Alexander Sukharevsky is the managing partner of QuantumBlack, McKinsey’s AI arm. Under his leadership, QuantumBlack has expanded within the firm to emerge as a recognized leader in applied artificial intelligence, helping organizations undergo end-to-end transformation, redefine their business models and improve performance through the responsible use of AI and technology.
In addition to his QuantumBlack work, Alexander is a leader of McKinsey’s research programs into areas such as generative AI, future technologies, and the use of technology for the common good. He publishes his research frequently, including several reports on the future of work and the influence of the Internet, information, and communications technology on economies. Alexander also serves on McKinsey’s Technology Council, which gathers top global technology leaders and McKinsey experts to assess and to track trends in business and technology. Prior to his leadership role of QuantumBlack, he led McKinsey’s Technology Practice in Europe, the Middle East, and Africa. Sukharesky, a speaker at the VivaTech global conference in Paris June 11-14, announced a new product suite called agents at scale – and launched a new report on seizing the agentic AI advantage. In an interview with The Innovator he spoke about CEOs can use agentic AI to make a difference in their bottom line.
Q: What are the key takeaways from QuantumBlack’s new report on seizing the agentic AI advantage?
According to McKinsey’s most recent Global Survey on AI1, more than 78% of companies are now using gen AI in at least one business function, up from 55% a year earlier. Nearly eight in 10 companies report using gen AI—yet just as many report no significant bottom-line impact. At the heart of this paradox is an imbalance between “horizontal” (enterprise-wide) copilots and chatbots—which have scaled quickly but deliver diffuse, hard-to-measure gains—and more transformative “vertical” (function-specific) use cases—about 90% of which remain stuck in pilot mode.
AI agents offer a way to break out of the gen AI paradox. That’s because agents have the potential to automate complex business processes—combining autonomy, planning, memory and integration to shift gen AI from a reactive tool to a proactive, goal-driven virtual collaborator that can turbocharge operational agility and create new revenue opportunities. But unlocking the full potential of agentic AI requires more than plugging agents into existing workflows.
Q: What will it take for companies to see an impact on their top line?
AS: Companies need to take time to step back and look at the processes end-to-end and think -considering the technological capabilities currently in place – how to reinvent them instead of just infusing technology into an anachronistic process. It is not trivial. They also need an orchestration layer between various systems, because data doesn’t sit in one system, it sits in many systems. And then they need a proper architecture and to do some change management. Failure to have all of this in place is why companies are not seeing an impact on their bottom line. It is why we are introducing agents at scale.
Q: What exactly is agents at scale?
AS: What we trying to do is literally take one hundred years of McKinsey’s domain experience and QB’s experience working with clients and provide a ready-to-go depository of reinvented processes that companies can take and build on to implement these new technologies, using agents within each process. But if you and I have the same components, the same agents, what’s your secret sauce? What we allow is for companies to also create their own agents that will reflect the DNA of their company, their commercial acumen and their proprietary data. This is what we do at the content level. Now, if we go to technology, there are more issues that we aim to help companies solve. Issue number one is the architecture. How do you ensure that you’re able to bring the latest and greatest as it develops and input it right into your IT architecture and landscape? Agents at scale offers flexible modules that can be swiftly tailored to specific industries, workflows, and existing IP. We are also solving other issues by creating the right orchestration layer between different ERP [enterprise resource planning] and other systems in your organization to get the data from all of them; offering advanced features for coordinating multiple agents, ensuring they work together efficiently, and a framework that promotes the reuse of components and adherence to industry standards, enabling faster deployment and scaling.There are no out-of-the-box solutions for all of this so the idea is to help companies create architecture orchestration on top of a great library of processes and agents, and then last, but not least, deal with organizational change and the human side of the equation.
Q: There are a lot of companies that are selling agentic AI products and services to large corporates. What is McKinsey’s differentiator?
AS: What we are introducing is the orchestration between various agents coming from various systems, avoiding suppler lock-in. While there are a lot of technology players that provide technology solutions, none of them have a hundred years of domain experience like McKinsey. Another differentiator is our very deep understanding of processes that could be deployed at scale based on our industrial experience worldwide.
Q: How does agents at scale translate into productivity?
AS: Let’s look at the numbers. If you think about any practitioner, let’s assume that the raw productivity potential is one, just for simplicity. If you augment these practitioners with some of the gen AI tools, the productivity potential maybe goes up 20% up, so 1.2 or maybe a maximum of 2x. That’s the situation that we are seeing today. The moment you have practitioners to supervise agents, meaning a practitioner builds and supervises a certain workflow, which is done by say15-20 agents, we see a 20x jump in productivity or more. In our report we cite the case study of a large bank that needed to modernize its legacy core system, consisting of 400 different pieces of software—a massive undertaking budgeted at more than $600 million. Large teams of coders went about it via manual, repetitive tasks, with difficulty coordinating across silos and relying on often slow, error-prone documentation and coding. While first-generation Gen AI tools helped automate code conversion, the project remained slow and laborious. Then, human workers were elevated to supervisory roles, overseeing squads of AI agents, each contributing to a shared objective in a defined sequence. These squads retro document the legacy application, write new code, review the code of other agents, and integrate code into features that are later tested by other agents prior to delivery of the end-product. Freed from repetitive, manual tasks, human supervisors guide each stage of the process, enhancing the quality of deliverables and reducing the number of sprints required to implement new features. The bank saw productivity increases of up to 50%, with a 30-40% reduction in costs.
Q: What kind of organizational change is needed?
AS: Once you put together a team with some human beings, some virtual members, and they are all working together, you need to define tasks between them in a way that best fits their skills. It was Thomas Edison that said genius is one percent inspiration and ninety-nine percent perspiration. I suggest it’s the reverse now. It’s about human workers’ creativity, allowing them to spend 99% of their time on creative things and leaving more mundane and routine workto the machine.
Q: What is the CEO’s role in solving the Gen AI productivity paradox?
AS: Technology is just a lever. It’s about complete business reinvention and it ends and starts with leadership. It’s not a project for the IT department or the data department or the digital department. The CEO needs to drive it. Unless the CEO or chairman or owner of the business is committed to do so, and they personally drive it, it will be close to impossible to implement because when you start a lot of things are unclear, and there are no quick wins here. You really need to undergo the journey to see the impact and it requires some consistency on everything from changing human beings, to developing skills, to allocation of the budget. I think the role of the leadership in this process is trying to create conviction within the organization that this thing is relevant. Overcoming fears or some of the political tensions is just as important as everything else, if not more important. Therefore, its ends and starts with the leader who is behind this effort.
Q: What advice would you give to companies that want to have more success with Gen AI and effectively implementing agentic AI?
AS: It starts with leadership and conviction. If you don’t have it, basically, don’t bother. The leader should be willing to put his or her reputation on the line, to commit to this journey and not to expect to get any quick wins. Number two, before jumping into technology, you need to step back and rethink, how you reinvent the core of your business and your processes. This is the part that we see many organizations miss in their rush to implement solutions. It’s difficult, because you need to challenge the conventional wisdom, you need to get your data right and you need to work with your IT architecture to allow the data to flow. You also need to decide on what type of agents you want to use and on the foundational models you would like to put in to power this channel. Then, you need to ask yourself: Do I have enough great technology talent to implement this and am I able to convince the rest of the organization to understand the technology, embrace it and co-develop it? You also need to put in place an organizational structure that allows you to have agents and human beings working hand in hand in an agile way. You can’t pick and choose between these elements. Companies need to do all of these things to get the desired outcome.
