Artificial intelligence is everywhere. If you work in a PMO you need to be thinking about how the PMO can prepare for the impact Artificial Intelligence (AI), the impact it will have on the operational and governance work in the PMO, the projects the PMO governs, and the staff themselves. In this article, we will provide you with a step by step framework to consider as part of your preparation in the PMO.
To prepare the PMO we need to think of five key areas. If you consider each of these areas, you will take a more comprehensive approach.
- Strategy: What are your goals of the PMO? How does AI help towards this? Are you chasing a fad, or can you use it for improvement in the PMO. This depends on why your PMO exists. A governance PMO needs AI such as detecting anomalies in projects, compared to a support PMO who may need a language-based service to help people find processes And you may need a mixture of these. Much like your PMO, it is not one-size-fits-all.
- Data: Everything is about data. AI works with an algorithm that takes in your data, learns something, and builds that learning into a model. That model is what provides the answers to your questions or the predictions you ask for. Your preparation has to start with data, as the AI grows from the data. Is your data good? Complete? Clean? Do you have historical data to learn from? Is your data good for reporting now? If not, what will you feed your AI?
- People: Besides the data, it’s about the people. Many people in the PMO are manually manipulating data, manually analysing data or looking for issues and providing support. How will their jobs change? Will the AI itself be a new persona and will it be accepted? The introduction of AI is often done in conjunction with change management and your people and culture team.
- Context: Will the output of the AI be accepted? Does AI actually feel right for your organisation and managers? Is now the time? Are you mature enough? If management doesn’t trust your current reports, forget about getting them to trust your AI. A key principle of AI is that it can be trusted and is transparent.
- Future: How will you keep the AI alive? Will the right data keep coming? Will you continue to have a need for AI as part of a process? Will you redesign the AI as your PMO processes change?
With these in mind, here is a 7-step process to help you understand what to do.
Step 1: Decide on your desired outcomes
Why do you want AI? AI isn’t some sort of magic problem solver. You need solid use cases or you will spend a lot of time developing ideas that don’t make sense.
To develop your use cases, ask yourself this:
- Where can you artificially introduce experience? Consider what you need to do but can’t. Work out how AI can help you achieve the goals of your PMO. Think of it like hiring a person. What would they do? What goal is not possible right now but is expected of you that they could help with?
- Where can you introduce intelligence? This goes beyond automation. Where in the process do you need time and space to think? Where is a process requiring you to make decisions that could be automated or improved based on wider input on a scale that AI can handle?
- Where there are inefficiencies. Consider where there is a high cost in time, stress and effort.
- Consider where there are errors, and how AI and automation could help.
These questions lead you to your use cases. But you don’t have to implement them all, only those that support your PMO goals and eliminate anything “cool” that doesn’t.
Step 2: Develop your plan
AI is popping up everywhere, especially when people start using things like ChatGPT. This may not be the most thought out approach for you. Implementing AI into your PMO is either going to be a dedicated project, or part of a wider PPM solution rollout. These need to be planned as a project. But don’t treat this as an IT project. AI is the tool which to support your process, people and strategy. So, treat it as a process project with a large change management component.
Step 3: Redesign the process to use AI
Inconsistently using AI as a bolt on tool will not lead to sustainable benefits. It needs to be introduced into the process. This comes back to your use cases. Examples of processes are using AI to create estimates when business cases are created; to help you justify the business cases; to find lessons before you begin etc. But these are steps in a process. You need to redesign the process to use the input and output of AI. This may lead to a change in the roles that use the process.
Step 4: Get your data ready
If you don’t have data, or that data isn’t clean and representative, then your AI will fail or be biased. Make sure there are sensible relationships in the data. The data needs to represent each process and you need to ensure you have permission to use the data. Consider how it was captured and how it will keep being captured. If it is once off data, your AI isn’t going to last. Consider how legacy data may have poor controls around it and be in legacy systems. This makes it worth less as an input and possibly harder to get to.
Most of your effort is going to go into preparing your data. Organisations are introducing a data supply chain officer to ensure they have sufficient information for the AI.
It is important to consider:
- Is the data central and ready reporting? If you don’t have a successful PPM solution with data you can trust then you are not ready to add AI to it.
- Do you have all the status report data?
- Is the data readable by someone with no assumed knowledge?
- Is the data full of shorthand, codes, and abbreviations?
- Does every project have the same volume and type of data? If not some will influence the AI more.
A positive side effect of this is that your PPM solution reporting is going to be so much better.
Step 5: Build the solution
This is where it gets more technical. There are specific processes you use to implement AI, and whoever is helping you with it will lead you through this.
What you build depends on the tool you select and how it is integrated into your PPM solution. It isn’t all ChatGPT. That is one specific approach for a specific set of problems.
Let your requirements drive the selection of and configuration of the tool. Working backwards from an interesting tool is a classic IT mistake.
Step 6: Train it
The tool is generic. You can’t just jump in and go live. You need to train the tool and let it learn from your data. This forms its experience, and then it can provide a useful service using new data in the future. Don’t underestimate the effort required.
You need to have a set of good data that you can use to train it. For example, if you want it to predict a project going red, give it examples of that. You also need a set of data that you can use to test it.
As you prepare you are going to find you need to re-prepare data until you get a level of accuracy you are happy with. You will adjust the AI model and the data a few times. The AI is training you at this point to understand it and give it the data it needs.
Step 7: Prepare the people
People need to know how to work with AI, and you need to ensure that the introduction of AI does not alienate your people.
- Think of the persona. The AI doesn’t have to be a piece of software, it can be thought of as a persona doing a job and this may help integrate it into the process.
- Manage the change and educate your employees about the changes AI will bring but do it so that you foster a positive mindset around AI. Show them examples and how it will help them succeed.
- Reskill and upskill your staff. Teach them about AI and how to use it.
- Help your staff understand the importance of data, and how data is the asset that together with the AI provides the value. Therefore, they need to keep data secure and not give it to AI tools outside of your control.
- Think about how performance metrics may change and how you measure humans versus AI.
Once you have your solution in place, keep focusing on the data and the people.