artificial intelligence

Acquiring AI Tools

AI technology for project management will test our ability to find and

implement the right solutions.

The PMO has a strategic responsibility to promote continuous

improvement in both the project maturity of the organization and the

success of projects. Implementing a new technology, such as AI, is

critical to moving forward on these responsibilities. The PMO will be

faced with many decisions, which include whether to buy tools or build

them internally and also to determine the strategy for ongoing

maintenance and support. 

important for the PMO to take a step back and assess the amount of

data that is available for AI tools. In fact, it is wise to perform some

initial data analytics on the available data as a first step. This can be

done with internal resources or with the aid of data analytics

consultants. It might be possible to gain valuable insights into project

issues based on data analysis alone. Regardless, a crucial first action

is to check the data. There were at least two major software

deployments that I worked on where the project was stalled due to

data issues. I wish the organization had taken more time before the

project started to investigate their data and fix most of the problems

before jumping into the acquisition phase. AI technology for project

management will test our ability to find and implement the right

solutions and, more importantly, to perform this activity the right way.

Acquiring an AI tool to use in a project requires a good

understanding of the data required, the process, and the desired

output. The most common machine learning algorithms use

supervised learning, which means that each dataset is labeled. Either

the performing organization or the vendor needs to have the data that

is required to feed the tool, and the amount of data and format of the

data must be clear. Insufficient data leads to the creation of a poor

model that results in a low accuracy of the output. The PMO also

needs to know if there are hidden biases in the data and how to

remedy or alter the data to mitigate this. A hidden bias might be due to

historical data that does not contain enough recent projects or project

data. Another issue is whether the labeled datasets are balanced. For

example, if there are two labels and 98 percent are one type, the

model will most likely predict that label for every new dataset. Can

unsupervised learning be used for training a model? This is an area

that is not as developed as supervised learning, so if this is being

suggested by a vendor, there must be a clear understanding of how it

works and what data is required over what duration of time.

An important factor to consider is the long-term strategy for

developing AI tools as an ongoing process. If the organization starts

with a smaller implementation, can the tools be scaled to a higher

volume or larger purpose? This is the ability to begin with one section

of the organization or a single project and then allow the tools to be

used by the entire global organization. Alternately, the strategy might

be a rollout of machine learning models from a single purpose in a

project to all aspects of project management. Regardless, the solution

needs to be scalable from a hardware, software, and data perspective.


There are only two main options for acquiring AI tools: build your own

or procure them from a vendor. The problem with buying from vendors

is that, in the current landscape, each vendor only has a solution to a

specific problem and that results in a disparate jigsaw puzzle of

isolated pieces that may or may not eventually fit together. The tools

could provide a way to improve the accuracy of project cost, maintain

the end date for a schedule, or classify risks. These are all targeted in

a specific area, and the problem is that there is no consideration for

the overall perspective of the project. AI tools need to have a holistic

perspective or integrated capability in the same way that we expect a

project manager to take that responsibility. Having specific tools is not

necessarily bad, but they can lead to blindly following outcomes that

are not tied to the overall benefit of the project. In the case of the

PMO, the piecemeal tools can have a negative impact on the overall

portfolio of projects because they are so shortsighted. On the other

hand, it is not wrong to use tools that have such a narrow focus as

long as they do not detract or obscure the big picture of achieving

project success and providing value to the organization.

Acquiring a tool from a vendor requires answers to specific

questions. Will the organization be responsible for providing all the

structured data, or will the vendor provide guidance or assistance?

The level of support in this case is important, as the data will drive

accuracy in the machine learning tools. Without proper data, the

results will be useless. The organization needs to seek answers to the

amount of data required to achieve accurate results, the level of

balanced data that is acceptable, the expected hidden bias in the

data, the learning approach being used, and how to manage it all. In

addition, there might be a requirement to create interfaces to other

systems in order to acquire and use more real-time data. A machine

learning tool should not be created from a single acquisition of data

and then ignored. Using ongoing data to provide continuous updates

is more appropriate in a project management environment, especially

with a fast-paced set of projects that are typical of the project

management function. For the PMO, this means that data from

several projects must be fed to the model, which creates the

requirement to proactively manage data movement .

اترك تعليقاً

لن يتم نشر عنوان بريدك الإلكتروني. الحقول الإلزامية مشار إليها بـ *

10 + 5 =

زر الذهاب إلى الأعلى