AI Archives - 2nd Watch

Data & AI Predictions in 2023

As we reveal our data and AI predictions for 2023, join us at 2nd Watch to stay ahead of the curve and propel your business towards innovation and success. How do we know that artificial intelligence (AI) and large language models (LLMs) have reached a tipping point? It was the hot topic at most families’ dinner tables during the 2022 holiday break.

AI has become mainstream and accessible. Most notably, OpenAI’s ChatGPT took the internet by storm, so much so that even our parents (and grandparents!) are talking about it. Since AI is here to stay beyond the Christmas Eve dinner discussion, we put together a list of 2023 predictions we expect to see regarding AI and data.

#1. Proactively handling data privacy regulations will become a top priority.

Regulatory changes can have a significant impact on how organizations handle data privacy: businesses must adapt to new policies to ensure their data is secure. Modifications to regulatory policies require governance and compliance teams to understand data within their company and the ways in which it is being accessed.

To stay ahead of regulatory changes, organizations will need to prioritize their data governance strategies. This will mitigate the risks surrounding data privacy and potential regulations. As a part of their data governance strategy, data privacy and compliance teams must increase their usage of privacy, security, and compliance analytics to proactively understand how data is being accessed within the company and how it’s being classified.

#2. AI and LLMs will require organizations to consider their AI strategy.

The rise of AI and LLM technologies will require businesses to adopt a broad AI strategy. AI and LLMs will open opportunities in automation, efficiency, and knowledge distillation. But, as the saying goes, “With great power comes great responsibility.”

There is disruption and risk that comes with implementing AI and LLMs, and organizations must respond with a people- and process-oriented AI strategy. As more AI tools and start-ups crop up, companies should consider how to thoughtfully approach the disruptions that will be felt in almost every industry. Rather than being reactive to new and foreign territory, businesses should aim to educate, create guidelines, and identify ways to leverage the technology.

Moreover, without a well-thought-out AI roadmap, enterprises will find themselves technologically plateauing, teams unable to adapt to a new landscape, and lacking a return on investment: they won’t be able to scale or support the initiatives that they put in place. Poor road mapping will lead to siloed and fragmented projects that don’t contribute to a cohesive AI ecosystem.

#3. AI technologies, like Document AI (or information extraction), will be crucial to tap into unstructured data.

According to IDC, 80% of the world’s data will be unstructured by 2025, and 90% of this unstructured data is never analyzed. Integrating unstructured and structured data opens up new use cases for organizational insights and knowledge mining.

Massive amounts of unstructured data – such as Word and PDF documents – have historically been a largely untapped data source for data warehouses and downstream analytics. New deep learning technologies, like Document AI, have addressed this issue and are more widely accessible. Document AI can extract previously unused data from PDF and Word documents, ranging from insurance policies to legal contracts to clinical research to financial statements. Additionally, vision and audio AI unlocks real-time video transcription insights and search, image classification, and call center insights.

Organizations can unlock brand-new use cases by integrating with existing data warehouses. Finetuning these models on domain data enables general-purpose models across a wide variety of use cases.

#4. “Data is the new oil.” Data will become the fuel for turning general-purpose AI models into domain-specific, task-specific engines for automation, information extraction, and information generation.

Snorkel AI coined the term “data-centric AI,” which is an accurate paradigm to describe our current AI lifecycle. The last time AI received this much hype, the focus was on building new models. Now, very few businesses need to develop novel models and algorithms. What will set their AI technologies apart is the data strategy.

Data-centric AI enables us to leverage existing models that have already been calibrated to an organization’s data. Applying an enterprise’s data to this new paradigm will accelerate a company’s time to market, especially those who have modernized their data and analytics platforms and data warehouses.

#5. The popularity of data-driven apps will increase.

Snowflake recently acquired Streamlit, which makes application development more accessible to data engineers. Additionally, Snowflake introduced Unistore and hybrid tables (OLTP) to allow data science and app teams to work together and jointly off of a single source of truth in Snowflake, eliminating silos and data replication.

Snowflake’s big moves demonstrate that companies are looking to fill gaps that traditional business intelligence (BI) tools leave behind. With tools like Streamlit, teams can harness tools to automate data sharing and deployment, which is traditionally manual and Excel-driven. Most importantly, Streamlit can become the conduit that allows business users to work directly with the AI-native and data-driven applications across the enterprise.

#6. AI-native and cloud-native applications will win.

Customers will start expecting AI capabilities to be embedded into cloud-native applications. Harnessing domain-specific data, companies should prioritize building upon module data-driven application blocks with AI and machine learning. AI-native applications will win over AI-retrofitted applications.

When applications are custom-built for AI, analytics, and data, they are more accessible to data and AI teams, enabling business users to interact with models and data warehouses in a new way. Teams can begin classifying and labeling data in a centralized, data-driven way, rather than manually and often-repeated in Excel, and can feed into a human-in-the-loop system for review and to improve the overall accuracy and quality of models. Traditional BI tools like dashboards, on the other hand, often limit business users to consume and view data in a “what happened?” manner, rather than in a more interactive, often more targeted manner.

#7. There will be technology disruption and market consolidation.

The AI race has begun. Microsoft’s strategic partnership with OpenAI and integration into “everything,” Google’s introduction of Bard and funding into foundational model startup Anthropic, AWS with their own native models and partnership with Stability AI, and new AI-related startups are just a few of the major signals that the market is changing. The emerging AI technologies are driving market consolidation: smaller companies are being acquired by incumbent companies to take advantage of the developing technologies.

Mergers and acquisitions are key growth drivers, with larger enterprises leveraging their existing resources to acquire smaller, nimbler players to expand their reach in the market. This emphasizes the importance of data, AI, and application strategy. Organizations must stay agile and quickly consolidate data across new portfolios of companies.

Conclusion

The AI ball is rolling. At this point, you’ve probably dabbled with AI or engaged in high-level conversations about its implications. The next step in the AI adoption process is to actually integrate AI into your work and understand the changes (and challenges) it will bring. We hope that our data and AI predictions for 2023 prime you for the ways it can have an impact on your processes and people.

Think you’re ready to get started? Find out with 2nd Watch’s data science readiness assessment.

Modern Data Warehouses and Machine Learning: A Powerful Pair

Artificial intelligence (AI) technologies like machine learning (ML) have changed how we handle and process data. However, AI adoption isn’t simple. Most companies utilize AI only for the tiniest fraction of their data because scaling AI is challenging. Typically, enterprises cannot harness the power of predictive analytics because they don’t have a fully mature data strategy.

To scale AI and ML, companies must have a robust information architecture that executes a company-wide data and predictive analytics strategy. This requires businesses to focus their data application beyond cost reduction and operations, for example. Fully embracing AI will require enterprises to make judgment calls and face challenges in assembling a modern information architecture that readies company data for predictive analytics.

A modern data warehouse is the catalyst for AI adoption and can accelerate a company’s data maturity journey. It’s a vital component of a unified data and AI platform: it collects and analyzes data to prepare the data for later stages in the AI lifecycle. Utilizing your modern data warehouse will propel your business past conventional data management problems and enable your business to transform digitally with AI innovations.

What is a modern data warehouse?

On-premise or legacy data warehouses are not sufficient for a competitive business. Today’s market demands organizations to rely on massive amounts of data to best serve customers, optimize business operations, and increase their bottom lines. On-premise data warehouses are not designed to handle this volume, velocity, and variety of data and analytics.

If you want to remain competitive in the current landscape, your business must have a modern data warehouse built on the cloud. A modern data warehouse automates data ingestion and analysis, which closes the loop that connects data, insight, and analysis. It can run complex queries to be shared with AI technologies, supporting seamless ML and better predictive analytics. As a result, organizations can make smarter decisions because the modern data warehouse captures and makes sense of organizational data to deliver actionable insights company-wide.

How does a modern data warehouse work with machine learning?

A modern data warehouse operates at different levels to collect, organize, and analyze data to be utilized for artificial intelligence and machine learning. These are the key characteristics of a modern data warehouse:

Multi-Model Data Storage

Data is stored in the warehouse to optimize performance and integration for specific business data.

Data Virtualization

Data that is not stored in the data warehouse is accessed and analyzed at the source, which reduces complexity, risk of error, cost, and time in data analysis.

Mixed Workloads

This is a key feature of a modern data warehouse: mixed workloads support real-time warehousing. Modern data warehouses can concurrently and continuously ingest data and run analytic workloads.

Hybrid Cloud Deployment

Enterprises choose hybrid cloud infrastructure to move workloads seamlessly between private and public clouds for optimal compliance, security, performance, and costs.

A modern data warehouse can collect and process the data to make the data easily shareable with other predictive analytics and ML tools. Moreover, these modern data warehouses offer built-in ML integrations, making it seamless to build, train, and deploy ML models.

What are the benefits of using machine learning in my modern data warehouse?

Modern data warehouses employ machine learning to adjust and adapt to new patterns quickly. This empowers data scientists and analysts to receive actionable insights and real-time information, so they can make data-driven decisions and improve business models throughout the company.

Let’s look at how this applies to the age-old question, “how do I get more customers?” We’ll discuss two different approaches to answering this common business question.

The first methodology is the traditional approach: develop a marketing strategy that appeals to a specific audience segment. Your business can determine the segment to target based on your customers’ buying intentions and your company’s strength in providing value. Coming to this conclusion requires asking inductive questions about the data:

What is the demand curve?
What product does our segment prefer?
When do prospective customers buy our product?
Where should we advertise to connect with our target audience?

There is no shortage of business intelligence tools and services designed to help your company answer these questions. This includes ad hoc querying, dashboards, and reporting tools.

The second approach utilizes machine learning within your data warehouse. With ML, you can harness your existing modern data warehouse to discover the inputs that impact your KPIs most. You simply have to feed information about your existing customers into a statistical model, then the algorithms will profile the characteristics that define an ideal customer. We can ask questions around specific inputs:

How do we advertise to women with annual income between $100,000 and $200,000 who like to ski?
What are the indicators of churn in our self-service customer base?
What are frequently seen characteristics that will create a market segmentation?

ML builds models within your data warehouse to enable you to discover your ideal customer via your inputs. For example, you can describe your target customer to the computing model, and it will find potential customers that fall under that segment. Or, you can feed the computer data on your existing customers and have the machine learn the most important characteristics.

Conclusion

A modern data warehouse is essential for ingesting and analyzing data in our data-heavy world. AI and predictive analytics feed off more data to work effectively, making your modern data warehouse the ideal environment for the algorithms to run and enabling your enterprise to make intelligent decisions. Data science technologies like artificial intelligence and machine learning take it one step further and allow you to leverage the data to make smarter enterprise-wide decisions.

2nd Watch offers a Data Science Readiness Assessment to provide you with a clear vision of how data science will make the greatest impact on your business. Our assessment will get you started on your data science journey, harnessing solutions such as advanced analytics, ML, and AI. We’ll review your goals, review your current state, and design preliminary models to discover how data science will provide the most value to your enterprise.

-Ryan Lewis | Managing Consultant at 2nd Watch

Get started with your Data Science Readiness Assessment today to see how you can stay competitive by automating processes, improving operational efficiency, and uncovering ROI-producing insights.

Analytics and Insights for Marketers

Analytics & Insights for Marketers is the third in a series of our Marketers’ Guide to Data Management and Analytics. In this series, we cover major terms, acronyms, and technologies you might encounter as you seek to take control of your data, improve your analytics, and get more value from your MarTech investments.

In case you missed them, you can access part one here and part two here.

In this post, we’ll explore:

Business intelligence (BI)
Real-time analytics
Embedded analytics
Artificial intelligence (AI)
Machine learning

Business Intelligence

Business intelligence refers to the process in which data is prepared and analyzed to provide actionable insights and help users make informed decisions. It often encompasses various forms of visualizations in dashboards and reports that answer key business questions.

Why It Matters for Marketers:

With an increasing number of marketing channels comes an increasing amount of marketing data. Marketers who put BI tools to use gain essential insights faster, more accurately define key demographics, and make marketing dollars last.

Marketers without access to a BI tool spend a disproportionate amount of time preparing, rather than analyzing, their data. With the right dashboards in place, you can visualize observations about customer and demographic behaviors in the form of KPIs, graphs, and trend charts that inform meaningful and strategic campaigns.

Real-World Examples:

Your BI dashboards can help answer common questions about more routine marketing metrics without spending hours preparing the data. In a way, they take the pulse of your marketing initiatives. Which channels bring in the most sales? Which campaigns generate the most leads? How do your retention rate and ROI compare over time? Access to these metrics and other reports can shape the big picture of your campaigns. They help you make a measurable impact on your customer lifetime value, marketing RPI, and other capabilities.

Real-Time Analytics

Real-time analytics utilizes a live data stream and frequent data refreshes to enable immediate analysis as soon as data becomes available.

Why It Matters for Marketers:

Real-time analytics enhances your powers of perception by providing up-to-the-minute understanding of buyers’ motivations. A real-time analytics solution allows you to track clicks, web traffic, order confirmations, social media posts, and other events as they happen, enabling you to deliver seamless responses.

Real-World Examples:

Real-time analytics can be used to reduce cart abandonment online. Data shows that customers abandon 69.57% of online transactions before they are completed. Implementing a real-time analytics solution can enable your marketing team to capture these lost sales.

By automatically evaluating a combination of live data (e.g., abandonment rates, real-time web interactions, basket analysis, etc.) and historical data (e.g., customer preferences, demographic groups, customer click maps, etc.), you can match specific customers to targeted messaging, right after they leave your site.

Embedded Analytics

Embedded analytics is the inclusion of a business intelligence functionality (think graphs, charts, and other visualizations) within a larger application (like your CRM, POS, etc.)

Why It Matters for Marketers:

The beauty of embedded analytics is that you do not need to open up a different interface to visualize data or run reports. Integrated BI functionality enables you to review customer data, sales history, or conversion rates along with relevant reports that enhance your decision-making. This enables you to reduce time-to-insight and empower your team to make data-driven decisions without leaving the applications they use daily.

Real-World Examples:

A member of your marketing team is reviewing individual customers in your CRM to analyze their customer lifetime value. Rather than exporting the data into a different analytics platform, you can run reports directly in your CRM – and even incorporate data from external sources.

In doing so, you can identify different insights that improve campaign effectiveness such as which type of content best engages your customers, how to re-engage detractors, or when customers expect personalized content.

Artificial Intelligence

AI is the ability for computer programs or machines to learn, analyze data, and make autonomous decisions without any major contributions from humans.

Why It Matters for Marketers:

Implementing AI can provide a better understanding of your business as it detects forward-looking data patterns that employees would struggle to find – and in a fraction of the time. Additionally, marketers can improve customer service through a data-driven understanding of customer behavior and with new AI-enabled services like chatbots.

Real-World Examples:

Customizing email messaging used to be a laborious process. You’d need to manually create a number of campaigns. Even then, you could only tailor your messages to segments, not to a specific customer. Online lingerie brand Adore Me pursued AI to mine existing customer information and histories to create personalized messages across omnichannel communications. As a result, monthly revenue increased by 15% and the average order amount increased by 22%.

AI chatbots are also making waves, and Sephora is a great example. The beauty brand launched a messaging bot through Kik as a way of engaging with their teenage customers preparing for prom. The bot provided them with tailored makeup tutorials, style guides, and other related video content. During the campaign, Sephora had more than 600,000 interactions and received 1,500 questions that they answered on Facebook Live.

Machine Learning

Machine learning is a method of data analysis in which statistical models are built and updated in an automated process.

Why It Matters for Marketers:

Marketers have access to a growing volume and variety of complex data that doesn’t always provide intuitive insight at first glance. Machine learning algorithms not only accelerate your ability to analyze data and find patterns, but they can identify unforeseeable connections that a human user might have missed. Through machine learning, you can enhance the accuracy of your analyses and dig deeper into customer behavior.

Real-World Examples:

One Chicago retailer used a centralized data platform and machine learning to identify patterns and resolve questions about customer lifetime value. In an increasingly competitive landscape, their conventional reporting solution wasn’t cutting it.

By combining data from various sources and then performing deeper, automated analysis, they were able to anticipate customer behavior in unprecedented ways. Machine learning enabled them to identify which types of customers would lead to the highest lifetime value, which customers had the lowest probability of churn, and which were the cheapest to acquire. This led to more accurate targeting of profitable customers in the market.

That’s only the beginning: a robust machine learning algorithm could even help predict spending habits or gather a customer sentiment analysis based on social media activity. Machine learning processes data much faster than humans and is able to catch nuances and patterns that are undetectable to the naked eye.

We hope you gained a deeper understanding into the various ways to analyze your data to receive business insights. Feel free to contact us with any questions or to learn more about what analytics solution would work best for your organizational needs.

How Machine Learning Can Benefit the Insurance Industry

In 2020, the U.S. insurance industry was worth a whopping $1.28 trillion. High premium volumes show no signs of slowing down and make the American insurance industry one of the largest markets in the world. The massive amount of premiums means there is an astronomical amount of data involved. Without artificial intelligence (AI) technology like machine learning (ML), insurance companies will have a near-impossible time processing all that data, which will create greater opportunities for insurance fraud to happen.

Insurance data is vast and complex. This data is comprised of many individuals with many instances and many factors used in determining the claims. Moreover, the type of insurance increases the complexity of data ingestion and processing. Life insurance is different than automobile insurance, health insurance is different than property insurance, and so forth. While some of the processes are similar, the data and multitude of flows can vary greatly.

As a result, insurance enterprises must prioritize digital initiatives to handle huge volumes of data and support vital business objectives. In the insurance industry, advanced technologies are critical for improving operational efficiency, providing excellent customer service, and, ultimately, increasing the bottom line.

ML can handle the size and complexity of insurance data. It can be implemented in multiple aspects of the insurance practice, and facilitates improvements in customer experiences, claims processing, risk management, and other general operational efficiencies. Most importantly, ML can mitigate the risk of insurance fraud, which plagues the entire industry. It is a big development in fraud detection and insurance organizations must add it to their fraud prevention toolkit.

In this article, we lay out how insurance companies are using ML to improve their insurance processes and flag insurance fraud before it affects their bottom lines. Read on to see how ML can fit within your insurance organization.

What is machine learning?

ML is a technology under the AI umbrella. ML is designed to analyze data so computers can make predictions and decisions based on the identification of patterns and historical data. All of this is without being explicitly programmed and with minimal human intervention. With more data production comes smarter ML solutions as they adapt autonomously and are constantly learning. Ultimately, AI/ML will handle menial tasks and free human agents to perform more complex requests and analyses.

What are the benefits of ML in the insurance industry?

There are several use cases for ML within an insurance organization regardless of insurance type. Below are some top areas for ML application in the insurance industry:

Lead Management

For insurers and salespeople, ML can identify leads using valuable insights from data. ML can even personalize recommendations according to the buyer’s previous actions and history, which enables salespeople to have more effective conversations with buyers.

Customer Service and Retention

For a majority of customers, insurance can seem daunting, complex, and unclear. It’s important for insurance companies to assist their customers at every stage of the process in order to increase customer acquisition and retention. ML via chatbots on messaging apps can be very helpful in guiding users through claims processing and answering basic frequently asked questions. These chatbots use neural networks, which can be developed to comprehend and answer most customer inquiries via chat, email, or even phone calls. Additionally, ML can take data and determine the risk of customers. This information can be used to recommend the best offer that has the highest likelihood of retaining a customer.

Risk Management

ML utilizes data and algorithms to instantly detect potentially abnormal or unexpected activity, making ML a crucial tool in loss prediction and risk management. This is vital for usage-based insurance devices, which determine auto insurance rates based on specific driving behaviors and patterns.

Fraud Detection

Unfortunately, fraud is rampant in the insurance industry. Property and casualty (P&C) insurance alone loses about $30 billion to fraud every year, and fraud occurs in nearly 10% of all P&C losses. Overall, insurance fraud steals at least $80 billion every year from American consumers. ML can mitigate this issue by identifying potential claim situations early in the claims process. Flagging early allows insurers to investigate and correctly identify a fraudulent claim.

Claims Processing

Claims processing is notoriously arduous and time-consuming. ML technology is the perfect tool to reduce processing costs and time, from the initial claim submission to reviewing coverages. Moreover, ML supports a great customer experience because it allows the insured to check the status of their claim without having to reach out to their broker/adjuster.

Why is ML so important for fraud detection in the insurance industry?

Fraud is the biggest problem for the insurance industry, so let’s return to the fraud detection stage in the insurance lifecycle and detail the benefits of ML for this common issue. Considering the insurance industry consists of more than 7,000 companies that collect more than $1 trillion in premiums each year, there are huge opportunities and incentives for insurance fraud to occur.

Insurance fraud is an issue that has worsened since the COVID-19 pandemic began. Some industry professionals believe that the number of claims with some element of fraud has almost doubled since the pandemic.

Below are the various stages in which insurance fraud can occur during the insurance lifecycle:

Application Fraud: This fraud occurs when false information is intentionally provided in an insurance application. It is the most common form of insurance fraud.
False Claims Fraud: This fraud occurs when insurance claims are filed under false pretenses (i.e., faking death in order to collect life insurance benefits).
Forgery and Identity Theft Fraud: This fraud occurs when an individual tries to file a claim under someone else’s insurance.
Inflation Fraud: This fraud occurs when an additional amount is tacked onto the total bill when the insurance claim is filed.

Based on the amount of fraud and the different types of fraud, insurance companies should consider adding ML to their fraud detection toolkits. Without ML, insurance agents can be overwhelmed with the time-consuming process of investigating each case. The ML approaches and algorithms that facilitate fraud detection are the following:

Deep Anomaly Detection: During claims processing, this approach will analyze real claims and identify false ones.
Supervised Learning: Using predictive data analysis, this ML algorithm is the most commonly used for fraud detection. The algorithm will label all input information as “good” or “bad.”
Semi-supervised Learning: This algorithm is used for cases where labeling information is impossible or highly complex. It stores data about critical category parameters even when the group membership of the unlabeled data is unknown.
Unsupervised Learning: This model can flag unusual actions with transactions and learns specific patterns in data to continuously update its model.
Reinforcement Learning: Collecting information about the environment, this algorithm automatically verifies and contextualizes behaviors in order to find ways to reduce risk.
Predictive Analytics: This algorithm accounts for historical data and existing external data to detect patterns and behaviors.

ML is instrumental in fraud prevention and detection. It allows companies to identify claims suspected of fraud quickly and accurately, process data efficiently, and avoid wasting valuable human resources.

Conclusion

Implementing digital technologies, like ML, is vital for insurance businesses to handle their data and analytics. It allows insurance companies to increase operational efficiency and mitigate the top-of-mind risk of insurance fraud.

Working with a data consulting firm can help onboard these hugely beneficial technologies. By partnering with 2nd Watch for data analytics solutions, insurance organizations have experienced improved customer acquisition, underwriting, risk management, claims analysis, and other vital parts of their operations.

2nd Watch is here to work collaboratively with you and your team to design your future-state data and analytics environment. Request a complimentary insurance data strategy session today!

3 Types of Employees That Can Use AI Offerings on Google Cloud

The Google Cloud Platform (GCP) comes with a number of services, databases, and tools to operationalize company-wide data management and analytics. With the insights and accessibility provided, you can leverage data into artificial intelligence (AI) and machine learning (ML) projects cost-efficiently. GCP empowers employees to apply their ideas and experience into data-based solutions and innovation for business growth. Here’s how.

1. Developers without Data Science Experience

With GCP, developers can connect their software engineering experience with AI capabilities to produce powerful results. Using product APIs, developers can incorporate ML into the product without ever having to build a model.

Let’s take training videos for example – Your company has thousands of training videos varying in length and across subjects. They include everything from full-day trainings on BigQuery, to minutes-long security trainings. How do you operationalize all that information for employees to quickly find exactly what they want?

Using Google’s Cloud Video Intelligence API, the developer can transcribe not only every single video, word-for-word, but also document the start and end time of every word, in every video. The developer builds a search index on top of the API, and just like that, users can search specific content in thousands of videos. Results display both the relevant videos and timestamps within the videos, where the keyword is found. Now employees can immediately find the topic they want to learn more about, without needing to sift through what could be hours of unrelated information.

Additional APIs include, Cloud Natural Language, Speech-to-Text, Text-to-Speech, Cloud Data Loss Prevention, and many others in ML.

2. Everyone without Data Science Experience, who isn’t a Developer

Cloud AutoML enables your less technical employees to harness the power of machine learning. It bridges the gap between the API and building your own ML model. Using AutoML, anyone can create custom models tailored to your business needs, and then integrate those models into applications and websites.

For this example, let’s say you’re a global organization who needs to translate communications across dialects and business domains. The intricacies and complexities of natural language require expensive linguists and specialist translators with domain-specific expertise. How do you communicate in real time effectively, respectfully, and cost-efficiently?

With AutoML Translation, almost anyone can create translation models that return query results specific to your domain, in 50 different language pairs. It graphically ingests your data from any type of Sheet or CSV file. The input data necessary is pairs of sentences that mean the same thing in both the language you want to translate from, and the one you want to translate to. Google goes the extra mile between generic translation and specific, niche vocabularies with an added layer of specificity to help the model get the right translation for domain-specific material. Within an hour, the model translates based on your domain, taxonomy, and the data you provided.

Cloud AutoML is available for platform, sight, structured data, and additional language capabilities.

3. Data Scientists

Data scientists have the experience and data knowledge to take full advantage of GCP AI tools for ML. One of the issues data scientists often confront is notebook functionality and accessibility. Whether its TensorFlow, PyTorch, or JupyterLab, these open source ML platforms require too many resources to run on a local computer, or easily connect to BigQuery.

Google AI Platform Notebooks is a managed service that provides a pre-configured environment to support these popular data science libraries. From a security standpoint, AI Platform Notebooks is attractive to enterprises for the added security of the cloud. Relying on a local device, you run the risk of human error, theft, and fatal accidents. Equipped with a hosted, integrated, secure, and protected JupyterLab environment, data scientists can do the following:

Virtualize in the cloud
Connect to GCP tools and services, including BigQuery
Develop new models
Access existing models
Customize instances
Use Git / GitHub
Add CPUs, RAM, and GPUs to scale
Deploy models into production
Backup machines

With a seamless experience from data to a deployed ML model, data scientists are empowered to work faster, smarter, and safer. Contact Us to further your organization’s ability to maximize data, AI, and ML.

Here are a few resources for those who wish to learn more about this subject:

Webinar: 6 Essential Tactics for your Data & Analytics Strategy
Webinar: Building an ML foundation for Google BigQuery ML & Looker

Sam Tawfik, Sr Product Marketing Manager

Maximizing Cloud Data with Google Cloud Platform Services

If you’re trying to run your business smarter, not harder, utilizing data to gain insights into decision making gives you a competitive advantage. Cloud data offerings empower utilization of data in the cloud, and the Google Cloud Platform (GCP) is full of options. Whether you’re migrating data, upgrading to enterprise-class databases, or transforming customer experience on cloud-native databases – Google Cloud services can fit your needs.

Highlighting some of what Google has to offer

With so many data offerings from GCP, it’s nearly impossible to summarize them all. Some are open source projects being distributed by other vendors, while others were organically created by Google to service their own needs before being externalized to customers. A few of the most popular and widely used include the following.

BigQuery: Core to GCP, this serverless, scalable, multi-cloud, data warehouse enables business agility – including data manipulation and data transformation, and it is the engine for AI, machine learning (ML), and forecasting.
Cloud SQL: Traditional relational database in the cloud that reduces maintenance costs with fully managed services for MySQL, PostgreSQL, and SQL Server.
Spanner: Another fully managed relational database offering unlimited scale, consistency, and almost 100% availability – ideal for supply chain and inventory management across regions and between two databases.
Bigtable: Low latency, NoSQL, fully managed database for ML and forecasting, using very large amounts of data in analytical and operational workloads.
Data Fusion: Fully managed, cloud-native data integration tool that enables you to move different data sources to different targets – includes over 150 preconfigured connectors and transformers.
Firestore: From the Firebase world comes the next generation of Datastore. This cloud-native, NoSQL, document database lets you develop custom apps that directly connect to the database in real-time.
Cloud Storage: Object based storage can be considered a database because of all the things you can do with BigQuery – including using standard SQL language to query objects in storage.

Why BigQuery?

After more than 10 years of development, BigQuery has become a foundational data management tool for thousands of businesses. With a large ecosystem of integration partners and a powerful engine that shards queries across petabytes of data and delivers a response in seconds, there are many reasons BigQuery has stood the test of time. It’s more than just super speed, data availability, and insights.

Standard SQL language
If you know SQL, you know BigQuery. As a fully managed platform, it’s easy to learn and use. Simply populate the data and that’s it! You can also bring in large public datasets to experiment and further learn within the platform.

Front-end data
If you don’t have Looker, Tableau, or another type of business intelligence (BI) tool to visualize dashboards off of BigQuery, you can use the software development kit (SDK) for web-based front-end data display. For example, government health agencies can show the public real-time COVID-19 case numbers as they’re being reported. The ecosystem of BigQuery is so broad that it’s a source of truth for your reports, dashboards, and external data representations.

Analogous across offerings

Coming from on-prem, you may be pulling data into multiple platforms – BigQuery being one of them. GCP offerings have a similar interface and easy navigation, so functionality, user experience, and even endpoint verbs are the same. Easily manage different types of data based on the platforms and tools that deliver the most value.

BigQuery Omni

One of the latest GCP services was built with a similar API and platform console to various other platforms. The compatibility enables you to query data living in other places using standard SQL. With BigQuery Omni, you can connect and combine data from outside GCP without having to learn a new language.

Ready for the next step in your cloud journey?

As a Google Cloud Partner, 2nd Watch is here to be your trusted cloud advisor throughout your cloud data journey, empowering you to fuel business growth while reducing cloud complexity. Whether you’re embracing cloud data for the first time or finding new opportunities and solutions with AI, ML, and data science our team of data scientists can help. Contact Us for a targeted consultation and explore our full suite of advanced capabilities.

Learn more

Webinar: 6 Essential Tactics for your Data & Analytics Strategy

Webinar: Building an ML foundation for Google BigQuery ML & Looker

-Sam Tawfik, Sr Product Marketing Manager

3 Ways McDonald’s France is Preparing their Data for the Future

Data access is one of the biggest influences on business intelligence, innovation, and strategy to come out of digital modernization. Now that so much data is available, the competitive edge for any business is derived from understanding and applying it meaningfully. McDonald’s France is gaining business-changing insights after migrating to a data lake, but it’s not just fast food that can benefit. Regardless of your industry, gaining visibility into and governance around your data is the first step for what’s next.

1. No More Manual Legacy Tools

Businesses continuing to rely on spreadsheets and legacy tools that require manual processes are putting in a lot more than they’re getting out. Not only are these outdated methods long, tedious, subject to human error, and expensive in both time and resources – but there’s a high probability the information is incomplete or inaccurate. Data-based decision making is powerful, however, without a data platform, a strong strategy, automation, and governance, you can’t easily or confidently implement takeaways.

Business analysts at McDonald’s France historically relied on Excel-based modeling to understand their data. Since partnering with 2nd Watch, they’ve been able to take advantage of big data analytics by leveraging a data lake and data platform. Architected from data strategy and ingestion, to management and pipeline integration, the platform provides business intelligence, data science, and self-service analytics. Now, McDonald’s France can rely on their data with certainty.

2. Granular Insights Become Opportunities for Smart Optimization

Once intuitive solutions for understanding your data are implemented, you gain finite visibility into your business. Since completing the transition from data warehouse to data lake, McDonald’s France has new means to integrate and analyze data at the transaction level. Aggregate information from locations worldwide provides McDonald’s with actionable takeaways.

For instance, after establishing the McDonald’s France data lake, one of the organization’s initial projects focused on speed of service and order fulfilment. Speed of service encompasses both food preparation time and time spent talking to customers in restaurants, drive-thrus, and on the online application. Order fulfilment is the time it takes to serve a customer – from when the order is placed to when it’s delivered. With transaction-level purchase data available, business analysts can deliver specific insights into each contributing factor of both processes. Maybe prep time is taking too long because restaurants need updated equipment, or the online app is confusing and user experience needs improvement. Perhaps the menu isn’t displayed intuitively and it’s adding unnecessary time to speed of service.

Multiple optimization points provide more opportunity to test improvements, scale successes, apply widespread change, fail fast, and move ahead quickly and cost-effectively. Organizations that make use of data modernization can evolve with agility to changing customer behaviors, preferences, and trends. Understanding these elements empowers businesses to deliver a positive overall experience throughout their customer journey – thereby impacting brand loyalty and overall profit potential.

3. Machine Learning, Artificial Intelligence, and Data Science

Clean data is absolutely essential for utilizing machine learning (ML), artificial intelligence (AI), and data science to conserve resources, lower costs, enable customers and users, and increase profits. Leveraging data for computers to make human-like decisions is no longer a thing of the future, but of the present. In fact, 78% of companies have already deployed ML, and 90% of them have made more money as a result.

McDonald’s France identifies opportunity as the most important outcome of migrating to a data lake and strategizing on a data platform. Now that a wealth of data is not only accessible, but organized and informative, McDonald’s looks forward to ML implementation in the foreseeable future. Unobstructed data visibility allows organizations in any industry to predict the next best product, execute on new best practices ahead of the competition, tailor customer experience, speed up services and returns, and on, and on. We may not know the boundaries of AI, but the possibilities are growing exponentially.

Now it’s Time to Start Preparing Your Data

Organizations worldwide are revolutionizing their customer experience based on data they already collect. Now is the time to look at your data and use it to reach new goals. 2nd Watch Data and Analytics Services uses a five-step process to build a modern data management platform with strategy to ingest all your business data and manage the data in the best fit database. Contact Us to take the next step in preparing your data for the future.

-Ian Willoughby, Chief Architect and Vice President

Listen to the McDonald’s team talk about this project on the 2nd Watch Cloud Crunch podcast.

Cloud Crunch Podcast: Data, AI & ML on Google Cloud

If you’re trying to run your business smarter, not harder, chances are you’re utilizing data to gain insights into the decision-making process and gain a competitive advantage. In the latest episode of our podcast, we talk with data and AI & ML expert, Rui Costa at Google Cloud, about why and when to use cloud data offerings and how to make the most of your data in the cloud. Listen now on Spotify, iTunes, iHeart Radio, Stitcher, or wherever you get your podcasts.

We’d love to hear from you! Email us at CloudCrunch@2ndwatch.com with comments, questions and ideas.

Top Enterprise IT Trends for 2021

Between the global pandemic and the resulting economic upheaval, it’s fair to say many businesses spent 2020 in survival mode. Now, as we turn the page to 2021, we wonder what life will look like in this new normalcy. Whether it is employees working from home, the shift from brick and mortar to online sales and delivery, or the need to accelerate digital transformation efforts to remain competitive, 2021 will be a year of re-invention for most companies.

How might the new normal impact your company? Here are five of the top technology trends we predict will drive change in 2021:

1. The pace of cloud migration will accelerate

Most companies, by now, have started the journey to the public cloud or to a hybrid cloud environment. The events of 2020 have added fuel to the fire, creating an urgency to maximize cloud usage within companies that now understand that the speed, resilience, security and universal access provided by cloud services is vital to the success of the organization.

“By the end of 2021, based on lessons learned in the pandemic, most enterprises will put a mechanism in place to accelerate their shift to cloud-centric digital infrastructure and application services twice as fast as before the pandemic,” says Rick Villars, group vice president, worldwide research at IDC. “Spending on cloud services, the hardware and software underpinning cloud services, and professional and managed services opportunities around cloud services will surpass $1 trillion in 2024,”

The progression for most companies will be to ensure customer-facing applications take priority. In the next phase of cloud migration, back-end functionality embodied in ERP-type applications will move to the cloud. The easiest and fastest way to move applications to the cloud is the simple lift-and-shift, where applications remain essentially unchanged. Companies looking to improve and optimize business processes, though, will most likely refactor, containerize, or completely re-write applications. They will turn to “cloud native” approaches to their applications.

2. Artificial intelligence (AI) and machine learning (ML) will deliver business insight

Faced with the need to boost revenue, cut waste, and squeeze out more profits during a period of economic and competitive upheaval, companies will continue turning to AI and machine learning to extract business insight from the vast trove of data most collect routinely, but don’t always take advantage of.

According to a recent PwC survey of more than 1,000 executives, 25% of companies reported widespread adoption of AI in 2020, up from 18% in 2019. Another 54% are moving quickly toward AI. Either they have started implementing limited use cases or they are in the proof-of-concept phase and are looking to scale up. Companies report the deployment of AI is proving to be an effective response to the challenges posed by the pandemic.

Ramping up AI and ML capabilities in-house can be a daunting task, but the major hyperscale cloud providers have platforms that enable companies to perform AI and ML in the cloud. Examples include Amazon’s SageMaker, Microsoft’s Azure AI and Google’s Cloud AI.

Edge computing will take on greater importance

For companies that can’t move to the cloud because of regulatory or data security concerns, edge computing is emerging as an attractive option. With edge computing, data processing is performed where the data is generated, which reduces latency and provides actionable intelligence in real time. Common use cases include manufacturing facilities, utilities, transportation, oil and gas, healthcare, retail and hospitality.

The global edge computing market is expected to reach $43.4 billion by 2027, fueled by an annual growth rate of nearly 40%, according to a report from Grand View Research.

The underpinning of edge computing is IoT, the instrumentation of devices (everything from autonomous vehicles to machines on the factory floor to a coffee machine in a fast-food restaurant) and the connectivity between the IoT sensor and the analytics platform. IoT platforms generate a vast amount of real-time data, which must be processed at the edge because it would too expensive and impractical to transmit that data to the cloud.

Cloud services providers recognize this reality and are now bringing forth specific managed service offerings for edge computing scenarios, such as Amazon’s new IoT Greengrass service that extends cloud capabilities to local devices, or Microsoft’s Azure IoT Edge.

4. Platform-as-a-Service will take on added urgency

To increase the speed of business, companies are shifting to cloud platforms for application development, rather than developing apps in-house. PaaS offers a variety of benefits, including the ability to take advantage of serverless computing delivering scalability, flexibility and quicker time to develop and release new apps. Popular serverless platforms include Amazon Lambda and Microsoft’s Azure Functions.

5. IT Automation will increase

Automating processes across the entire organization is a key trend for 2021, with companies prioritizing and allocating money for this effort. Automation can cut costs and increase efficiency in a variety of areas – everything from Robotics Process Automation (RPA) to automate low-level business processes, to the automation of security procedures such as anomaly detection or incident response, to automating software development functions with new DevOps tools.

Gartner predicts that, through 2024, enhancements in analytics and automatic remediation capabilities will refocus 30% of IT operations efforts from support to continuous engineering. And by 2023, 40% of product and platform teams will use AIOps for automated change risk analysis in DevOps pipelines, reducing unplanned downtime by 20%.

Tying it all together

These trends are not occurring in isolation. They’re all part of the larger digital transformation effort that is occurring as companies pursue a multi-cloud strategy encompassing public cloud, private cloud and edge environments. Regardless of where the applications live or where the processing takes place, organizations are seeking ways to use AI and machine learning to optimize processes, conduct predictive maintenance and gain critical business insight as they try to rebound from the events of 2020 and re-invent themselves for 2021 and beyond.

Where will 2021 take you? Contact us for guidance on how you can take hold of these technology trends to maximize your business results and reach new goals.

-Mir Ali, Field CTO

3 Steps to Implementing a Machine Learning Project

Once you understand the benefits and structure of data science and machine learning (ML), it’s time to start implementation. While it’s not an overly complicated process, planning change management from implementation through replication can help mitigate potential pitfalls. We recommend following this 3-step process.

Step 1: Find Your Purpose

It can be fun to tinker around with shiny, new technology toys, but without specific goals, the organization suffers. Time and resources are wasted, and without proof of value-added, the buy-in necessary from leadership won’t happen. Why are you implementing this solution, and what do you hope to get out of the data you put in?

ML projects can produce several outcomes contributing to decisions fueled by data and gaining insights into customer buying behavior, which can be used to optimize the sales cycle with new marketing campaigns. Other uses could include utilizing predictive search to improve user experience, streamlining warehouse inventory with image processing, real-time fraud detection, predictive maintenance, or elevating customer service with voice to text speech recognition.

ML projects are typically led by a data scientist who is responsible for understanding the business requirements and who leverages data to train a computer model to learn patterns in very large volumes of data to predict outcomes while also improving the outcomes over time.

Successful ML solutions can generate 4-5% higher profit margins, so identify benchmarks, set growth goals, and integrate regular progress measurements to make sure you’re always on track with your purpose in mind.

Step 2: Apply Machine Learning

The revolutionary appeal for ML is that it does not require an explicit computer program to deliver analytics and predictions, it leverages a computer model that can be trained to predict and improve the outcomes. After the data scientist’s analysis defines the business requirements, they wrangle the necessary data to train the ML model by leveraging an algorithm, which is the engine that turns the data into a model.

Data Wrangling

Data preparation is critical to the success of the ML project because it is the foundation of everything that follows. Garbage in equals garbage out, but value in produces more value.

Raw data can be tempting, but data that isn’t clean, governed, and appropriate for business use corrupts the model and invalidates the outcome. Data needs to be prepared and ready, meaning it has been reviewed for accuracy, and it’s available and accessible to all users. Data is typically stored in a cloud data warehouse or data lake and it must be maintained with ongoing governance.

A common mistake organizations make is relying on data scientists to clean the data. Studies have found that data scientists spend 70% of their time wrangling data and only 30% of the time implementing the solution and delivering business value. These highly paid and skilled professionals are scarce resources trained for innovation and analyzing data, not cleaning data. Only after the data is clean should data scientists start their analysis.

ML Models

The data scientist’s core expertise is in selecting the appropriate algorithm to process and analyze the data. The science in ML is figuring out which algorithm to use and how to optimize it to deliver accurate and reliable results.

Thankfully, ML algorithms are available today in all the major service provider platforms, and many Python and R libraries. The general use cases within reach include:

Classification (is this a cat or is this not a cat) using anomaly detection, marketing segmentation, and recommendation engines.
NLP (natural language progression) using autocomplete, sentiment, and understanding (i.e., chatbots).
Timeseries using forecasting.

Algorithms are either supervised or unsupervised. Supervised learning algorithms start with training data and correct answers. Labeled data trains the model using the algorithm and feedback. Think texting and autocorrect – the algorithm is always learning new words based on your interaction with autocorrect. That feedback is delivered to the live model for updates and the feedback loop never ends.

Unsupervised learning algorithms start with unlabeled data. The algorithm divides the data into meaningful clusters used to make inferences about the records. These algorithms are useful for segmentation of click stream data or email lists.

Some popular algorithms include CNN (convolutional neuro network), a deep learning algorithm, K Means Clustering, PCA, Support Vector Machine, Decision Trees, and Logistic Regression.

Model Quality

With everything in place, it’s time to see if the model is doing what you need it to do. When evaluating model quality, consider bias and variance. Bias quantifies the algorithm’s limited flexibility to learn the pattern. Variance quantifies the algorithm’s sensitivity to specific sets of training.

Three things can happen when optimizing the model:

Over-fitting: Low bias + high variance. The model is too tightly fitted to the training data, and it won’t generalize data it hasn’t seen before.
Under-fitting: High bias + low variance. The model is new and hasn’t reached a point of accuracy. Get to over-fitting first, then back up and reiterate until the model fits.
Limiting/preventing under/over-fitting: There are too many features in the model (i.e. data points used to build the model), and you need to either reduce them, or create new features from existing features.

Before unleashing your ML project on customers, experiment first with employees. Solutions like virtual assistance and chat bots that are customer-facing can jeopardize your reputation if they don’t add value to interactions with customers. Because ML influences decision-making, accuracy is a must before real-world implementation.

Step 3: Experiment and Push into Production

With software projects, it either works or it crashes. With data science projects, you have to see, touch, and feel the results to know if it’s working. Reach out to users for feedback and to ensure any changes to user experience are positive. Luckily, with the cloud, the cost of experimentation is low, so don’t be afraid to beta test before a full launch.

Once the model fits and you’ve pushed the project into production, make noise about it around the organization. Promote that you’re implementing something new and garner the attention of executive leadership. Unfortunately, 70% of data projects fail because they don’t have an executive champion.

Share your learnings internally using data, charts, results, and emphasizing company-wide impact. You’re not going to get buy in on day one, but as you move up the chain of command, earning more and more supporters, your budget will allow for more machine learning solutions. Utilize buzzwords and visual representations of the project – remember data science needs to be seen, touched, and felt.

Ensure ML and data science success with best practices for introducing, completing, and repeating implementation. 2nd Watch Data and Analytic Solutions help your organization realize the power of ML with proper data cleaning, the right algorithm selection, and quality model deployment. Contact Us to see how you can do more with the data you have.

-Sam Tawfik, Sr Marketing Manager, Data & Analytics