Is AI SAFe®? Strategic AI with Scaled Agile Methodology

While much of the news of AI in financial services has focused on Fintech innovation, such as I Know First (a Big Data, Self-learning Algorithm for investing), AI has been in mainstream financial institutions for years.

According to Bank of America, since launching in 2018, Erica® had helped nearly 32 million of their clients manage their financial lives and had breached 1 billion client interactions (nearly 1.5 million per day) by October of 2022.

Since its launch, Erica has expanded to include new features and functionality:

• Clients viewed 37 million proactive insights to help them review their finances and cut recurring subscription charges that may have increased unexpectedly, know when they’ve received a merchant refund, or have duplicate charges.
• More than 4 million proactive notifications about eligibility for Preferred Rewards have helped clients enroll in the program and enjoy the benefits.
• 60 million Spend Path insights have helped clients understand their finances with a weekly snapshot of spending.
• More than 98% of clients get the answers they need using Erica. In September 2022, the bank launched Mobile Servicing Chat by Erica to connect clients for a live chat with representatives to answer more complex servicing questions, with more than 170,000 chats having already taken place.
• Coming in the first half of 2023, Erica will connect clients to financial specialists when they have questions about new products and services, such as a mortgage, credit card, or deposit account.

In the last quarter of 2021, Fidelity released an AI-powered prospect analysis and lead conversion platform. Born in the Fidelity Labs incubator, the new startup, known as Catchlight Insights, tested with a small group of wealth management firms in November of 2022, it is already being integrated with leading CRM platforms such as Wealthbox.

Catchlight utilizes institutional-grade data and AI to automate, scale, and repeat lead research that was previously manual and time-consuming. It streamlines cold-calling and emailing by helping advisors prioritize leads using data from five for-profit databases. Their lead-evaluation tool assigns up to 1,000 attributes per record and has data on over 255 million U.S. consumers. The tool combines client attributes, algorithms, and advisors’ success rates to rank leads and provide tailored sales pitches.

Vanguard’s implementation has been more focused on customer service and investment strategy to benefit their customers. When considering AI solutions, they look for answers to two questions. First, how does it help their participants? Second, how does it help them improve a process or service?

They use AI to look at everything they know about each participant to determine which actions will benefit which participants, and then use behavioral finance to craft messages that motivate participants to take those actions.

The active equity team is using natural language processing (NLP) to analyze a vast amount of new information each day, including SEC filings, annual reports, and quarterly earnings. The NLP program, which processes up to 1.25 million words per minute, evaluates and interprets content, and generates reports based on the analysts’ criteria.

By assisting the active equity team in analyzing new and complex information in a fraction of the time and on a large scale, the AI system can help fund managers make better investment decisions that will in turn benefit participants.

Combining AI with Scaled Agile = Business Agility

Business Agility is crucial in the digital age as it allows organizations to quickly respond to market changes and emerging opportunities with innovative, digitally-enabled business solutions. The pace of change in the digital world is incredibly fast, with customer desires, competitive threats, technology choices, business expectations, revenue opportunities, and workforce demands all happening at blistering speeds. The ability to achieve customer delight at the speed of market changes requires validating innovations with customers and being willing to pivot quickly when necessary.

Agile techniques can provide a powerful opportunity in the right conditions, particularly in the growing area of AI and advanced analytics. However, many organizations struggle to move beyond the proof of concept stage for AI projects because they lack a documented delivery strategy or try to force the process into traditional IT delivery methods. Hence the desirability of implementing AI with Scaled Agile methods.

Agile is best suited for complex problems with initially unknown solutions and product requirements that are likely to change. It works well when the work can be modularized, close collaboration with end-users is feasible, and rapid feedback is possible. Creative teams typically outperform command-and-control groups, making agile techniques particularly useful in the realm of AI and advanced analytics, where poorly defined solutions are best iterated in cycles of rapid discovery.

Adopting agile techniques can lead to faster speed-to-market, competitive advantage, and better collaboration for organizations looking to scale AI. However, there are two common traps businesses fall into. The first is trying to embrace an agile mindset while using traditional tools like spreadsheets and email, which can hinder agile values such as prioritizing interactions over processes and tools. When implementing AI with Scaled Agile the failures are even more astronomical! Instead, collaboration tools like Microsoft Teams™, Slack™, Trello, and JIRA prioritize adaptability and visibility to optimize delivery outputs. The second trap is investing in agile technologies without aligning them with business priorities, which can result in isolated pockets of value. To avoid this, businesses need support from senior stakeholders and cross-functional teams to scale agile technologies and ensure they are applied to the right business use cases.

Develop AI Competency Strategically

Many organizations lack the necessary expertise to effectively implement Artificial Intelligence (AI) initiatives. Even those with some AI projects underway may not have the right skills in the right areas. In order to successfully navigate AI opportunities, organizations need to foster collaboration between business and technology professionals and develop AI competency. SAFe® emphasizes the importance of constant interaction between these groups, with techniques like strategic portfolio Kanban systems, PI planning, System and Solution Demos, and Inspect & Adapt. Continuous exploration activities, design thinking, and work definitions also leverage the overlap between business and technology representatives. It should be emphasized that when executing AI with Scaled Agile, these techniques must be used even more rigorously to avoid costly disconnects.

In order to establish a productive solution development process for AI systems, organizations must apply a clear decision-making framework. Many AI initiatives fail to deliver the expected results because of poor decision-making around how and why to use AI. Some organizations adopt AI simply because they believe it is popular, without fully understanding the effort required to implement and scale it, its impact on the organization, or whether it will deliver the intended benefits. With a sound framework in place, organizations can successfully build the AI plane while they fly it.

AI Decision Framework

Align Strategy: In order to achieve beneficial outcomes for the business, it is important to align AI initiatives with the overall strategy. This involves assessing economic viability, using lean budgeting, and implementing a portfolio Kanban system and lean business case to ensure alignment with strategy. Furthermore, PI Planning can be used to continually align AI strategy with implementation.

Customer Experience: Customer centricity is crucial to ensure that an AI initiative is solving a genuine customer problem. Design thinking can be applied to AI capabilities to help define the customer problem and integrate intrinsic AI capabilities, such as neural networks or natural language processing, into a favorable customer scenario.

Continuous Learning: A Continuous learning culture that encourages continuous exploration and improvement is necessary for a successful AI-powered solution. Uncertainty in solution development is always high, but it is especially so in the case of AI. The SAFe® Lean Startup Cycle can help by creating a clear business hypothesis, building an AI MVP, and validating it against appropriate measures.

Measurable Objectives: Empirical milestones are crucial in guiding the development of a successful AI solution. AI capabilities must be seamlessly integrated with the rest of the solution throughout the iterative development process. Incremental releases are leveraged to gather valuable customer feedback, allowing for continuous improvement of the solution.

To effectively scale AI, a cultural shift is necessary, in addition to the implementation of agile methods and a data culture. Simply hiring data scientists with agile experience is not enough without a strategy to foster an organization-wide culture that embraces the insights generated by AI.

Redefining how success is measured is also crucial. Metrics at the highest levels of the organization must be recalibrated with data and AI in mind, including redefining the frequency, methods, and parameters used to measure the value of AI. Without this shift in mindset, agile methods and a data culture will not be effective in scaling AI.

Ultimately, the success of scaling AI depends not only on the methods and technology used but also on the organization’s culture and approach to measuring value.

Build an AI Solution Path

While these may vary by organization, the process of implementing AI with Scaled Agile requires five key steps that must be followed for a successful outcome.

Identify: The first step is to define the business opportunity and understand the customer’s needs, and then assess which suitable AI architectures could support the two. This requires involvement from both technology and business stakeholders.

MVP: The second step is to pilot the AI capability, which involves building a Minimum Viable Product (MVP) and testing it to ensure that it solves the business problem. It is important to apply basic AI capability monitoring and measurement at this stage.

Effectuate: If the MVP is proven viable, the third step is to operationalize the AI capability by designing and implementing it in a way that allows for full integration with existing systems and support for required business scenarios. It is important to plan for the investments required to continuously monitor, adjust, and retrain AI components.

Magnitude: The fourth step is to scale the AI capability to support increased volumes of customers. This introduces unique challenges, such as the possibility of the model, data, and business parameters growing out of alignment. Adjustments to the model, learning algorithm, and data processing may be necessary to sustain the model through scaling and beyond. Additionally, scaling may lead to the emergence of new scenarios that require updates to the AI capability or even a drastic change to the ML model or learning algorithm.

Control: The final step is to govern the AI capability, which involves establishing a governance process that spans multiple solution trains in the portfolio or even multiple portfolios in the enterprise. This process should account for areas that are particularly important with the advent of AI, such as effective data management, privacy and security, computing power, the bias problem, and traceability. Ethical implementation of AI is a dominant topic in this domain and requires serious investment by organizations leveraging this technology.

To achieve a successful outcome, organizations must carefully follow each step in the process of scaling AI. Each step requires involvement from different stakeholders and careful planning to ensure that the AI capability is integrated and supported in a way that aligns with the business’ goals and values.

It’s still about the people!

Proficient AI systems and algorithms necessitate teams of experts who monitor and adjust them to ensure they perform as intended. In fact, a 2020 MIT and Boston Consulting Group survey revealed that companies that effectively implement AI do so by investing in AI talent.

AI is only a tool, it’s the teams of people working every day who ensure that this valuable business technology is operating in the best interest of the company and its customers. Creating the right teams and the proper environment for typically disparate teams to work together is critical.

Data science teams know what AI techniques are most effective, but the customer experience, user design, strategy, and development teams know what will have the most impact. Involving the right stakeholders and creating powerful partnerships are critical to maximizing the customer experience and/or financial benefits and aren’t possible if the groups work in silos.

Addressing the need for talent and diversity, Vanguard’s head of Participant Intelligence Andrea Needham emphasized, “When Vanguard uses artificial intelligence (AI) in retirement plans, our goal is to positively affect participants’ financial health. To do this, AI will always need humans. Bias is a challenge we face every day. Our teams are constantly checking the validity of our AI’s outputs—the recommendations we make to participants. We monitor to what extent the recommendations are taken, and we constantly feed that information back into the system. Without this robust bias oversight, it wouldn’t be possible to use AI to effectively improve our participants’ financial lives.”

The same is true for Agile transformations. Neglecting the cultural and change-management aspects of agile is a major mistake often made by large organizations. Achieving a successful transformation involves more than just changing the way teams work at the grassroots level; it also requires a shift in the way executives operate, as their actions have a significant impact on the organization’s culture.

Taken together, properly implementing AI with Scaled Agile will more effectively improve customer centricity, collaboration, learning, responses to competitive threats and business opportunities, and other key organizational objectives.