Could AI solve that problem? Speed that process? Five important things you should ask to unearth AI opportunities in your organization
Any IT leader with a pulse already understands that AI will have an increasing impact on their business in 2019, as well as in their broader industry. The same goes for the related disciplines that tend to get lumped in under the AI umbrella, such as machine learning and robotic process automation (RPA).
But that’s not really an actionable insight. It says “be prepared,” but it doesn’t weigh in much on where to get started. And where to get started with AI is the most pressing question today, especially given that what constitutes the best applications of AI (a term we’ll use in blanket fashion here) will necessarily be organization-specific, or at least industry-specific.
We asked a group of AI experts for just that: Actionable insights. How can IT and business leaders think clearly about where AI might be a good fit? That question produced, well, more questions, but productive ones.
Here are five important things you should be asking to help unearth viable, results-oriented AI opportunities in your organization.
1. Where can we make better decisions?
One of the fundamental opportunities for AI or AI-augmented solutions lies in any process or area where your organization can improve its decision-making. In particular, where would your organization benefit from moving to a more data-driven model for making business decisions rather than relying entirely on human instinct and input? (You’ll need to determine criteria for measuring this kind of improvement.)
“The goal should be to ask: ‘Can better decisions be made?’” says Michael McCourt, research engineer at SigOpt. “When a company has people with authority willing to make changes and revisit assumptions using data-driven AI models, then there is an opportunity for a successful AI project.”
McCourt notes that this can cause conflict in organizations that don’t already have a data-oriented culture, especially if people perceive this line of thinking as a threat to their job. If that’s the case, you’ll need the right champions and executive involvement.
Note that your organization can and should determine for itself what makes a decision “better.”
“IT leaders and their business counterparts can break down the business into data and the decisions overlaid across that data,” says John Sprunger, senior architect at West Monroe Partners. “They can determine where automating those decisions or improving the outcome of those decisions will provide value or competitive advantage.”
2. Where are we most inefficient?
Amy Hodler, AI and graph analytics program manager at Neo4j, says that the best first step for uncovering AI opportunities is to look for areas where things aren’t running optimally today.
“Identify which of your processes have measurable inefficiencies,” Hodler advises.
A mostly universal example would be to look for critical processes or tasks in your business that rely heavily (or entirely) on manual data entry.
“If companies are still relying on manual data entry for critical functions, they are putting themselves on a fast track to becoming a laggard, or worse yet – extinct,” says People.ai CEO Oleg Rogynskyy. “Having high levels of manual data entry leads to heightened error rates, slow turnaround time, and required quality checks – these are just a few of the antiquated byproducts that AI solutions can solve.”
Inefficiency isn’t just marked by time-consuming tasks or bottlenecks; it can just as well be measured by the repetitive practice of simple tasks.
“It’s critical to keep an eye out for areas of your organization that generate data and need decisions that require a human to deliberate for a second or less,” says Bill Brock, VP of engineering at Very. “If a person can do a mental task with less than one second of thought, these tasks are ripe for automation.”
Sprunger from West Monroe Partners advises IT leaders to look closely at repeatable, low-complexity, and data-driven human-computer interactions (such as manual data entry) and customer service interactions as possible AI opportunities.
“These are prime use cases for offloading to AI technologies like RPA or AI chatbots,” Sprunger says.
3. Where do we have a lot of relevant data?
AI depends upon the information you feed it. While that could ultimately come from all manner of sources, the early phases of identifying promising AI opportunities will likely be better served by considering areas in which you have robust, reliable, and accessible data.
Data is step two in Hodler’s holy trinity for uncovering good AI opportunities.
“Determine which processes you have the most information on, prioritizing those where you have the most relevant information about the elements and relationships involved,” Hodler advises.
That “relevant information” Hodler mentions is of fundamental importance. Without it, you might be pursuing problems rather than results.
“Make sure you have access to data,” says Tom Wilde, CEO at Indico Data Solutions. “Labeled data that captures the desired outcome is the single most important ingredient for success. Ideally, this data is present inside your enterprise because that type of data is much more digestible than trying to scrape data from across the Internet.”
Those are “boil the ocean” type of projects, and Wilde says they are doomed to fail in most companies.
4. What business outcome do we want to achieve?
As with other technology hype cycles, don’t make the mistake of force-fitting AI into the business rather than letting business needs and goals dictate sensible applications of AI.
The final step in Hodler’s trinity: Always align your AI and ML initiatives with your overall business strategies.
“It makes no sense to invest in an area that does not have a clear impact on business goals,” Hodler says. “If you ignore this, you will never make it to production and be stuck in proof-of-concept purgatory.”
Hodler points to example business goals or impacts such as new business opportunities, decreasing waste or fraud, or accelerating efficiencies. Again, you get to decide internally which goals and impacts matter most – and you should be sure to have more than one perspective in the room.
“If an organization is sussing out ways to effectively implement AI in their business operations, its first step should be to ask business leaders which challenges keep them up at night or which opportunities they think might grow their business substantially – 10x is a good rule of thumb, but not definitive,” says Arti Garg, emerging market and technology director at Cray.
The evaluative work doesn’t stop once you’ve lasered in on a particular goal or problem (or several of them.)
“Once the business problems have been identified, the organization should take a deeper dive to understand what is driving current bottlenecks or preventing progress on high-value opportunities,” Garg says. “If those obstacles can be removed with an AI solution and removing those obstacles would solve the problem, then the project is worth pursuing as an AI project.”
5. Will this actually solve our problem?
By Garg’s count, too few companies ask this question during their AI exploration.
“It is common to identify business problems for which there is a relevant AI project, but where the project will not solve the problem,” Garg says. “In my experience, this is common in AI projects aimed at driving better business efficiency.”
Morten Jørgensen, VP customer solutions at Arundo Analytics, offers another way of framing this question: Is this project actually feasible for us? It’s as important as focusing on business goals for determining AI use cases.
“Our view is that where you start working [on AI] should be driven by a combination of two factors: One, business value – demonstrating that predictive analytics, for example, or AI will yield real value for the business – and two, feasibility – the availability of required data and the ability to use machine learning to solve the problem at hand.”
Garg from Cray points out that when the answer to this question is no, organizational roadblocks are often as much the reason why as of a lack of data or technical expertise.
“The problem is [often] misaligned incentives across business units,” Garg says. “Even if the AI solution offers better insight into the problem, structural barriers will prevent the business from acting on those insights.”
Bonus tip: Ask the above questions during project retrospectives
Brock, the VP of engineering at Very, offers some extra advice for agile teams: Project retrospectives can be especially fruitful times to ask these questions.
“As your team reflects on what happened in the iteration and identifies actions for improvement going forward, you often uncover the largest pain points in your processes, projects, and organization,” Brock notes. “By using these assessments and practicing continuous improvement, the team will begin to naturally focus on problems that can have the highest impact. Take those pain points and look for data, labels, and whether or not the pain point is something that can be solved by [AI and] data science.”
Date: February 19, 2019
Source: The Enterprisers Project