These articles about AI and ML project design are intended to address the
real
difficulties
people experience when trying to define and scope their projects.
Project
design issues
are widely regarded as a key cause of project failure. So to have the best chance of
success,
read our tips before you start, and review them regularly during your project.
These articles are organised as a set of questions you should ask yourself about each aspect of your
project. Most importantly, you do not need to be an AI or ML expert to answer the questions.
We also provide a free tool - a kind of Business Model Canvas for AI and ML projects:
Project designer
Using the tool, you can record your answers to these questions.
List teams, systems and individuals whose responsibilities will be affected by deployment of the solution.
The number of Stakeholders may be larger than your project team. You don't need to have them at every meeting - but you do need to work with them to ensure your project doesn't create problems for them. There's another question below which identifies key SMEs who will need to be regularly providing input to your project. The aim of this question is to produce a list of everyone and everything potentially affected by, or benefiting from, your solution.
Teams
Teams affected might be employees or customers who either consume solution outputs, will contribute to the operation or the solution, or use it directly. They might be a couple of steps "downstream" from the solution; or perhaps they are "upstream", and produce data used in the solution where new Quality Control processes will need to be applied. Usually you can find representatives of these teams to include in your stakeholder group.
Systems
Systems might be synonymous with some teams, but not always. There may be systems which are highly automated and used by many groups, which would have to consume outputs from your solution - presumably replacing existing outputs from the old solution or process. It's important to identify these systems affected to have a good understanding of the potential risks and consequences of deploying your solution.
Individuals
There may also be people who aren't directly going to use or be affected by data from your solution, but who are still affected by the risks or workload the solution will generate. For example - IT departments; Chief Information Officers (CIOs) - who are responsible for data security; Chief Technology Officers (CTOs) - because you're going to be introducing or using key new technologies and platforms; and support managers, because your deployment will change employee or user experience and generate new and potentially more numerous tickets!
Champions are crucial to maintaining project momentum and have a clear vision of the solution. Who is your Champion? Note: The champion is often also a stakeholder.
Simply list some candidate Champions in your response.
We suggest that having an explicit champion for your project is very important. The Champion should be invested in your project to motivate them to keep pushing for its success. Often, this motivation is because the Champion is also a key stakeholder - but not always. They might simply be passionate about it. Either way, the Champion will be responsible for maintaining momentum.
Another role of the Champion is maintaining a clear vision of the solution and the direction of the project. You might also call the Champion a Project Director, but the role is more activist than a normal director. There will be many choices and compromises to make; the Champion ensures that these choices don't lose track of the project value proposition.
The Champion might also sponsor the project (giving it budget) but this is not always the case.
SMEs know key data, systems, or business-processes better than almost anyone else. List SMEs whose knowledge and experience will be important to your project.
A good AI/ML project will usually cross many domains and disciplines. Typically, the experts in Business Analysis, Data Science, AI and ML won't know anything about the problem you're solving. That's where Subject Matter Experts (SMEs) come in.
Usually, the SMEs won't know much about data science or ML. But that's OK - as long as the various experts talk to each other a lot!
That's where having explicit SMEs on the project team comes in. SMEs should understand the problem or opportunity deeply. They should understand how existing processes work, their strengths and weaknesses, and the problems often encountered. They might also know about ideas that have been tried in the past, and failed. All this wisdom and experience will guide the project.
The aim of this question is to identify a handful (at most) of SMEs who collectively really deeply understand the area the project is working in. They should be regularly involved in the project throughout its duration, helping to come up with ideas to solve problems and giving feedback on the quality of results. The SMEs will also help to identify data sources, and understand the characteristics of these data. Finally, SMEs will also know who or where to ask for additional information.
Estimate the size and composition of the project team, including:
Note that in smaller projects, one person might fill multiple roles. Answering this question may require help from AI/ML and software development experts. That's OK - feel free to guess and refine the team later.
From experience, we can imagine some different teams of varying scale.
A minimal team might be two individuals:
Like any good movie, there is always a protagonist who does things, and an antagonist who reviews and directs the work and understands the vision of where you need to get to. That separation is important to keep the project on a good track as the technical situation evolves.
A small team for a PoC (Proof of Concept) without production delivery objectives might look like this:
Let's say you do want to carry the solution to production. A small team including development of production software might additionally have dedicated software engineering resource[s]:
If you require strong maintenance and support guarantees (e.g. Service-Level Agreements), you might want to add some hours or a resource for DevOps or technical support.
Teams can grow much larger; it's not unusual for larger projects to occupy 10-20 people for months. But if you've got a team that big, you don't need to read this article - you'll have enough experience on tap.
Mentoring
One practice we often see on complex projects is a "senior" data-science/AI/ML specialist who reviews the work of a "junior" data-science/AI/ML practitioner or software engineer. It's important to have regular reviews of methodology both within and across disciplines. This helps to de-risk the project and it helps less experienced team members to grow.