Companies dreamed big in 2018 in response to external pressures and a changing IT landscape. But many of those...
dreams met a harsh reality in the form of expenses, insufficient resources, cultural resistance and the realization that digital transformation is not easy.
That's according to a series of Forrester Research reports that address the gap between IT ambition and execution in 2018 -- especially when it came to implementing AI -- and predict what's in store for 2019.
The stakes for IT teams remain high, but Forrester believes things are looking up in 2019, as CIOs take a more pragmatic approach to digital transformation and focus on building a more durable and effective foundation upon which to innovate and scale operations.
What's that mean for enterprise AI journeys? Forrester predicted pragmatic AI -- to augment, automate and personalize -- will take hold in 2019, as CIOs let go of their grand, long-term AI ambitions and realize they have to work with what AI can do today, not what it will do tomorrow.
This will help enterprises rise above the widespread epidemic known as AI washing -- i.e., when a company's brands and products claim they involve AI, but the connection is tenuous or nonexistent.
"Swapping out old algorithms with an AI algorithm only provides limited and short-term lift," said Michele Goetz, analyst at Forrester and co-author of the report.
Forrester's 2019 AI predictions report focuses on its five top predictions, based on the thousands of questions Forrester clients have asked them about AI in 2018 and the firm's in-depth research.
1. Data quality will remain a challenge
Forrester said the No. 1 challenge for AI adopters is sourcing quality data. The firm predicted "data doldrums" will continue to drown the majority of firms embarking on AI in 2019. For this reason, Forrester said the tables will turn from AI to IA -- information architecture -- for the majority of firms that have already dabbled in some form of AI, as they realize you need an AI-worthy data environment to utilize AI.
"Data doldrums are and will continue to be on the list for the foreseeable future," said Niel Nickolaisen, vice president and CTO at human resource consulting company O.C. Tanner, based in Salt Lake City. "Data is messy, and it takes times and effort to cleanse data. I expect data will always be messy."
Goetz named data quality as the aspect of AI that is most pertinent to CIOs -- and the most essential of Forrester's AI predictions.
"Data is a digital twin of the business, not digital exhaust," she said, explaining that CIOs must address the data problem in AI in a new way. Simply migrating data to the cloud for data scientists to work with ignores semantic design principles that allow AI to gain a deep understanding of the business and customer.
"Data needs to be interpreted outside of what database, file or table it comes from and be representative of environment, influences, intent, behaviors, decisions, actions and outcomes," Goetz said.
2. Companies will bring humans back into the loop
Forrester predicted that 10% of firms using AI will bring human expertise back into the loop in 2019. Machine learning is great at analyzing data to create models that make predictions, recognize patterns and automate decisions, but it lacks human reasoning capabilities, the report stated.
"Just as we have management and governance oversight of our workforce, AI should also be put under this umbrella," Goetz said. "A human in the loop is both the expert that can support the preproduction training of AI, as well as be the colleague and manager of the AI robot once in production."
"Having the business intelligence, GRC [governance, risk and compliance] and human-to-machine collaboration capabilities to see and manage the robot as a virtual team member is going to de-risk AI actions, while also ensuring that AI can continuously learn from human team members and managers how to do its job better and avoid ethical and moral issues, as well as bad decisions," she said.
Nickolaisen, however, said he thinks bringing humans into the loop has the potential to negate some of AI's power and drive.
"I have always thought that the power of AI was the ability to quickly process the vast amounts of data and variables and deliver what an informed human could do, but much, much faster," he said. "At times, there are data holes, so the AI might present its best decision and let the human 'polish' the decision by intuiting what the missing data might imply -- but this creates its own risks."
3. Enterprises will compete to use AI in the race for AI talent
Michele Goetzanalyst, Forrester Research
It's no secret there's an AI talent shortage in the enterprise not just in the form of data scientists and machine learning architects, but also in the legal, customer experience and operational expertise required to train and manage AI systems. Two-thirds of AI decision-makers struggle with finding and acquiring AI talent, and 83% struggle with retention, according to Forrester's research.
In 2019, the firm predicted we'll see companies start to tackle the AI talent shortage by applying AI to recruitment, as traditional recruiting practices continue to fall short.
Nickolaisen said AI-driven recruitment sounds compelling, but he's not sure where the algorithms will gather the data to propose or present recruitment options.
"I think the more likely way to address the talent needs is firms that will create more usable AI tools. That also implies that the tools will be narrowly focused more vertically," he said.
4. More RPA-plus-AI technology
Forrester predicted that robotic process automation (RPA) and AI will join forces to create digital workers for more than 40% of enterprises. Companies have been treating these sets of technologies distinctly -- RPA for automation and AI for intelligence. But to truly innovate and create breakthrough opportunities, the research firm said it believes a combination of the two is needed.
The report gave a few use cases already doing this: analytics that solve nagging platform issues, chatbots that "boss around" RPA bots, IoT events that trigger digital workers and text analytics that organizes unstructured data into clean files for RPA tasks. The latter leads the four RPA categories in terms of actual deployments.
5. Increasing interest in 'explainable AI'
According to Forrester, some machine learning algorithms are transparent and easily understandable, but others, like neural networks, are opaque.
That won't always fly in 2019. In its AI predictions report, the research firm said it expects regulations like the GDPR’s Recital 71, which states that subjects of automated decision-making have the right "to obtain an explanation of the decision reached," will drive interest in explainability from both enterprises, as well as vendors, creating an emerging supply-side market.
Forrester listed examples that point to this growing demand for explainable AI: DARPA's investment in it and the rollout of explainable AI services and features from vendors like Equifax, IBM and Pegasystems.
Nickolaisen said he has his hesitations about explainable AI, as well.
"I agree that there needs to be some transparency into the algorithms, but does that weaken the capabilities of the [machine learning] to test different models and create the ensemble that best links cause and effect?"