Driving AI Adoption: An 8-Step Blueprint for Your Team’s Success
3AI August 21, 2023
Featured Article:
Author: Ganes Kesari, Innovation Titan
A Telecom major was grappling with high customer attrition. The firm was one of the largest Telecom companies in the world and a market leader in Asia.
The marketing team’s heuristics-driven approach to customer retention was dated and ineffective. Reviewing the business performance in a weekly huddle, the CEO knew they had to do something different.
The firm turned to data science to solve this challenge. Machine learning (ML) algorithms were trained to predict customer churn. Simple algorithms such as decision trees used attributes such as “bill amount” and “outgoing call pattern” to improve customer retention by 39%.
While the marketing team was thrilled with these results, the data science team turned to advanced black-box algorithms such as Neural Networks that pushed accuracy even higher. Pilot tests run on high-value customers turned out to be a resounding success – Artificial Intelligence (AI) delivered 66% higher customer retention than the traditional approach.
The solution was ready for rollout, or so it seemed. Then, things turned south.
The marketing product managers flatly refused to use the solution. They found it hard to trust an algorithm that spat out a set of customer names with little explanation. Many of these recommendations were counter-intuitive and the entire process felt wrong.
Despite the data-backed results, they gave the data science solution the cold shoulder. The graveyard of AI projects is filled with such advanced, accurate, and well-meaning yet unused solutions.
What are some simple yet practical approaches you can adopt to turn the tide and make users fall in love with your Data & Analytics (D&A) solution? This article will share actionable tips about how and when to get started.
Why do data science solutions fail to get adopted?
In a bid to deliver organizational value from analytics, leaders often obsess over the kind of projects to choose or the technology needed to deliver them. While both of these are important, the elephant in the room is organizational inertia.
Often, end users are unwilling to change their ways or are uninterested in learning how to use yet another tool. However, these are really symptoms of the problem, not the root cause.
In most cases a key challenge is that leaders do an ineffective job explaining why D&A is needed, how it will enable business goals, who should get involved, and what is expected of each stakeholder.
4 shifts needed to enable decision-making using data
Adoption of data science, four areas need attention. Inspired by Brent Dykes’ four-pillar framework for data culture, here’s why they distinguish success from failure:
1. Shift Process-set: Organizational priorities and business challenges trigger the need for data & analytics solutions. While designing data science solutions, it is critical to reimagine the business processes around these solutions for effectiveness and ease of adoption.
2. Shift Toolset: Picking the appropriate data science tools, hiring technical talent, and investing in technology solutions are essential. Organizations that cannot architect change in their toolset will lose out to competition and perish over time.
3. Shift Skillset: Given the magnitude of shifts in technology and business processes, people must be enabled with the right skillset. End users should be equipped with data literacy while technology teams must keep pace with skills needed in a fast-changing landscape.
4. Shift Mindset: As we saw earlier, the three shifts above will amount to nothing without a proportionate, meaningful shift in human mindset. “It’s critical that all levels of the organization participate in this change,” says Bob Dunmyer, Partner, Advanced Analytics at Guidehouse.
“This sounds like a lot of work. When should I start?” one might ask. Given the magnitude of work involved, an early start is crucial. According to Bob Audet, Partner, IT Strategy & Transformation at Guidehouse: “Leadership should plan for change management and adoption during the strategic planning stages of an initiative.”
He adds: “Leaders should define the ideal future state for the organization and develop a change vision and strategy to execute a successful transformation. During the strategic planning stage, leaders must assess the current state of the organization, including people’s readiness for transformation.”
With clarity on the organization’s current state and end goals, bridging the gap between the two becomes obvious. For example, “based on this assessment, if resistance is predicted to be strong, additional creative tactics can be planned early to deal with it,” recommends Audet.
8 steps to improve the adoption of AI initiatives
We’ve seen why you need an early start to improve AI adoption. This timeline can be divided into four broad phases: Pre-implementation (or planning), Implementation (or solution-build), Go-Live (or rollout), and Post Go-Live (or maintenance).
We’ll now look at eight important steps you must plan across these stages. Since the theme of the article is adoption, we will not delve into the technology aspects of solution-building, namely, “Shifting Toolset.”
A. Pre-Implementation
1. Onboard executive commitment and ownership [Shift Mindset]
In most organizations, change starts at the top. Executives must define the vision for D&A and storytell why it is crucial for the business. They must do more than fund the initiatives. Leaders must own the strategic outcomes and actively participate throughout the initiative.
2. Definite target outcomes and KPIs [Shift Process-set]
While picking initiatives, assess whether they are aligned with the organizational goals and D&A vision. “One less-talked-about reason why analytics projects fail is poor problem formulation,” says Hitesh Sethi, Head of Analytics and Business Advisory at Dun & Bradstreet India. “There’s a gap between the business problem and how it’s formulated due to factors such as poor communication, not having ‘lived’ in the problem domain, or a mad rush to apply complex techniques such as deep learning,” he clarifies.
When there’s clarity on the business problem, it is easy to identify the target outcomes and paint a picture of what success should look like. Ensure stakeholders agree on the success KPIs and how they will be measured.
B. Implementation
3. Tailor business workflows and plan integration [Shift Process-set]
Anything that introduces friction in a user’s workflow will dampen adoption. To improve the likelihood of usage, embed the AI solution deep into the business workflow. This often calls for redesigning processes and re-engineering connected technology solutions. Plan time for them.
4. Identify data champions and upskill users on data literacy [Shift Skillset]
Leaders often wonder who should lead change management on the ground. “Business divisions must own the change efforts since they know where the pockets of resistance exist and who may be the earlier adopters and supporters,” says Audet.
Identify data champions by picking business users interested in technology who can passionately evangelize D&A solutions within the organization. Initiate data literacy training to help all users get comfortable with reading, interpreting, and communicating with data.
C. Go-Live
5. Launch organization-wide communication and roadshows [Shift Mindset]
Leaders often make the mistake of downplaying communications about D&A initiatives. Data science initiatives need aggressive internal marketing through roadshows, executive presentations, and creative campaigns to inspire users.
6. Measure business impact and report ROI [Shift Process-set]
As the popular saying goes, “nothing succeeds like success.” Track and report outcomes by measuring the pre-agreed KPIs. Attributing this success back to the AI initiative calls for explicit experiments such as A/B tests or split tests to validate impact.
Design experiments within an organization wherein one team or geography use the AI application, and this is compared against another that maintains a business-as-usual process. The incremental lift achieved in results helps confirm that the benefits were generated by D&A.
D. Post Go-Live
7. Incentivize usage and celebrate small wins [Shift Mindset]
“Traditional change management relies on a simple concept: If you educate people, they will change. Yet humans consistently disprove this concept,” notes Dunmyer. “We frequently make irrational choices — for example, not eating healthy or failing to save for retirement.”
People respond well to incentives — monetary or non-monetary. Leverage this behavior to reward adoption and acknowledge progress by celebrating early wins.
8. Expand the solution and onboard new teams [Shift Process-set]
Once you taste success with a rollout, it is critical to sustaining the hard-earned momentum. Revisit strategic plans to identify the next best set of products, teams, or geographies that could benefit. Learn from the experience and improve the process to build a data-driven organization.
Logistics firm saves $1.2 million by adopting a data science solution across its warehouses
A large cold-chain firm in the U.S. was tackling the challenge of delays in turn times across its warehouses. The executive leadership planned its foray into data science by building an intelligent appointment scheduling system.
In the pre-implementation phase, the team picked this initiative from a list of 100 ideas and chalked out its target outcomes and KPIs. Leadership identified three technology-savvy warehouse managers to pilot the initiative. Through weekly meetings, these data champions shared business inputs to shape the solution, design the workflows, and plan the rollouts.
The solution went live with great fanfare and was communicated in executive town halls, emailers, and internal magazines.
Quick wins delivered by the application through efficient scheduling were acknowledged through infographics and video testimonials. Soon, the solution was rolled out to a few dozen warehouses across the country. With strong executive commitment, the solution saw robust continued adoption and delivered an annual ROI of $1.2 million through saved penalty costs.
Note: This article was originally published on Forbes.
Title picture: freepik.com
