KPI of a publisher is to increase traffic on the site. If an ABI tool will be able to direct bot-made articles to the right users and personalize the content for them, ?the function of the editorial ?team will turn obsolete. What does this mean to the future of publishing in general?
The tendency is to think that ML/AI will replace the work of low skills/low income employees. However, I question if ML and AI will also replace or commoditize the work of highly-paid, experienced managers.
Competition: ABI may change the way we look at competition. Let’s take the publishing ABI example above and say that there are three main publishers sharing one market. What will happen to their competitive landscape if all companies adopt the same publishing ABI tool?
First, you must consider timing – if one adopts the tool earlier than others, the market share split may change by the time all adopt it.
What will the tool do to affect ranking? I can imagine three potential and conflicting scenarios:
- Scenario 1:?Will the #1 publisher just get bigger and dominate the business as the ABI tool will have more information to keep this publisher at the #1 position? We see this trend already taking place in today’s economy where few large companies dominate businesses like advertising (Google, Facebook), retail (Amazon), Smart devices (Apple), etc.
- Scenario 2: Will such a tool actually even the market share since all publishers will ultimately optimize the articles and site layout in the same manner? Or,
- Scenario 3: Will such a tool actually drive the #1 publisher to remain the leader but will optimize the #2 to attract different, 2nd largest, audience to secure the #2 position etc.? i.e. the ABI tools will force market share stagnation by definition?
We can use existing ABI markets like Algo trading to predict how competition will emerge in the presence of ABI platforms – while each Algo trading company develops its own technology, it is becoming evident that the market is entering a maturity stage where almost all companies use similar deep learning techniques, making it difficult to improve performance. It will be interesting to witness what this technology maturity will mean to the market share among active competitors in the next few years.
Mutual influence: Most people think of ABI tools operating on a specific enterprise, in an isolated fashion. But when ABI tools will take actions, they are likely to interact with other ABI tools of affiliated enterprises, much ?similar to the way human operators interact today.
If we reuse the example above of the ABI tool trying to optimize gross margins, that tool will need to interact with potential ABI tools at the components and line manufacturing vendors to find the best margin optimization method. Such interactions will happen, making the tasks of performance monitoring — whether it will be done manually or by a smart BI tool — much more complex as it creates layers of inter-dependencies.
Security: The magnitude of damage that a security breach can create is well known based on recent events such as the complete destruction of nuclear facility centrifuge in Iran or massive levels of identity thefts.
As ABI tools gain insights and actionable powers, the damage a security breach can create will be tenfold of what we know today. This is another place where ABI performance monitoring tools will be required to determine wrong operations.
The Big Opportunities
I believe that unlike VBI, where the biggest winners had been generic platforms like Tableau and Qlik, the biggest winners in ABI will be sectorial, i.e. there will be very few leading vendors in each market, specializing in the nuances of each business line.
This is similar to the present SaaS marketplace where 2-3 top vendors are servicing each vertical in the enterprise like IT, service desk and HR. The winner(s) of each vertical has the potential to create huge outcome as they will become the next generation engines that power tomorrow’s highly efficient enterprises. Also, as hinted above, an ABI performance monitoring tool will have a significant value proposition, hence business opportunity, in the next generation enterprise operating sophisticated ABI tools.
Also published on Medium.