What does the $250 M Series A for Automation Anywhere mean for Enterprise Start-ups ?

In the last week, my twitter feed has been filled with stories about how Automation Anywhere (AA) has raised a $250 M Series A to “automate mundane jobs” and “make business processes self-running”. The articles are filled with buzzwords like “AI” and “bots”. In a press release, Mihir Shukla, the Automation Anywhere CEO claims that their bots can automate 80% of an enterprise’s business processes. Anyone reading these articles would think that “the machines are taking over” and AA is on the verge of building general AI.

Automation Anywhere is an impressive company with over $100 M in revenues, 1000+ customers and 100%+ YoY growth. At the same time, its product is relatively simple and involves almost no machine learning today. Its success is validation of the opportunity that exists for enterprise chatbots and “business process elimination” more broadly.

The proliferation of data, SaaS apps and custom code has created massive fragmentation in most enterprises. A sales rep may have to access 3–4 apps and continuously cut and paste data between them. In HR and Finance, the fragmentation is even worse, with employees having to toggle between a dozen applications in a single day. A few decades ago, SAP popularized the concept of building a system of record. Now every application aspires to be a system of record, and as a result, there is no notion of a unified customer or employee record.

The proliferation of data and apps is not stopping anytime soon. However, ML offers a new approach to addressing these issues. I believe that there are 3 broad opportunities for ML start-ups focused on the enterprise that want to address this issue”

  1. Create “meta system of records”: Convincing half a dozen fiefdoms within an enterprise to work together and create common system of record is a lost cause. Instead start-ups like Quantic Mind and Eightfold are creating “meta systems of record” for customer and employee data. They suck in data from multiple enterprise systems, de-dupe and unify the data, apply machine learning to the data to extract insights, and expose the data to other enterprise systems (including some apps that they have developed). While these companies create a unified customer/employee record and therefore effectively a system of record, they don’t attempt to replace existing systems of record.
  2. Chatbots: The interface for most enterprise apps is horrendous. Salesforce is the canonical example — its simply unusable! In addition, apps like ServiceNow often lack context and force you to re-enter the same data multiple times. Messaging is already the dominant consumer interface. Chatbots can be context aware, abstract away the complexity of existing software and automate basic responses.
  3. Process Elimination vs. Process Automation: Automation Anywhere reduces the need to cut and paste data from one screen to another screen. Tools like MuleSoft and Tray.io attack the same problem by creating connectors and rules-based work flows on top of existing APIs. The future is about eliminating human data entry by using NLP to extract the data from ambient signals. For example, start-ups are starting to pre-populate CRM fields by extracting information from emails and phone calls.

For me, the Automation Anywhere financing is a reminder of the scale of opportunity to improve the efficiency of Fortune 500 companies. In the last 50 years, traditional enterprise software has increased productivity 2–3 X and enabled companies to become global corporations/leverage global talent. In the next 50 years, ML based enterprise software has the potential to increase productivity tenfold. If you are excited about building such a company, give me a shout!