Some of the quick outcomes around this important enabling technology can further drive the business case for more investment in broader digital transformation and innovation initiatives.
Preparing and implementing an AI project can be a multi-year process. The latest data shows that only 28% of respondents said they passed the AI planning stage in their first year. This is due to a variety of factors, including the relative maturity of the technology (at least in the expanding industry use cases), the level of complexity involved (e.g., extensive integration requirements, limited enterprise experience, and lack of in-house skills), concerns about AI bias and governance, risk, and compliance issues, extensive change management requirements, and more.
With such an emphasis on demonstrating quick wins, whether as part of a corporate innovation initiative or a digital transformation initiative, an excessively long AI project could impact the reputation of a larger initiative, not just their own. As CIOs shift from "project to product" in their product management approach, these lengthy AI projects can also delay the release of innovative new products, internal or external.
To achieve rapid success around this important enabling technology, and further drive the business case to invest more in broader digital transformation and innovation initiatives, CIOs can fast-track their AI implementation in five ways:
While we focus on AI machine Xi (ML) programs and examples related to lending decisions in financial services, these recommendations apply to many other AI programs and industries as well. ）
1.Build or buy based on whether AI will become a core competency for an organization.
One of the first decisions to be made is whether to build or buy. While we've heard a lot about the various platforms, infrastructure, and frameworks for building AI on your own, Unsung Heroes tend to be more niche professional AI vendors who offer cloud-based AI services that can be quickly trained and deployed for your specific needs. Use case. The decision to build or buy really depends on how important AI is to your organization as a core competency in the future.
For example, while every financial services firm should be concerned about the looming digital and financial divide between the "haves" and "have-nots" in AI (see "Taking a counterintuitive approach to business strategy and technology deployment"), not every company needs to build its own algorithms in-house. Smaller stores can be very effective in focusing more on the business benefits and outcomes of incorporating third-party AI technologies into their core workflows, such as loan underwriting, without having to build their own in-house AI machine Xi expertise.
2.When it comes to data, "the more the merrier", quality is key.
It has been said that success is 10% inspiration and 90% perspiration. When it comes to AI, a successful implementation is typically 10% AI and 90% data. Any dataset used to train AI ML algorithms to reflect human decision-making needs to be as large and clean as possible.
In simple terms, this means that 10,000 rows of data with 1,000 properties per row are more useful for the ML algorithm than 100 rows of data with 100 properties per row. underwrite.Marc Stein, CEO of AI, said the company is focused on applying advances in artificial intelligence to provide lenders with non-linear, dynamic credit risk models, however, it's not as simple as "more is better." The data type and quantity must match the algorithm type. Deep Xi requires a large number of records to be effective, and statistical-based algorithms can better handle smaller data sets.
If you're using AI to simulate human decision-making, get as much data as possible, make sure every data field counts, and value data quality and consistency. This can be time-consuming, especially when drawing from multiple different **, but if done early and thoroughly, you can avoid a lot of costly rework.
3.Spend time on change management and training on how to best interpret results.
While it's technically simple to call an AI API to pass a new data set and receive scores, it's more difficult to enable business analysts to best interpret those scores and incorporate new processes into the change management and training workflows required for day-to-day work.
While some forms of AI may generate automated decision-making, such as making a "yes" or "no" decision on a new loan based on credit history, machine Xi-learned algorithms often provide more nuanced responses as well. This response may need to be used in conjunction with existing manual processes to make the best lending decisions. As an example, the AI "score" can be on a scale from "A" to "D" and "F". "A" and "F" may be unambiguous "yes" or "no" decisions that can be fully automated for real-time decision-making, but "B" to "D" may still require human underwriters to be involved.
Just as you take the time to train analysts on new financial models and how best to interpret model results, the same is true for AI-based results. Business analysts can spend weeks or even months observing the results returned by the machine Xi algorithm, so they have a baseline of how best to interpret the score. If you're working with an AI vendor, they can provide guidance on how to interpret the results and how to train employees to get the most out of the new system.
Mr. Stein believes that it is crucial to understand that artificial intelligence is not magic. It's just a process of discerning patterns of past behavior in order to be more accurate about the future**. A business can only succeed if it has well-defined problems and easy-to-understand success metrics. For example, "We need to reduce loan defaults as measured by loss ratios" or "We need to raise the current 32."5% conversion rate" and so on. If you don't fully understand the problem, you won't understand the solution either.
4.Take an assumption and test approach, not success or failure.
Since every AI implementation is unique, it's important to approach each project with a "what ifs and test" mindset, rather than treating the project as a complete success or failure. By making assumptions at each step and applying the lessons learned at each step to the next iteration, you can quickly refine your AI deployment until it becomes a viable solution that can deliver meaningful results.
While assumptions and test methodologies lengthen project deployment times, the benefit is that you can continually fine-tune your solution to incorporate real-life lessons learned, align with customer and employee requirements, and keep moving on to the most compelling business cases that will make your solution sustainable.
5.Incorporate all forms of automation into your vision for the future.
As you begin your initial AI pilot, proof-of-concept, or MVP, keep in mind that your organization's future vision for enterprise-grade AI could be from a mix of fully manual processes to multiple types of automation, from those adopting machine process automation (RPA) to more sophisticated AI. It's often necessary to reinvent business processes from scratch and then apply the tools that work best for the job at each new step. Simply inserting RPA or artificial intelligence into an existing business process that hasn't changed is likely to miss out on the art of possibility.
Another important factor is the switching that takes place between each tool. This can be man-to-machine or machine-to-machine. By optimizing handoffs and making them fast, seamless, and reliable, you can further enhance future business processes to be as cost-effective and competitive as your business goals and what the market demands.
The good news is that AI implementations can happen quickly, but that doesn't necessarily mean AI is getting smarter. It's about making the right choices, such as building or buying, becoming obsessed with data quality (and therefore customers), spending enough time on change management and getting the business involved early, taking a "what ifs and test" approach, and ultimately incorporating multiple automation technologies into your vision for the future.
If your AI project is taking quite a long time, be patient and stick to it. You can also take advantage of some of the suggestions here to help you get to the finish line quickly. Of course, just like digital transformation, the race never ends.
This article is from *** CIO Information (non-profit organization;Information Officer Information;Information Officer Recruitment Part-time Information;Assist IT practitioners to better engage in career development).