The two main components of AI that will have the greatest impact on
project management are machine learning and natural language
processing. A machine learning algorithm uses a mathematical
formula based in calculus that attempts to find the least error between
correlations in the data. This is also known as minimizing the cost
function. Based on the correlations, the AI tool creates a model that
can be used to classify or make a prediction based on a new set of
data. A machine learning tool is written in software code and often
uses utilities, or precoded functional blocks of code, to create
decision-making algorithms, the most popular and effective one being
a neural network. Learning occurs in several ways, but the most
common are as follows.
Supervised learning is when a dataset is labeled and the algorithm is
trained to correlate each dataset with the labeled result. The algorithm
is capable of modifying itself until it has the most accurate model. It is
then used on test data to verify the accuracy of the model. For
projects, we can also label our datasets. There are successful
projects, well-designed risk plans, and communication plans that
result in high stakeholder satisfaction. There are also negative results
for each example. Supervised learning is commonly used in the field
of health care to diagnose X-ray results and can provide higher
accuracy than a trained technician.
5 The algorithm is trained on X-ray
images labeled as either clear or showing evidence of a condition. A
new X-ray image with a unique pattern is used as input to the
algorithm and it diagnoses or predicts the result.
Unsupervised learning is when the data is not labeled but has a
sufficient number of clues so that the algorithm is able to classify the
data effectively. If the data indicates that an object has leaves, has a
trunk, and has branches, then the algorithm will correctly classify it as
a tree. The main benefit of unsupervised learning is clustering. The
algorithm has no labels and simply groups similar items together. How
can this be applied to projects? Risks can be clustered or grouped so
that when one of the risks from a cluster occurs, there is a strong
possibility that a similar risk will occur.
Reinforcement learning is when the algorithm learns through trial and
error to make proper predictions. A common example is typing on a
smartphone, with the full word appearing after you’ve entered only two
or three letters. The program learns your pattern through a series of
repetitions in which you did not select the word it suggested because
that was not the word you wanted. How can this apply to project
management? All algorithms must be updated on a regular basis to
stay accurate and this is one of the methods to do so.
There are several different methods used to create a machine learning
algorithm, including a neural network, random forest, support vector
machines (SVM), and a Naive Bayes classifier.
6 The algorithm creates
a model based on historical data and then uses the model to make a
classification or prediction based on a new set of data. Another basic
capability of AI is NLP, which is a computer program’s ability to
interpret human language and classify communication into a meaning
or, as it is called in NLP, an intent. This includes the ability to interpret
emotion behind the words, which is a skill known as sentiment
analysis. NLP is also used to search documents and extrapolate
meaning and to determine correlations and anomalies. NLP-based
algorithms can be integrated into project management tools. For
example, NLP can be used to search the text of a document and
compare it to documents from similar projects. If the documents from
similar projects are labeled as either accurate or incomplete, NLP
results can be used by a machine learning algorithm to classify a
current document based on the labels. NLP can also be used in scope
documents or project proposals to search for errors or inconsistencies.
NLP is used to create models that allow a computer to interact
with a human who is speaking. When humans talk or when they
create written text, such as an instant message or an email, it is called
an utterance. NLP uses utterances to determine a positive or negative
sentiment with regard to a project environment. NLP can also be used
to uncover utterances that reveal the threshold levels of stakeholders.
In other words, the project status in terms of a late schedule, for
example, can be above the tolerance level for the project customer or
project sponsor who has the authority to ask for a complete
justification and analysis before further funding is provided.
An intriguing benefit of a machine learning algorithm is that the
same basic algorithm can be used with different data to achieve a new
objective. Pedro Domingos uses an analogy to explain this in his book
one opposable thumb—and yet it can make and use an infinite variety
7 Machine learning tools allow us to deliver increasingly larger
and more complex projects and still achieve a successful result. There
is a significant difference between regular programming code and
machine learning algorithms. Normal programming code must ensure
that all possibilities are taken into account in the code. Typically, if a
particular event happens, then the code performs an action. Regular
software programs are filled with logic that is required for every
possible event; if something unforeseen happens, the code will stop.
With machine learning code, the program uses data to arrive at a
probability and then uses that result to perform an action. If a new
event occurs, the code uses the model based on data to make a
decision and continues. There is logic in the code but far less is
required than for non-AI-based programs.