My Journey to AI
As Chief Digital Officer for IBM Cloud and Cognitive Software, I’m in a unique situation. I work with an incredibly smart team who take our software offerings and digitize the discovery, engagement, purchase, and adoption experience. At the same time, I get to work with customers on their Journey to AI helping them collect, organize, and analyze data to infuse AI into their business. I meet with CDOs (Chief Digital Officers and Chief Data Officers) asking how IBM is helping customers democratize data, govern data, and deliver self service empowerment. All of this is to lay a solid foundation for AI in the business. Back to the unique situation, while I talk to customers about our strategy, I am living it within my own organization.
I wanted to lay out a few examples of what we’ve done in IBM Cloud and Cognitive Software as we embrace our own Journey to AI.
The Journey to AI is simple. Organizations want to take advantage of AI, but they need a trusted foundation and access to the data. As our GM, Rob Thomas says, “there is no AI without IA (Information Architecture)”. That foundation includes this:
Collect: This is how you get access to data.
- Structured, unstructured, semi structured.
- Proprietary and open source.
- In the cloud. On premises.
- Any of it. All of it.
Organize: This is how you govern your data.
- Discover your data assets.
- Catalog your data assets.
- Profile and categorize your data assets.
- Policy based enforcement to access.
- One source of truth.
Analyze: This is how you become insight driven.
- Describe the plan.
- What happened? How does it compare against plan? Why did it happen?
- What will happen next and how do I change my trajectory?
- Automate and optimize the decisions.
Infuse: This is how you embed AI into your business.
- Build your machine learning models.
- Deploy your models.
- Manage your models with trust and transparency.
We are running projects that include some of these steps and projects that include all of these steps.
- Real time dashboards for insight
Collect, Organize, Analyze
My product managers lean on a dashboard called “Pearl” for real time insights on our marketing funnel, orders, revenue, and campaigns. Pearl was built, supported, and driven from our Chief Marketing Office and Digital Business Group, and I am a consumer of the content. It’s the most fundamental example of becoming a data driven business. At any time, on any day of the week, we can log in and see the performance and health of my business. We don’t have to rely on powerpoint charts that are days old with data, and I don’t have to bug my team for what’s happening. I can log in and see for myself. As an executive, I try to lead by example. If I go into the dashboard to get insights, there is no reason from my team not to do the same.
How we did it?
There are 16 data sources that collect data types ranging from chats, emails, web visits, and our finanical ledger. We organize that data into a centralized data lake and run the queries from there. The data is visualized in a dashboard with core reports for marketing campaign managers and offering managers. Here’s a 5 minute clip of Michelle Peluso, IBM CMO and SVP of Digital Sales, and team talking to CNBC about Pearl: https://www.cnbc.com/video/2019/01/02/ai-is-increasing-understanding-of-customers-at-ibm.html
2. Affinity Models
Collect, Organize, Analyze, Infuse
For greater opportunities to up-sell and cross sell additional services, our data science team have built affinity models. These models tell us what products and offerings are used together by our successful customers to various solutions. We take this insight and build e-nurture campaigns for the users, predicted by these models, encouraging the use of the offerings and services with strong group affinities with the offerings they are currently using.
How we did it?
As our customers interact with IBM Cloud, we collect the data of various IBM Cloud offering pages, events, and developers’ milestones. We clean and organize this data in the form of journeys of customers’ success. These journeys are analyzed to extract the insightful user sessions of sustained activities composed of transitions from one offering to another. Using these transitions, we build AI based Affinity Models, which can be used to predict the next best suitable offering for a given client.
Fig X. Watson Machine Learning centric affinity model
3. Virtual Agents
Infuse
Virtual Agents (a.k.a. chat bots) are conservation tools that interact with users answering common questions. Our data shows that customers are 4.2x more likely to have a conversation on a page with a Virtual Agent than just a live chat button with a human behind it. We have over 18 Virtual Agents supporting our Digital portfolio. Prospects that contact us are 5x more likely to convert to a customer. In one example, chats with users are 53% more likely to end in a sale after interacting with the Virtual Agent. These agents provide significant coverage, especially on weekends, when we don’t have as many humans behind live chat.
How did we do it?
Working collaboratively across Digital Offering Management and Marketing, we built a corpus of questions and answers that would likely come up in the conversation with a user. We infused the Virtual Agent on our web pages.
4. Usage Insight for Growth Hacking
Collect, Analyze
Growth Hacking is rapid experimentation in the marketing funnel, in product, and in the sales process to get to growth. We run Growth Hacking experiments weekly to figure how we attract and retain more users. In a digital model where we never speak to 90% of our customers, this is critical. Examples of tests include:
- Implementing an upgrade button within Watson Studio converting upgrades in a 30 day period
- Trial email “nudge” to remind users to download their trial driving a 36% increase conversion rate
- Changing a hard requirement on a registration form driving 5x increase in MQ on Cloud users
How we did it?
There are many examples of growth hackings, but it starts by having our products instrumented. We collect data based on milestones activities (ex. web page visit -> % clicked the trial -> % purchased). We analyze this data in user friendly dashboard. All of the user data is anonymous but we can still get great insight in aggregate.
These are just a few examples of what we’re doing in our digital transformation journey. It’s very encouraging and rewarding when the team has discussions and prioritization meetings based on data they are getting from these projects. Would love to hear how you are empowering yourself and your team with data driven decisions.
