Applied Artificial Intelligence: An Introduction For Business Leaders

by Mariya Yao

Ask them if they have run into problems with scalability, stability, security, compatibility, or ease of use.

Competitive Landscape

Building a fully customized, in-house solution means you can build an end-to-end system that addresses all required aspects of your business workflow and technical integrations.

In transitioning from buy to build, an important consideration is whether there exists open-source software that will allow you to affordably customize and extend pre-built features without paying hefty costs to a third party. Open-source solutions exist for virtually all aspects of a machine learning platform, ranging from solutions that manage your data layer, model development and management, and higher-level analytics and reporting.

Because the scope, ongoing

maintenance costs, and maturity level of AI technologies vary widely, it is difficult to produce a generalized methodology to quantify ROI for AI.

An alternate measure may be to examine how AI technology unlocks business value. Typical metrics center around tangibles such as increased revenue and decreased costs, as well as intangibles ...

Increasing ...

AI can identify new potential customers for the sales team, facilitate personalization to improve conversion rates and decrease churn, and power customer service bots to provide higher quality service and generate repeat business.

AI may also allow you to offer new products or services that weren’t previously possible.

Decreasing Costs

Measuring the ability to reduce costs is another popular way to assess returns on AI investments.

Keep in mind that automation generally only eliminates a fraction of an employee’s responsibilities, so be sure not to overestimate potential gains.

Measuring ROI

Machine learning solutions do not need to be perfect to provide value.

Portfolio Approach

View these early investments as research and development (R&D) ventures and assume that lots of failures will accompany

each success.

Pick the Right “True North” Metric

Choosing a more specific metric helps you to qualify your AI strategy and also check that your decisions align and advance your business and technology in the right direction.

Is this a metric everyone can understand?

You want to align your entire company,

Be sure your true north metric is an accurate proxy for success.

Is this a leading or a lagging indicator?

Is this a relative or absolute metric?

Is this metric actionable?

Is the metric tracked and measured correctly?

Is our true north metric really aligned with our business goals?

Once you have confidence that a metric is productive, however, it's best to stay consistent until you have strong reasons to optimize for a new goal. Constantly switching true north metrics will confuse your team and hinder your execution.

data is not reality.

Even with the same dataset, two people can form vastly different conclusions.

This is because data alone is not “ground truth," which is defined by machine learning experts as observable, provable, and objective data that reflects reality.

Your ability to solve a problem with artificial intelligence depends heavily on how you frame your problem and also whether you can establish ground truth without ambiguity.

What people call “data” can be carefully curated measurements selected purely to support an agenda, haphazard collections of random information with no correspondence to reality, or information that looks reasonable but resulted from unconsciously biased collection efforts.

Gathering huge but messy volumes of data will only impede your future analytics,

You will need to discuss your expectations and set appropriate parameters in order to collect the information that you actually want.

Mistakes here can result in capturing incorrect or accidentally biased data.

Measurement errors occur when the software or hardware that you use to capture data goes awry, either failing to capture usable data or producing spurious data.

Common errors include missing a particular filter that may have been used on the data, such as the removal of outliers; using different accounting standards, as in the case with financial reporting; and simply making calculation errors.

Coverage error describes what happens with survey data when there is insufficient opportunity for all targeted respondents to participate.

Sampling errors occur when you analyze data from a smaller sample that is not representative of your target population.

Two types of inference errors can occur: false negatives and false positives.

Assuming you have a clean record of ground truth, calculating inference errors will help you to assess the performance of your machine learning models.

The foundation of artificial intelligence is data.

Accuracy gives the percentage of classifications that were correctly made.

precision tells us the model’s ability to correctly classify the instances that we care about in a dataset.

Recall measures the percentage of true outcomes that were correctly classified as being true. In other words, recall characterizes a model’s ability to identify all of the instances that we should care about in a dataset.(80)

There’s a trade-off between optimizing the precision and recall of a model. The nature

of your task will determine whether it’s more important to maximize precision, to maximize recall, or to achieve a balance between the two.

Emphasizing recall minimizes false negatives,