Aged just 47, Andy MacMillan has already been a chief executive for nigh on a decade. The new head honcho at AI and data analytics company Alteryx, MacMillan began his career as a developer at Electronic Data Systems, the IT outsourcing company spun out of General Motors. It was there he cut his teeth building websites in the dotcom boom, before earning his MBA and joining Stellent as VP of Product Marketing. “Product management is a crazy job,” says MacMillan. “You are right in the middle of the decision-making process, and you’re responsible for everything but in charge of nothing. I loved that job.”
MacMillan would end up assuming even more responsibilities, becoming a vice-president when Stellent was bought by Oracle and running a billion dollars of business by his early 30s. Opportunities were plentiful. MacMillan soon moved to Salesforce, becoming GM of its Data.com division, before hopping over to Act-On Software for his first stint as a chief executive.
After a period helming UserTesting, Alteryx came calling, setting MacMillan the challenge of making the company the go-to enterprise platform for data-driven AI workflows and agents – to be achieved, he told Tech Monitor in the following interview, edited for length and clarity, through its AI Data Clearinghouse. “Thinking about my interests as a kid, it is no surprise this has been my career,” he says. “I always liked technology and building things, but I had an entrepreneurial streak, too. So, ‘tech CEO’ was not an unlikely outcome. For me, the business side is just as interesting as the tech side.”
Tech Monitor: What most excited you about taking over as Alteryx CEO?
Andy MacMillan: The role of CEO takes up so much time, so I had planned to take some time off to rest. I have three kids, and I’m good at being involved in their lives, so I wanted to take time to think about my next move. At UserTesting, we rolled out an LLM across the company and were solving problems with AI – I built some custom GPTs myself – and I could see how it would change the way companies operate.
AI was good at using creative content to make more creative content – you can train it on marketing material to write marketing material – but it wasn’t good at using data to identify which service teams were suffering a productivity challenge. Our LLM provider told us we had to write lots of custom handlers or make hundreds of API calls. Ultimately, I wanted to find a way to solve the problem of there being limited LLM use cases without lots of fine-tuning. It was then that I got the call from Alteryx, which has a data workflow tool wherein business customers don’t write code, they just drag and drop data, thereby turning it into metrics AI can access. I also had the chance to hire a new management team to pursue a new market opportunity, so I ended with only two months off, not the six I had planned.
What skills does the role require, and how did your previous experience prepare you for it?
People often ask me how I got to be a CEO at my age, and I’ve been fortunate along the way. I got to be very senior very young. The first people I managed at Oracle were other VPs who were older than me. I set them missions and gave them what they needed to get them done. I’ve never managed frontline employees, only senior leaders, so I had to work hard to get their respect.
The job is not about showing people how smart you are or telling them how to do their job. It is about finding out how I can help them do their job. The CEO is not the head of sales or finance or any other team, but aligns all those folks to hit the numbers. I operate between all the functions, keeping teams aligned and making sure they can talk to me and each other about any problems.
Empowering senior leadership teams to be successful is an important skill set. I spend a lot of time on alignment using the V2MOM model: vision, values, methods, obstacles, and measures. I advise writing down the missions and the obstacles to accomplishing goals, and letting leaders tell you about their goals and challenges. My job is to facilitate conversations across the teams to achieve alignment. It is not about solving each individual team’s problems. I am not a go-between, and I am not good at that. I bring the teams together. I hire smart people and get them aligned, which is 90% of the challenge.
People talk more and more about agentic AI, but do you think there is a widespread understanding of what it is and what it can do?
We are starting to see the development of a better understanding of what agentic AI could be. Business leaders are starting to imagine a world where it can do a lot of useful things for them, but they currently have no idea how to get there, at least not in such a way that they can trust its outputs. We are being sold on the idea of running before we can walk or crawl.
At Alteryx, we are trying to give people visibility, through the AI Data Clearinghouse, of how agentic AI can fit into the workflow, so they can understand its potential, even if they have no idea how to start. Maybe they want an agent that can help them close the books every month, rather than just a tool that can write them a slightly better email. That is the kind of progress we are talking about.
What are the biggest opportunities and the biggest risks with agentic AI?
For some people, the biggest meta-risk stems from the fact that whenever there is major technology disruption, there is a shift in the kind of jobs we do. For example, when we automated making bricks, we no longer needed brickmakers and started needing more bricklayers. We have to look closely at how agentic AI will change jobs, whether we are moving up the value chain, and how we need to retrain and upskill.
We should all be thinking about the impact of AI. It will certainly be disruptive but, on aggregate, there are very few teams in any workplace where people don’t want more hands so they can get more done. There is always a way to absorb a big productivity lift of the kind that agentic AI can offer. So, we can be optimistic about productivity gains while also being aware of disruption, and we can help people upskill.
With that in mind, what impact do you think agentic AI will have over the next five to ten years?
It will reduce the amount of time businesses spend on data entry and retrieval. That will certainly change, and we will feel that the technology is working for us, rather than the other way around. After all, data entry means we are working for the robots, feeding them data. When the robots are giving us useful prompts to make us more productive in sales, finance or any other business line, then they are working for us.
Ultimately, agents will be able to use your data to reach out and engage with your employees, and I think we will get there fairly quickly. We’ll get to a world where even the Luddites appreciate how AI is helping us. Any survey of employees will show that people don’t like doing the admin stuff. They want to do the parts of the job they enjoy.
With any AI, but particularly the increasing levels of autonomy of agentic AI, governance and risk management are important considerations. How can companies achieve the right level of oversight and accountability?
That is part of what we are trying to address with the AI Data Clearinghouse. A business will say it wants to be an AI-first company, but then say none of its data will go into the AI. We enable business users to build workflows that are visual in nature, so anyone can look at a workflow that’s plugged into AI and quickly understand the data involved in the process. You can build workflows by putting some data into an LLM or agentic AI, but then change your mind and take that data out. You can work to find what data is helpful. Companies that take that kind of approach will be successful with AI quickly. That is the crawl and the walk. It is important that you can see how it works, because AI can’t just be a black box. Companies that simply load all of their data into AI will learn some very hard lessons, very publicly.
Read more: The philosophical barrier to corporate AI adoption is real, says Box’s Yashodha Bhavnani
