AI isn’t science fiction or a future technology we’re waiting to adopt. It is, right now, affecting every aspect of our daily lives, and that includes how we develop applications, products, and services.

Every few years, there’s a new buzzword technology that drives mass hype as it promises to disrupt the status quo: software, mobile, IoT, 3D printing, virtual reality, blockchain. In 2016, every company desperately wanted to latch on to artificial intelligence (AI).

So while the earliest innovators (think Alan Turing) were studying how computers could mimic humans in the 1950s, we just recently witnessed a hype cycle triggered by the potential for AI to cause the next generational shift in computing. Everyone claimed to use AI in their marketing materials, to the point where investors and consumers alike became disillusioned, and interest waned.

But during this AI winter, tech companies continued to pour investment into AI, which led to an inflection point and brings us to where we are today. AI is a household name that’s transforming businesses and working behind the scenes in consumers’ everyday lives. To a large extent, AI is the new pseudonym for automation, as app development is for software development.

As a result of this hype, AI became an umbrella term that is rarely comprehended by the mainstream and applied differently by businesses, depending on the goals of the company or industry employing it. AI encompasses everything from algorithmic automation to learning and cognitive systems; the meaning is that broad. When it comes to the futuristic, autonomous robotics that consumers’ minds automatically jump to when they hear AI, we’re not there yet. Sophia, the AI robot, is potentially no more than a sophisticated chatbot that runs on a decision tree.

To break down the current AI landscape and examine its practical applications, let’s simplify to two buckets: AI as a service companies (or companies that offer AI as a service) and companies where AI powers the core service delivery. As a potential customer, it is vital to understand where the companies sit in this landscape as they will have a fundamentally different approach to product development. Conversely, as a customer, you need to make sure you are adopting these technologies at the right time.

AI (as a service) companies are providing their fully baked AI solutions to help their customers with infrastructure, analytics, machine learning, and more. AI, and in particular, learning, is the focal point of their value proposition as they are helping their customers automate intelligence to support their digital transformations. Google Cloud AI, for example, is providing infrastructure and AI tools to enterprise customers. Nvidia and SambaNova are focused on helping companies deploy AI at scale.

These companies need to have established core models, data sets, and must be fairly advanced in the development of their technology. Ideally, as a customer who doesn’t want to be part of an early trial or data set, you want to be engaging this when the companies have either trained their systems with external data or are established with core data from existing customers (i.e., computer vision companies telling you that your photo has a dog in it because they have seen hundreds of thousands of photos with dogs in it). These companies usually over-index on their product development because the product is the manifestation of the model.

When it comes to companies with services powered by AI, they’re using AI under the hood to automate part or all of their otherwise human process to scale and meet the needs of their customers. For example, Engineer.ai leverages human-assisted AI to build tailor-made software using the collective knowledge of what has been made before and assembling the best team of developers and designers to build it. Today, on average, 60 percent of the software they produce is using common building blocks, and 40 percent is custom-built by humans.

It’s important to note that fully autonomous development, while in some cases the end goal, is still at least a decade away. In the same way, Zendesk, a customer service and engagement platform, is using AI to automate its ticketing process to deliver the best customer service at scale. Its product roadmaps undoubtedly include more automation as it looks to scale its capacity without scaling costs or skimping on the end products it delivers to customers.

These companies often take a very different tack to product development; intrinsically because they are using the AI to help them scale; they need the scale to help them decide what to build. “When we think about what we are building; how we are thinking of deploying AI, automation vs. intelligent systems, we have modeled how we need to scale inorganically thus avoiding the pitfalls of exponential increase in human operations cost,” explains Sachin Dev Duggal, CEO of Engineer.ai, when describing how they are close to 50 percent now of its automation journey.

In the case of Zendesk and Engineer.ai, customers here must ensure that delivery of the service meets their requirements and they are less concerned with explicitly how that company delivers the service. After all, they are getting a custom software project or a quick response to a customer – whether a monkey does it, algorithm or cognitive learning shouldn’t matter as long as the value proposition (faster, better, easier) is delivered.

Contrary to AI as a service companies, these types of organizations are delivering parts of a fully built-out model much earlier in the process, and any overheads associated with it are de facto costs of product development. One could argue, companies that over-index on building out the intelligent platform will run the risk of failed delivery because, quite frankly, the data sets are so vast and varied.