Chapter 3. The Business Transformation: Bringing AI to Every Business Process
Although an AI transformation can spring from a technology-first approach, the full impact of AI is best achieved if technology and business come together to rethink the primary functions in the organization. Every business process can be optimized or completely redefined with AI. The question is where to start.
Transforming Your Business Units
Just like the evaluation of potential scenarios in which you can bring AI to your applications, the first step in transforming your business is to inventory and categorize your processes. There are many frameworks in the industry for that, but I particularly like the one defined by management consultant and author Geoffrey Moore in his book Zone to Win: Organizing to Compete in an Age of Disruption (Diversion Books). Geoffrey states how every organization has to carefully balance four different zones:
- Incubation
- Focused on evaluating multiple new opportunities enabled by new technology waves to identify future viable business options
- Transformation
- Focused on scaling a particular disruptive opportunity identified in the incubation zone to a material revenue
- Performance
- Focused on maximizing the current primary revenue stream of the company
- Productivity
- Focused on delivering the programs and systems to support the performance zone
This model is very effective for understanding and balancing the different goals an organization has to pursue at any given time, and it applies perfectly to AI. As you identify potential use cases for AI you should always keep in mind the associated zone for each. A balanced strategy will address use cases that are representative of all four zones, so you can effectively transform your entire organization:
-
The incubation zone will include moonshots enabled with AI. These moonshots are managed differently from other projects; they don’t follow the same metrics (e.g., they are not pressured with short-term ROI goals) and are extremely agile and flexible.
-
The transformation zone will focus on projects (ideally only one at a time) that are developed in the incubation zone, but are being scaled to the entire organization. These are moonshots that we bring to market. They need special attention and a strong push to escape the inertia of the existing business in the organization.
-
The performance zone will focus on AI initiatives that can increase the revenue in the existing primary business. These initiatives can be managed at the existing rhythm of the business and follow the same ROI requirements and investment prioritization process.
-
The productivity zone will contain use cases that can increase the effectiveness or decrease the costs of supporting processes. Just like in the performance zone, these use cases can be treated in the context of the existing processes for increasing the efficiencies in your organization.
As you prioritize the use cases in your organization, you will again have to consider the zones they belong to. Applying the same criteria for use cases in different zones is a common mistake that will result in an unbalanced approach for your strategy. If your main criterion is cost savings, you will end up with only use cases that are in the productivity zone, and you will miss potential areas where you can disrupt the business (or even worse, find yourself disrupted by competitors). The opposite is also true: focusing only on criteria that favor incubation projects will mean you miss opportunities to make your current business more competitive.
Identifying and Evaluating Use Cases
With that framework in mind, the next step is to identify as many use cases as possible, and evaluate them to prioritize the ones that are most relevant to your business. It is critical that this process be done in a partnership between the technical departments and the business units. Technical departments can bring AI to your applications on their own, but they cannot bring AI to business processes without the involvement of the business units. And that can happen only with the right internal structure and motivation.
A common approach is to create a multidisciplinary virtual team with representation from the technical departments and each of the business units. In some cases that isn’t needed, especially if your organization is mature in its role as a software company, because there are already mechanisms in place to bring together technology and business—but no matter how you do it, the identification and evaluation of use cases for AI should be a joint exercise between the technical departments and business units. We’ll discuss different organizational models to support that concept in more detail in Chapter 6.
Once the internal construct is in place, you can use the four zones (incubation, transformation, performance, and productivity) to identify the use cases in your organization. At Microsoft, we use a very simple framework to help customers in that process called Agile Value Modeling. In a highly iterative process between business and technology, projects are identified and distributed depending on their zone, with each represented with a bubble proportional to its potential value to the business.
The iterative nature of the process will help you to move through the quadrants to identify new projects to try. An incubation moonshot project may result in several smaller, tactical projects in the productivity area. A tactical project in the performance area can turn into a strategic transformational project. During the conversation, the business units and technical departments can help each other by sharing their different approaches:
- From technology to business
- Start with what AI can do. Basic training for the business users on the three primary types of AI capabilities (learning, perception, and cognition), as covered in the cheat sheet presented in Chapter 1, can be extremely helpful. They can then apply these concepts to their business model, with the technical members in the virtual team further defining the details iteratively.
- From business to technology
- Start with business scenarios, and jointly explore how they can be improved or redefined with AI. For example, business users can share the main challenges in the business, the long-term revenue plans, the primary cost drivers, and any other relevant considerations. A list of common use cases in organizations can be helpful for this approach; you can use one of the lists provided in this chapter as a starting point.
By now you should have a long list of potential business use cases for AI. It’s now time to evaluate and prioritize them. Remember that it is critical to have representation across multiple zones, and the criteria for the prioritization should be modified depending on the zone. For incubation and transformation use cases, you should focus on criteria that favor the long-term opportunity:
- Market differentiation
- Will this use case create a strong, sustainable differentiation for the organization in the market?
- Market opportunity
- What total addressable market is associated with this use case?
- Unique assets
- What unique assets (data, processes, customer base, etc.) does the organization have to address this use case in a different way than its competitors?
- Investment required
- Can the organization sustain the long-term investment required to successfully develop the use case and move it to the transformation zone?
- External disruptions
- Is there an external disruption that the organization may miss out on if it doesn’t target this use case? What is the potential negative impact of that?
- Network effect
- Can this use case ignite a virtuous cycle with exponential growth?
For performance and productivity use cases, you should focus on criteria that favor short-term impact and viability:
- Cost
- What is the estimated cost of the implementation, including internal and external resources?
- Return on investment
- How profitable can we estimate the use case to be, in terms of incremental sales, reduced costs, productivity gains, or improved quality?
- Payback period
- What’s the estimated time it will take for the investment to pay for itself?
- Nonmaterial impact
- What other metrics might be affected by the use case (customer satisfaction, loyalty, brand perception, increased barriers for competitors, etc.)?
- Risk
- How likely is the implementation of the use case to be successful? What dependencies and uncertainties are involved?
- Readiness
- Is the organization in a good position to implement this use case? Does it have the necessary resources (data, infrastructure, human resources, organizational resources) for it?
These criteria should provide a good framework for the combined technical and business team to evaluate potential use cases and discuss them with the company’s management. However, at the end of the day, the final prioritization should be based on the organization’s expertise. Never underestimate the power of intuition or of the passion of the team behind the idea. Making the person or team that proposes an idea the “CEO” of that idea is a powerful concept that can make a big difference to its success.
The criteria used should also drive the definition of success. Learning is as important as implementing the use cases themselves, and we won’t learn if we don’t measure the success of each initiative we target. Consider the following areas when establishing the metrics for your AI use cases:
- Metrics associated with the business impact
- These are closely aligned to your selection criteria and they are different for projects in each of the four zones. Some examples of business impact metrics are adoption, revenue, cost savings, and productivity increase.
- Quality metrics
- You’ll want to measure the level of quality of the final product, as well as its evolution. Metrics in this group can include the accuracy of the system, user satisfaction, deployment ratios, and volume of support incidents.
- Metrics related to the implementation
- To improve future execution and readiness, you should also measure the implementation itself: for example, schedule performance, budget performance, employee satisfaction, and completeness of requirements.
These metrics should be carried throughout the entire lifecycle of the project and owned by the entire multidisciplinary team associated with it. As you will learn in Chapter 5, a team that is aligned with the business outcomes of a project is the essence of a successful AI implementation.
Typical AI Use Cases
The following sections present a selection of popular use cases for you to use as inspiration. They are separated into two categories of business processes:
-
Horizontal processes, which are similar across any industry. These are well-known functions like marketing, sales, or HR. The use cases behind these functions are usually in the performance and productivity areas. They won’t disrupt your business, but they can provide ideas for low-risk, short-term projects. In many cases, you can even leverage an out-the-box solution from a third-party provider, or reusable artifacts like accelerators or AI models.
-
Vertical processes, which are unique to your industry vertical (such as manufacturing, health care, or retail). These processes are usually highly unique in every organization, and they rely on years of expertise, IP, and proprietary datasets. They have a strong potential to disrupt your business and increase your differentiation, but they are higher-risk, longer-term projects involving custom AI development.
We will start by covering use cases for the primary horizontal business functions. Then we’ll look at vertical use cases, and I’ll provide examples from different industries. (Feel free to jump directly to the section that covers your industry; I won’t be offended!)
Each use case is mapped to the primary AI capability that you will use to implement it. Looking at use cases like “predictive maintenance” or “sales forecasting” can be useful to provide inspiration for a “business to technology” approach, whereas focusing on AI capabilities like “vision” or “classification” can help you find new ideas for the “technology to business” approach. Either way, you can use these lists as inspiration while strengthening the deep partnership between business and technology required to identify the best use cases for your own company.
Horizontal Processes
As mentioned previously, horizontal processes are well-known business functions that span industries, such as marketing, IT, and customer service. Microsoft itself is going through a transformation of those functions, so I’ll share examples of our own journey for each of them.
Marketing
Marketing is probably the function in any company that has experienced the biggest digital transformation in recent years. Digital marketing now accounts for a significant part of most companies’ marketing resources, and even traditional advertisement is heavily driven by data nowadays. With that strong digital foundation, it’s not a surprise that marketing is a function with so much potential for AI.
AI is great at providing insights into the immense amount of data generated by marketing activities. Table 3-1 lists some of the most common use cases for this scenario, along with the primary AI capability enabling each of them.
| Use case | AI capability |
|---|---|
| Brand insights: Understand in real time how customers are talking about your brand, identifying sentiment and extracting key topics and drivers. |  |
| Customer insights: Identify customer segments, patterns, and trends in customer interests by analyzing sources such as search engines, social media, etc. |  |
| PR analytics: Understand the impact of your own PR activities and those of your competitors, identifying themes and sentiment. | |
| Feedback insights: Centralize the feedback channels in the company (email, social media, websites, etc.) and identify hot topics and dissatisfaction drivers. | |
In other cases, AI not only can inform us so that we can make better decisions, but directly act on our behalf to optimize our marketing activities. Table 3-2 gives some examples.
Beyond insights and actions, AI can also have a strong impact in the transition from marketing to sales. A few examples of this are given in Table 3-3.
Microsoft uses a lot of these scenarios internally with positive results. Our Global Demand Center, which centralizes our customer acquisition and engagement marketing engine globally, has been using AI to optimize the leads passed to sales. Fake contacts are removed with AI, and opportunities are scored and ranked with thousands of variables in real time. Following qualification, marketing hands off the leads to the appropriate sales channels. Using this approach, nonworkable leads were reduced by two-thirds and conversion rates increased 108% year over year.
Sales
AI not only can improve the number and quality of the leads passed to sales, it can also help during the sales process itself. This can be done by directly assisting the sellers in tasks such as those listed in Table 3-4.
| Use case | AI capability |
|---|---|
| Pipeline prioritization: Identify customers in the pipeline at risk of churn or likely to move the opportunity forward with an action. |  |
| Cold call lists: Generate customer lists for cold calls, identifying fit with product offerings, upselling opportunities, or renewal opportunities. | |
| Next best action: Recommend the best potential next action by the seller for a specific account, given the past actions and the pipeline state. |  |
| Meeting assistance: Provide guidance to the seller prior to a meeting or presentation to a customer, with context, opportunities, and recommended approaches. | |
AI can also be effective for broader scenarios that may be relevant to sales management (see Table 3-5).
| Use case | AI capability |
|---|---|
| Sales forecasting and quota setting: Predict the team’s sales forecast based on internal and external conditions for reporting purposes or quota setting. |  |
| Best practice identification: Analyze behaviors and practices correlated to top sales, to implement as best practices in the team. | |
Microsoft’s extensive number of cloud services and diverse portfolio for enterprises require a thorough understanding of the customer’s needs at all times so that the conversation is relevant for the customer and aligned with their interests. An AI application called Daily Recommender helps every seller at Microsoft to attain their quota by predicting the likelihood of a customer to buy products, consume services, or churn. Sellers get recommendations on each customer, including content to share, email templates, and call scripts. This approach increased the recommendation success rate fourfold, resulting in a 40% recommendation-to-opportunity rate.
Finance
Finance is undergoing a transformation, from a pure operational function to a proactive, forward-looking partner for the rest of the business. AI can play a role in that transformation, elevating finance within the organization in a number of ways (see Table 3-6).
| Use case | AI capability |
|---|---|
| Demand planning: Anticipate the demand for products or services to manage inventory, investments, or geographical distribution. | |
| Revenue forecasting: Complement traditional approaches to forecasting with AI-based methods based on historical sales data, information about deals in progress, and external conditions. | |
| Compliance: Optimize auditing practices to automatically identify high-risk transactions before they are closed and detect process anomalies. | |
| Expense reporting: Increase employee productivity for common financial processes such as expense reporting, and help automate the processing and auditing. | |
| Contract management: Understand and mine contracts in the organization, identifying contract obligations and auditing contract terms. | |
Finance was actually one of the first organizations in Microsoft to apply AI internally. The company’s revenue has been forecast with AI for several years now, with quite accurate results. Predictions are centralized and exposed with custom views for each role. For instance, sellers can gain insights on actuals versus forecasts by segments, subsegments, pricing levels, and products. After implementing AI, the variance between our overall revenue forecasts and actuals reduced from 3% to 1.5%, and this variance was reduced even more in businesses with a lot of transactions, such as those in the small or medium business space. After this success, Microsoft Finance began using AI extensively in other processes, such as compliance analytics, anomaly detection, and text analytics on global economy documents like company earnings and government reports.
Human Resources
HR is another example of a function within many companies that’s transforming from the earlier model of a reactive, process-oriented cost center to a more proactive and strategic function at the service of the business and the employees. AI can make core processes in HR more productive, so HR agents can focus on high-value activities. Table 3-7 gives some examples of core processes that can be augmented with AI.
| Use case | AI capability |
|---|---|
| Hiring: Support the identification of appropriate candidates or discover potential candidates either internally or externally. | |
| Subject matter expert (SME) identification: Support or automate the skill profiling in an organization to better match employees with projects, enable employees to connect with SMEs, or identify the right contacts for customers, projects, or products. | |
| Retention risk assessment: Predict retention risks in employees to enable early engagement and improve job satisfaction. | |
| HR analytics: Understand the organization’s health in real time by monitoring employee feedback, coming either from regular pulses across the organization or from proactive feedback. | |
| Employee support: Increase HR agent productivity by automating common HR requests and tasks with conversational assistants. | |
| Smart buildings: Improve office building management, access control, and efficiency by combining sensors and AI. |  |
At Microsoft, we aim to transform HR by providing employees with what they need in real time. For instance, we’ve created an HR bot that can automate common tasks requested by employees, like the creation of more than 5,000 travel letters every month. Employees’ offices at the Microsoft campus use AI to optimize energy consumption, do preventative maintenance, and even predict employees’ complaints on temperature conditions and act before they’re made.
IT and Software Development
Technical departments play an important role in the overall AI transformation in the company, but they can also leverage AI for their existing functions. Internal IT operations can benefit from AI in multiple processes, as illustrated in Table 3-8.
| Use case | AI capability |
|---|---|
| Security protection: Analyze activity and behavioral data inside the organization to identify bad practices and security attacks. | |
| Helpdesk bots: Provide conversational agents for employees to get automated support for the common cases. | |
| Datacenter management: Optimize the utilization, energy efficiency, reliability, and performance of on-premises datacenters. | |
Software development can also benefit from AI applied to the tools and services used by developers, as shown in Table 3-9.
Being a technology company, Microsoft has implemented many of these use cases for its own software development, and there are many others that it provides to its customers. Azure itself uses AI heavily to manage aspects such as hardware health prediction, resource optimization, performance, and utilization. AI is also infused into the DevOps processes at Microsoft, helping developers analyze their telemetry data and even suggesting appropriate fixes for issues.
In the realm of security, Microsoft uses AI to detect phishing attempts, protect against malware, and identify security threats, analyzing more than 6.5 trillion global signals each day. Microsoft Security Risk Detection uses AI to discover security bugs in software; it was responsible for finding one-third of the “million dollar” security bugs in Windows 7.
For software development, Microsoft is already making available many of the internal tools for any developer to use. The latest version of Visual Studio includes many features powered by AI for developer productivity. The AI can be trained on the vast repository of open source public code on GitHub or your own company’s code, and it provides features like smart code completion and assisted code reviews.
Customer Service
Customer service is probably one of the first functions explored by most organizations looking to bring AI into the business. They are looking for approaches that can create a better customer experience, identify customer dissatisfaction issues, and help customer agents be more productive. Table 3-10 lists some of the use cases typically addressed for customer service.
At Microsoft, we have implemented several AI-based solutions across our support organization. One of them is able to flag a case as high-risk based on multiple indicators from past interactions. This enables the agent to prepare for a call even more thoroughly when needed, or pass the case to another agent with deeper expertise on the topic. Customer interaction and feedback analysis is used to coach the support team and identify unresolved cases. Recovery managers contact those customers to ensure a resolution, driving a 180% average increase in sentiment score compared with the initial interactions.
Finally, we also provide a virtual agent that fields customer queries in our support center. The AI chat experience covers a broad range of products including Windows, Office, and Xbox. It can converse in complex multiturn dialogues that require comprehension and reasoning. It then offers instant answers and recommends actions to solve the issue, or it escalates it to a customer agent (who will be assisted as well by the virtual agent). Since its launch, the virtual agent has handled more than 100,000 support cases per day, with an increase in customer satisfaction of 31% and a 3× reduction of agent-to-agent transfers because of the intelligent routing.
Vertical Processes
We have explored many horizontal processes, with examples of how we have addressed some of them in our AI transformation at Microsoft. Because these processes are shared across industries, they are likely to also be candidates for your organization. But what about your unique processes? Those depend on your industry and your company’s unique expertise, and they have the potential to increase your differentiation and even disrupt your market. In this section, you will find real examples of customers redefining these vertical processes. Use these examples as inspiration, but ultimately finding these types of use cases for your organization requires deep exploration in partnership between technical departments and business units.
Manufacturing
Manufacturing is a great example of an industry that has been radically transformed by technology over the past century. The three industrial revolutions—steam, electric power, and electronics—were all born in manufacturing, so it’s no surprise that it’s one of the industries with the fastest adoption of the fourth industrial revolution: AI.
The ongoing digitization of manufacturing through the Internet of Things (IoT) provides real-time data that can be combined with AI to create the concept of a smart factory. However, the impact of AI can go beyond the factory and include other areas such as product design or supply chain management (see Table 3-11).
Jabil is a global manufacturing company spanning numerous industries such as consumer products, networking, and aerospace. Jabil was able to optimize its factory operations to be more competitive by connecting its equipment, sensors, and people. The company also uses AI perception capabilities to improve its processes: for example, automated optical inspection in the production line scans for any sign of defects, ensuring all potential anomalies are detected early.
In addition, Jabil analyzes millions of data points from machines running dozens of steps to predict failures earlier in the process. Using AI, the company was able to achieve 80% accuracy in the prediction of machine slowdowns or failures. Its AI solution is not only able to predict a condition, but also to explain why it was predicted, allowing Jabil to optimize its operations to avoid these conditions in the future.
Financial Services
Financial technology, or fintech, is already disrupting traditional banking. The internet and mobile devices transformed every single frontend and back office process, from deposits to small payments, loans to insurance, trading to portfolio management.
New players are entering the market, benefiting from this transformation and the low barrier of entry provided by technologies such as the cloud and blockchain. These new players are putting enormous pressure on the incumbent players, redefining the industry with new approaches such as peer-to-peer lending, crowdfunding, and mobile payments. Traditional banking now has to balance a heavily regulated and supervised industry with the agility and nimbleness of those startups disrupting the market.
AI can leverage the deep digital transformation already in place in the industry to create a new layer of differentiation from competitors and new players. A common starting point for AI in financial services is the customer-focused processes that are already being transformed. Table 3-12 gives some examples.
However, the impact of AI in financial services is deeper than the customer-focused scenarios, and has the potential of redefining core back office processes. Some examples are given in Table 3-13.
| Use case | AI capability |
|---|---|
| Fraud detection: Use AI to spot unusual patterns by analyzing multiple data points to identify fraudulent transactions that rule-based analysis may miss. | |
| Compliance assurance: Analyze structured transactions in conjunction with unstructured data such as documents, emails, or voice orders to identify internal practices compromising compliance. | |
| Trading and investment management: Optimize trade execution strategies and mergers and acquisitions analysis, augmenting agents and analysts with intelligent insights across multiple data sources. | |
| Trend identification and simulation: Mine vast amounts of unstructured data from economic reports, company earnings, news, social media, and other sources to identify trends that can impact the investment portfolio, or simulate conditions to assess a potential impact on it. | |
QuarterSpot is an online lending platform for small businesses that uses advanced models that incorporate real-time data from various sources—including business bank accounts—to assess applicant risk. Based on the results, QuarterSpot decides on whether to approve the loan and the appropriate interest rate for the risk. QuarterSpot then posts the loan in its marketplace, where investors can purchase portions of that loan in increments of $25.
For a model like this to work, QuarterSpot needed to drastically reduce the time taken to accurately predict the risk and approve a loan. To achieve both goals, the company applied AI models that were continuously improved in an agile MLOps cycle (covered in detail in Chapter 6). In just two years QuarterSpot was able to lower default rates by more than 50%, while increasing borrower approval rates by more than 15% and decreasing lending costs by 83% compared with traditional methods. Today, QuarterSpot also provides its AI models to more traditional lenders through a lending platform, creating an entire new business model for the organization.
Retail
If there’s a single sector that knows what technology disruption means, it’s retail. Ecommerce revolutionized the sector’s core services, introduced new players in the market, and forever changed customer expectations. Today, retailers have to compete with born-in-the-cloud players like Amazon, which can utilize its digital footprint to mine data on purchases and customer preferences to offer an extremely personalized and effective experience. Shoppers are now demanding similar (if not better) experiences from traditional retailers, which must create their own intelligent experiences with the added complexity of a hybrid physical/online store presence.
It is, however, within that diversity of channels that retailers are finding an interesting area to provide a differentiated value. AI can be applied to the online presence, but it can also be used to redefine the physical presence and extend it to the online sphere, offering a unique value proposition for customers through a multichannel, seamless experience.
For the online presence, there are multiple well-known AI scenarios already familiar to users. Table 3-14 gives some examples.
But by expanding these scenarios to physical stores, retailers can differentiate themselves from online-only competitors. Table 3-15 contains some examples.
Retailers can also leverage AI to optimize their operations across both their online and physical channels. Table 3-16 presents some examples.
Kroger is America’s largest grocery retailer. It’s a great example of a traditional retailer reimagining the customer experience across its physical and online channels, connecting them in a seamless experience powered by AI through the EDGE Shelf technology, a shelving system that uses digital displays instead of traditional paper tags. Kroger centralizes the data captured in its physical stores and through its web and mobile applications to optimize the customer experience in the stores, including the ability to optimize pricing in near real time.
Kroger can also provide unique experiences to its customers with this technology. The electronic shelves can use customer preference data, personalized promotions, and shopping lists to provide a unique guided shopping experience for customers.
Using video analytics, Kroger is also able to assist its store employees with restocking, notify them of incidents like spills, and even help them optimize product placement or provide custom advertisements based on customer demographics. For example, a family shopping together could receive recommendations for certain promotions on the digital shelves as they walk down an aisle.
Public Sector
When thinking about the public sector, one of the first areas that comes to mind is efficiency. The amount of procedures that the public administration has to manage for citizens and businesses is a burden that often limits the high-value services that could be provided. Repetitive procedures are a paradise for AI and present a great opportunity to obtain short-term benefits, as the examples in Table 3-17 demonstrate.
Beyond the foundational administrative services, public institutions also manage massive quantities of assets. AI can help optimize this. Table 3-18 lists some examples.
Probably the most visible scenario in which AI is applied in the public sector is to transform the relationship of a government with its citizens. Just as it can redefine the interactions between companies and customers, AI can also bring government closer to citizens in multiple ways, such as those listed in Table 3-19.
| Use case | AI capability |
|---|---|
| Conversational assistants for citizens: Use conversational interfaces to guide citizens through service requests, provide general information, or assist on city services such as transportation. | |
| Social listening: Understand citizen’s opinions, sentiment, and complaints using natural language understanding techniques on social media and any other feedback channel. | |
| Tourism promotion: Enhance the experience for tourists through multilanguage assistance, augmented experiences for points of interest, or informational bots. | |
The City of Los Angeles provides a conversational assistant called Chip (City Hall Internet Personality) that can gather and present a collection of information about any given topic or area, outline city resources and opportunities that are available to residents, and assist with filling out forms.
One of the states in Southern India, Andhra Pradesh, minimizes school dropout rates in more than 10,000 schools by using AI to predict dropouts, enabling early intervention by personnel.
Imec, a leading R&D and innovation hub in nanoelectronics and digital technologies, has partnered with Antwerp to launch a digital twin of the Belgian city. The digital 3D replica is created with real-time sensor information providing feedback on air quality, traffic, and noise pollution. AI provides a predictive view of the situation in the city where the impact of planned measures can be simulated and tested, to understand, for example, how an action by the city planners might impact traffic, noise levels, or air quality.
Health Care
Despite the infinite possibilities that AI brings to every industry vertical, health care is without a doubt the one for which it has the biggest potential to have a positive impact on society. The amount of data available in the health care sector, most of it highly unstructured, can now be leveraged at scale to redefine medicine. Table 3-20 provides some examples.
| Use case | AI capability |
|---|---|
| Image diagnosis: Augment and scale specializations such as radiology, oncology, or ophthalmology with AI applied to medical image analysis, helping with diagnosis, improving treatment and surgery planning, and bringing health care to populations without access to specialists. |  |
| Medical record mining: Analyze the vast amount of unstructured data in medical records, including test results, medical images, and doctor’s notes, to identify patterns that can help diagnose a disease and recommend the next best actions. | |
| Precision medicine: Combine AI with genomics to treat patients individually with customized treatments that are optimized for their genetic content, predicting which drug will produce the best effect for a patient. | |
| Predictive care: Detect the likelihood of a disease based on the patient’s genome, symptoms, blood samples, or any other personal information. Then proactively intervene with preventive medicine, treatments, or lifestyle changes. | |
| Drug development: Assist scientists in the drug discovery process, by either facilitating access to the vast amount of information needed for drug creation or optimizing the process of clinical trials by prioritizing potential molecular structures to focus on. | |
These use cases have huge potential to redefine medicine as we know it, but AI can also improve the most mundane processes and operations, enabling short-term results for a better patient and professional experience (see Table 3-21).
| Use case | AI capability |
|---|---|
| Personal health assistants: Use conversational AI technologies to assist patients with monitoring and treatment, provide first-level advice based on symptoms and medical history, or coach patients with chronic diseases or risk conditions on lifestyle habits. | |
| Dictation assistance: Assist health professionals with capturing documentation, using speech-to-text techniques to automatically enter the information in the electronic health record (EHR) systems. Use natural language processing to extract knowledge, such as symptoms, and assist health professionals with follow-up explorations and diagnosis candidates. | |
| Remote monitoring: Combine AI with IoT devices to perform remote monitoring of recovering patients or at-risk populations such as the elderly, people with disabilities, or chronic patients. | |
| Physical facility optimization: Optimize constrained resources in hospitals and other facilities by estimating progression of inpatients through the system, optimally redistributing staff, and predicting readmissions or bounce-backs in advance to minimize them. | |
BlueMetal, an Insight company specializing in health care solutions, worked with Steward Healthcare to predict the length of patients’ hospital stays to help doctors and nurses with schedule planning. The same system is able to use external factors such as seasonality and flu activity and other data sources such as the CDC social media to predict with 98% accuracy what their volumes will look like one and two weeks out, enabling the planning and optimization of hospital resources.
BlueMetal also partnered with Vivli, a nonprofit organization focused on sharing individual participant data from clinical trials, to work on a solution involving AI techniques to allow all this global data to be easily searched and analyzed, accelerating scientific discovery for health care.
Adaptive Biotechnologies is combining AI with the recent breakthroughs in biotechnology to map and decode the human immune system. With this technique, the company is aiming to create a universal blood test that reads a person’s immune system to detect a wide variety of diseases in the earliest stages, including infections, cancers, and autoimmune disorders.
Aurora Health Care is a nonprofit health care system with 15 hospitals and more than 150 clinics. With its conversational AI experience users can answer a set of questions about themselves and the symptoms they are experiencing; the AI agent adapts to the answers with follow-up questions, provides possible causes, and suggests a next action, including whether the patient should go to urgent care or see their primary doctor. The virtual concierge can even schedule the appointment directly from within the conversational experience.
The examples discussed here across horizontal and vertical processes highlight the breadth of use cases that business units can target in any organization. But what if we enable not only business units but every employee to create and target their own scenarios? We’ll explore that concept in the next chapter—but first, let’s take a look at Tanya’s story, which embodies this idea.