How not to fail at your AI CoE

The idea of a Center of Excellence is not new. The world has seen many successful and non-successful CoE attempts for various needs. When they are successful, they create both governance and enablement within the business. When they fail, they do neither well… and the business just keeps doing what its doing. The heart of this question is about “how do I gain ground in AI in an environment that needs both protection and success?” Enter the AI CoE… the idea that a group can force multiply the organization to achieve success. Let’s explore how we achieve success and how to avoid common mistakes.

What are the common mistakes?

I’ve seen these common mistakes when starting an AI Center of Excellence:

1. Thinking AI success means Data Science training
2. Putting onerous controls that limit forward progress
3. Spending too much time in “ivory tower” vs. on the business
4. Not considering the “end-in-mind” and achieving movement
5. Architecture for architecture-sake
6. Failing at effective training patterns
7. Not enough Executive Support (trying to run it from IT)

What does an AI CoE need to be successful?

1. Executive support for your AI strategy. The AI strategy is NOT an IT project or initiative. It isn’t like adopting the cloud or deploying Office 365. AI is an organizational objective that needs to be prioritized from the business leadership down. It reflects a change of every individual in the company and should impact every part of the organization. Do not make the mistake of trying to launch a CoE that does not have top-down support from the business. If you do, it will not achieve the gains you are hoping for.
2. A clear mandate for ROI. The CoE’s goal is to govern and enable AI throughout the business. Too often CoEs get stuck in the governance part and operate in a theoretical understanding of what this means. The successful CoEs have a healthy paring of governance with value. The governance’s goal is to accelerate innovation (AI) by applying a smooth road with guardrails to prevent spinning out.
3. Investment. I’ve seen some organizations totally underfund their AI efforts, or give up too quickly when an idea needs more work to reach success. We are in an early period of companies needing to experiment. You should absolutely build your efforts around achieving value, but be careful about underfunding the same effort or not following through when more work is necessary. You need to capture and keep momentum toward the goal of accelerating every person with AI.
4. Ability to affect change. Your AI CoE needs to have the ability to create change within the organization. This might seem obvious, but some are simply not capable, are understaffed, or do not have the organizational support to move efforts forward. In some cases that ability to affect change might include the ability to actually fund, or partially fund solutions within the business. In a sense, like an investor to gain ground across the organization.
5. Partnership. The AI CoE can’t do it without engagement from the business. They need to be both effective at creating value and at fostering partnerships to create positive outcomes. You can see how without #1 (Executive Support) you might not get as far as you like with #5. This whole program needs to be a priority for the business, which fosters partnership, which fosters engagement and outcomes.

How do I measure an AI CoE?

You need to start with the “end in mind”. What are you trying to achieve? The critical elements to measure an AI Center of Excellence on are:

1. New revenue in real dollars. Did your AI Center of Excellence result in creating new value for the organization in the form of business capabilities brought to market that results in real payoff? There are too many opportunities in AI to chase those that don’t have value.
2. Operational savings in real dollars. How did the efforts result in real dollars saved in current or future costs? For instance, a recent project automated the processing of 500,000 invoices yearly. This saved many tens of thousands of person hours per year, resulted in 5x more efficient collections, and improved customer relationships.
3. Employee time saved in hours. The optimization of employee time translates into dollars, but even more is understanding how many AI capabilities, especially commodity AI, relates into real impact on people. Microsoft has already embedded in Copilot the ability to measure impact of the deployed estate and complemented that with initial research. Measuring that direct impact in relationship to key scenarios being tracked is going to be key to achieving success and enabling others.
4. Employees using commodity AI. There was a point where we didn’t know the internet was a thing. Many of us lived through the moment where it was and those companies that harnessed it quickly were able to accelerate their growth. This will have a similar (if not more pronounced impact) on every person. A company should measure the extent to which every employee is able to become capable at leveraging AI to achieve more, as well as unlocking human skills that build on their unique capabilities as a person (not as a machine).
5. AI projects running in production. The holy grail is getting mission-driven AI projects into production and seeing value. I don’t mean POC or Pilot. I mean getting all the way to a production deployment that is supported via and delivering regular value to the business. The successful AI CoE will enable this. Note that this is inclusive of ML Ops. Also note that I don’t have a measurement directly on building or implementing governance or training. Those are traps. Focus your team on the end, not the steps in between.
6. Guardrails around commodity AI that enables access to the right information and accelerates semi-custom development, while protecting critical information. This often means better information governance than the organization does today and rolling out critical ops capabilities for even semi-custom platforms when they hit production.

So… what to do? How do I get this off the ground?

Focus on the following ideals as you navigate getting started.

1. Start with envisioning… “what does the mission of my organization look like through the lens of AI?”. Every organization needs to ask themselves this question to consider both the threat and opportunities that AI provides.
2. Challenge your organization to enable AI for every employee. Stop and think… are there truly any employees in my business who do not deserve to have AI as an asset?
3. Invoke new skills in every employee to “be more”. This is an important time of transition for every team member on your staff. Your AI CoE has an accountability for that transition
4. Measure the engagement of AI impact for every employee. You will know the results of your efforts by the fruits. Are your employees actively engaged and leveraging AI as part of their daily work? If so, you’re on the right track. If not, they might need some help getting there.

The creation of a business sponsored AI center of excellence is a critical asset in the journey of every organization. Apply intentionality and human-centric focus to arriving at success.