For a startup that doesn’t have a large number of customers or adequate data sets, it is very difficult to leverage machine learning as they don’t have enough data. iStock
For a startup that doesn’t have a large number of customers or adequate data sets, it is very difficult to leverage machine learning as they don’t have enough data. iStock

Startups love flaunting artificial intelligence but the real challenge is scaling up

  • The absence of software development ecosystem for AI and regulations around it makes it difficult for entrepreneurs to launch the product they want
  • While many startups are able to develop a viable software, there are many trying to sell a half-baked product

New Delhi: The interest in artificial intelligence (AI) and machine learning (ML) has grown exponentially in the last couple of years. It is no surprise that every company pitch made before investors revolve around AI.

However, there are many trying to cash in on the hype around AI. A March 2019 survey by a London venture capital firm, MMC, indicates that 40% of 2,830 AI start-ups surveyed in Europe don’t actually use AI. The companies using AI in their pitch are likely to get 15-50% more funding. Most of the AI deployments were in chatbots (26%) and fraud detection solutions (21%).

According to the Q4 2018, MoneyTree Report by PwC, investment in AI startups has grown significantly in the last few years. AI startups in the US alone raised a record $9.33 billion in 2018, which is almost 10% of total venture capital (VC) investments in 2017.

“Most investors don’t completely understand the core technology involved. A simple kind of automation is passed as AI and with their limited understanding they end up believing so. Besides, investors are not looking at technology but the business potential of these startups," said Sachin Dev Duggal, founder and chief executive officer, Engineer.ai. AI or no AI, if a startup has a clear course of action, provides value for customers and stakeholders, and looks promising, investors will continue to increase their investments across the startup ecosystem, Duggal.

Mahesh Makhija, partner and leader, emerging technology, EY India, pointed out, “a platform that is automating a significant part of the machine learning toolkit, will be probed by investors as they would want to understand the underlying technology. They will want to figure out if the platform is able to prepare data, run a range of AI models and choose the right one." Here investors would want to find out what’s the underlying technology: Are they able to run different models? How do they differentiate between the efficiency of these models?

“However, if an investor is contemplating investment in a cab hailing or food delivery startup, which is using advanced data analytics, they are more focused on the parameters of the business as it is a given that advanced analytics / AI will be used as part of the solution," Makhija noted.

For investors in those kinds of startups, ultimately, the goal is to provide the customer with something that is relevant. There, the focus is less on the actual technology and more on the ability to run the business. Deploying an AI solution isn’t the same as it was two years ago. Open source platforms like TensorFlow and Intel’s Neural Stick, have allowed many developers to build and offer low-cost AI solutions. However, scaling them is the real challenge.

Software development isn’t easy as it seems. While many startups are able to develop something viable, there are many trying to sell a half-baked product. Duggal notes, the absence of software development ecosystem for AI and regulations around it also makes the job difficult for entrepreneurs to launch the product they want.

“Not every AI-based platform has been created the same. Various pieces of functionality have their own databases. For building or integrating AI into software products, prices can range from $40,000 to $1,50,000 depending on the complexity of software and the needs of the end customer," Duggal.

“To be able to really get power from machine learning solution is not very easy. One of the biggest problems is data. Not only a large but a good quality data set is required to make a difference. Big tech companies have large data sets. But for a startup that doesn’t have a large number of customers or adequate data it is more challenging," Makhija said.

For big firms like Google or Microsoft, there is so much data available that machine learning is a way of life. But for a startup that doesn’t have a large number of customers or adequate data sets, it is very difficult to leverage machine learning as they don’t have enough data. Duggal is of the opinion that long-term AI-powered businesses can expect to remain a dominant force within the software development industry when they actually fulfil their customers’ needs and remain customer-centric at their core.

So the next time you hear a startup it’s doing AI, it’s okay if you ask ‘Why’.

Close