In any industry, technology, or application area, change is often caused by a select number of trendsetters. We conducted 17 interviews with the intention of taking a closer look at the adoption of artificial intelligence in various industries and sectors.
We set out to answer questions that are critical to business leaders considering AI’s value today, namely:
- How AI is being adopted across industry?
- What’s holding adoption back?
- What are the timelines of adoption in different sectors?
In order to help readers find what they’re looking for, we’ve broken this article down into three major sections, each of which aims to tackle the questions above. In the article below, we lay out the main findings from our interviews, coupled with relevant quotes directly from AI executives and researchers. We did our best to build this article entirely around the insights from our interviewees.
A visual representation of the phases of innovation adoption (source)
General Adoption Trends Across Industry
No matter the size, application area, or budget of our interviewed executive’s companies, there exists a select number of commonalities about general adoption trends. These trends were not exclusive to any industry, technology type, or business function.
Rather, these trends were observations about artificial intelligence adoption as a whole.
1 - Underestimating Resources / Staff Needed
In our October 2016 research on machine learning misconceptions, 30 artificial intelligence researchers were asked “What do you believe to be the biggest misconception that executives and businesspeople have in applying machine learning to business opportunities?” 30% of the respondents noted that underestimating the resources and staff needed was the biggest misconception. This idea is mirrored by Securinox Chief Scientist Igor Baikalov who notes, “Everybody wants to do it … because there is a serious shortage of talent in this area and attempts to close [the talent gap] have been somewhat difficult.”
2 - Need for Proper Data
Another common theme amongst AI executives and researchers alike is a distinct need for proper data, both in volume and quality. This idea is reiterated by a number of our interviewed guests, demonstrated below:
“The key thing to understand about this space is that you’re accessing sensitive data. You need volumes and volumes of data to get these things working, and that’s the frustrating thing.” - Alan O'Herlihy, Founder and CEO, Everseen
“We’re able to get machine learning working on the job. What I’ve come to find out is that you see other technologies only working very large environments that require training in massive data sets.”- Nicole Eagan, CEO, Darktrace
“Most Fortune 500 companies are dabbling in (IT tools performance, corporate real estate churn, diversity performance, etc) at some level. I would say for the next 2-3 years it’s going to be by and large fortune 500 companies / companies with at least 1,000 employees -- the application is eventually going to be there for SMBs, but it is really going to take time for the technology to mature to a point where frankly you have enough data to be valuable for those kinds of businesses. That is the challenge.” - Ben Waber, President and CEO, Humanyze
“More broadly we see a lot of these technology companies that have always been data driven, that have been born in the age of big data, they are starting to really adopt some of these sophisticated techniques, models, and big data approaches.” - Scott Clark, CoFounder and CEO, SigOpt
3 - A Focus of Specific Problem Solving and Customized Solutions
Business solutions are often aimed at finding a smart solution to a tough problem. These problems vary by industry, but the desired outcome of increased ROI remains. Artificial Intelligence is sometimes incorrectly thought of as a “plug and play” or “black box” solution, when in fact this is not the case. This focus on problem solving and customized solutions is mirrored by our esteemed guests below:
“People are looking at AI as more of an exotic thing and sometimes miss the understanding that AI is a way to solve a problem.” - Or Shani, Founder and CEO, Albert
"The industry understands that there are problems that are hard to solve, and they might need AI to solve it. Low margins means automating processes as possible is important." - Jay Perrett, CTO, Aria Networks
“To allow full autonomy you need to solve the whole problem. You need to the last .01% of cases. If you miss the traffic light or pedestrian in one out of a thousand cases, this is not good enough for full autonomy.” - Igal Raichelgauz, CEO, Cortica
“What we’ve built is an intelligence system that solves specific problems using disparate sources of data.” - Ben Waber, President and CTO, Humanyze
“Initially, our first clients are going to have to have a pretty big pocketbook, because we are going to charge them for all of the customization of our technology to what they’re doing. Initially, most of what we are doing is customization.” - Mark Ring, CoFounder and CEO, Cogitai
“What we have to do is leverage [retailers’] existing infrastructure (camera and server processing). We put that together and rebuild it -- it’s very customized.” - Alan O'Herlihy, Founder and CEO, Everseen
How are Technology Buyers Thinking?
Artificial intelligence is poised to transform the technology buying process, creating a more complex buying process and sales environment. Featured below is a selection of trends and insights gleaned from our interviewed AI executives.
1 - Hungry for Efficiency
Increasing efficiency has long been a strategy among successful businesses, be it operational, organizational, or commercial. AI has been particularly linked to efficiency, according to our interviewed experts below:
“Efficiency really starts to matter, especially in the AI and deep learning space, because as people start collecting more data and starting using more sophisticated and complex tools, it becomes more time consuming and expensive to try different things.” - Scott Clark, CoFounder and CEO, SigOpt
“Automation is often synonymous with efficiency (savings, getting things done on time) and that thinking solved a certain set of problems, but a whole bunch of other vexing cross industry problems remained." - Joy Dasgupta, Senior VP, Rage Frameworks Inc
2 - Bandwagon Effect
The “bandwagon effect” is a psychological phenomenon in which people do something primarily because other people are doing it, regardless of their own beliefs, which they may ignore or override (Investopedia). AI, according to our interviewed experts, has created a bandwagon effect in the digital space. These ideas are demonstrated below.
"Large law is where] you can see [AI] can make a lot of sense. You get adoption from the right couple of firms and everything else falls in line. They act as a group eventually.” - Gary Sangha, Founder, LitIQ
“There is obviously a huge bandwagon effect -- People are looking for the next wave of efficiency.” - Joy Dasgupta, Senior VP, Rage Frameworks Inc
3 - Transparency and Ease of Use
Ease of use and technology has gone hand-in-hand in the digital age. From swiping, to voice commands, quick and easy to use interfaces have been an integral part of successful technological innovations in recent years.
This idea is echoed by ExClone Founder Riza Berkan, who notes that “Any conversational AI that some company will adopt will depend on deployment ease and transparency. Without those two things it would be very difficult for these tools to flourish.”
4 - Challenges in Adoption
When considering the perspective of a technology buyer, skepticism is surely to become a topic of conversation at one point or another. According to some of our interviewed executives, there appear to be myriad of “hoops to jump through” regarding corporate acceptance. The specifics regarding the adoption challenges is outlined below:
“At the moment, only the early adopters are using [AI] in my opinion. Most of the industries are not there yet. Companies that are most technologically oriented and digital. Young companies like social gaming adopt [AI] much quicker. You can see an adoption rate in that vertical that is much higher [than other industries]." - Yohai Sabag, Chief Data Scientist, Optimove
"There is lots of skepticism about the ability to deliver because what we’ve seen... is lots of [companies that are presenting at conferences] put the best batch of [results] without anything to show for it.“ - Igor Baikalov, Chief Scientist, Securinox
“Adoption is the biggest challenge. Even with a crazy ROI you are still having to deal with organization structure, budgets, etc.” - Alan O'Herlihy, Founder and CEO, Everseen
What are the Adoption Timelines Across Industries?
Trends in artificial intelligence adoption depend on a plethora of considerations including budgets, data sensitivity, as well as business type. It is important to note that the insights below are specific to their particular industry or business function, and do not necessarily reflect the state of AI adoption at large.
An astute reader should note that executives of AI companies are likely to be optimistic about it’s possibilities in their sector, because (a) it behooves their sales goals, and (b) in order to found the company in the first place they likely became rather enamored with the possibilities of their technology - a kind of enthusiasm that often precedes a real “market” for a technology.
With those considerations in mind, below you’ll find what AI executives have to say about adoption in their sectors.
1 - Industries on the Cusp of Adoption
While some industries have appeared to be slower in adopting AI technologies, there are a number of other industries who are much closer to widespread adoption, either by choice or by market force. Below you’ll find quotes from AI executives who believe that their industry is on the cusp of adoption.
“I would say that within 2 years, 50% of public companies or more will have a form of artificial intelligence application deployed, either in bolt on to the existing system or a dedicated [contract discovery and analytics] system.” - Kevin Gidney, Founder and CTO, Seal Software Group
“People are much more comfortable with consuming data that has been machine filtered and we’re kind of taking advantage of that now.” - Peter Yared, Co-Founder and CTO, Sapho
“As we move forward there is a race between all the automotive manufacturers to build more advanced capabilities and gradually in the next 3-5 years expand these capabilities from safety to partial and full autonomy.” - Igal Raichelgauz, CEO, Cortica
"Within 5 years, almost everyone in telecom will be using it, they just might not be aware of it. It’ll solve the problems excel can’t. They’ll be using some software, service, or tool that integrates with AI, but they won’t necessarily think of it as AI. The networks NEED some degree of autonomy to maintain margin." - Jay Perrett, CTO, Aria Networks
“Retailers are very slow to move [towards AI]. I think it will take 3-5 years for the market [to force adoption]. - Alan O'Herlihy, Founder and CEO, Everseen
“Our customers are B2C companies because the interaction is more dynamic. We are more relevant when a company has a lot of interaction and communication between the company and the customer.” - Yohai Sabag, Chief Data Scientist, Optimove
“I would say that across all industries, those who are dealing directly with consumers have had to come to grips with [AI for the enterprise adoption], but the back office and middle office that is required to surface that experience, the revelation there hasn’t kept pace.” - Joy Dasgupta, Senior VP, Rage Frameworks Inc
2 - Finance and Baking are Trendsetters in AI Adoption
Finance, banking, and insurance were noted as industries ripe with AI adoption among our interviewed executives. This may have to do with the fact that finance is so dependent on data, or perhaps the fact that the data being stored is so valuable. Below is what our interviewed executives had to say on the matter.
“From a sector perspective, the banking sector attracts a lot of money, they are the most progressed, and it is also, from our perspective, the most competitive.” - Joy Dasgupta, Senior VP, Rage Frameworks Inc
“What we were seeing is that industries are more of a timeline than a focus. We started with finance, then insurance, healthcare, now retail / CPG … and it’s coming down the value chain.” - Matthieu Rauscher, VP, Yseop
“I would say financial services have the most hunger [for our services] as their models tend to be directly tied to money.” - Scott Clark, CoFounder and CEO, SigOpt
“2 years into the future, big brands and finance are the major players [in media monitoring].” - Miguel Martinez, CoFounder and Chief Data Scientist, Signal
Concluding Thoughts / Summary of Findings on AI Adoption
Although adoption timelines vary across industries, it is very likely that companies with the following four characteristics will have the best shot at making use of AI in the near term:
- Substantial budget for emerging technology projects that may not yield an immediate ROI. For the majority of AI use-cases, larger firms will often “flesh out” the genuinely valuable applications through trial-and-error (a process which requires a tremendous budget in a space like AI where talent is massively expensive). It is no surprise that smaller firms “look up” to see what their Fortune 500 counterparts are doing with AI - this is a fruitful exercise in determining what might be genuinely valuable. There will be cases of very well funded, very AI-talented small teams will build out genuinely useful applications, but the true test of these initial innovations will be their tests with the market - a job that’s likely to be done first by larger players.
- A large amount of organized data upon which to train AI systems. Industries that have been organizing and learning from data for decades (i.e. Banking) may have an easier time with this initial adoption than industries which are not driven by quantitative processes (i.e. Brick-and-mortar retail). Firms with data infrastructure and the talent to manage this infrastructure will have the best chance of making use of AI in the near term. In our recent research on the ROI of machine learning in business, 16 of our 35 polled AI researchers executives noted “Sufficient Data” as [a] “criterion needed for a company to drive maximal value from the application of machine learning in a business problem”.
- In house talent of data scientists, and AI experts. Firms with no AI talent are engaging in wishful thinking if they believe that outside vendors will be capable of transforming their core processes and way of doing business. Serious change must come from within. I’ve covered this specific topic in greater depth in a recent Inc Magazine article.
- Ability to acquire talent. This includes a firm’s ability to pay top salaries, and the brand appeal to top AI talent. Companies with the sex appeal of the “cutting edge” (often venture-backed) have an advantage here, as do high-tech behemoths with fantastic salaries (Google, Facebook, Amazon, Netflix, etc). Companies with a higher profit-per-employee (just compare Facebook to IBM, for example) will almost inevitably have an edge here as well.
A small handful of smaller (probably well-funded) AI firms will continue to churn out innovative solutions, but we believe that the vast majority of these solutions will require substantial “pilots” and testing with large firms before the use-cases (whether they be tumor detection in medical images, or marketing segmentation on social media) become accessible or viable for most mid-market or small firms.