A quantum step in the agentic direction for McKinsey

Recently, PAC had the opportunity to attend a McKinsey event in London focused on their QuantumBlack division. Next year will be the 100th anniversary of McKinsey, which is internationally known for its business-led management consulting capability serving businesses, governments, and institutions focused on addressing the strategic and operational challenges organisations face. This context is provided because that is broadly how the organisation is typically seen, and whilst technology plays a vital part in the outcomes delivered, the company is not recognised as a technology company first and foremost. However, this recent event PAC attended showed a side to McKinsey that demonstrates the company understands how the role of AI will continue to affect, and evolve rapidly, the core types of service delivery they provide for organisations. However this would have been more of a challenge had the company not acquired QuantumBlack in 2015, 6 years after it was formed, and subsequently allowed it to evolve as the sophistication and use cases of AI have progressed in the subsequent years.

At the time of its 2015 acquisition, PAC considered QuantumBlack an acquisition to compete in the data science and advanced analytics space in which companies like Palantir had gained significant dominance. What was interesting about the recent session in London is how far QuantumBlack has evolved into delivering engagements focused on ever-evolving advanced AI use cases in addition to its data science expertise. A conversation with the head of QuantumBlack Labs was particularly interesting to PAC. They discussed how their approach is not just to identify and create innovations but also to determine when an AI-related service they offer has been superseded so it can be retired. To PAC, this was a refreshing perspective in a world where the AI conversation is only ever about the next innovation because, time and time again, CxOs struggle with legacy technology footprints and the cost and complexity challenge of retiring them.

At the event, QuantumBlack provided a range of AI transformation use cases that they had delivered both for their clients and as internal capabilities to further empower how McKinsey operates. What was particularly interesting was how QuantumBlack was passionate about their role of enabling a client to skill up and retain expertise relating to the engagements they deliver, so there is no further dependence on them after a business outcome has been achieved. PAC finds this, especially regarding AI, to be a refreshing engagement approach because it builds greater trust between them and their clients with regard to the delivery of outcomes.

Innovating migrations with agentic AI

Of course, an AI-related event in the current climate would not be relevant if agentic AI were not a significant focus. In this regard, QuantumBlack provided what PAC considers an extremely pertinent use case for agentic AI to support clients in migrating from mainframes; this would have been welcome in prior enterprise-based roles this analyst has performed. For decades, the migration from one platform to another has been hindered by complexity, typically in the form of recoding, because many organisations deploy code that becomes obsolete and then look to migrate, but are faced with knowledge and documentation issues.

From personal experience, this analyst has worked for many organisations where the cost, complexity, and time to perform a migration often outweighed the benefit in the near term. This perpetuates legacy technologies in organisations long past their technological value and relevance. For many organisations, a great example is the use of mainframes. Over the last several decades, the knowledge required to migrate from mainframes has significantly diminished because of the onset of client/server and cloud enterprise architectures. This has left many organisations, that used mainframes for core business functions, with significant technical debt because of a high cost to migrate due to skill scarcity and, all too often, technical documentation that either did not exist, was not complete enough, or was not understandable.

For too many years, this type of “skeleton in the closet” has been a challenge that weighed down organisations because there were no viable means to resolve it cost-effectively. However, the McKinsey event clearly showed that the QuantumBlack team understood the opportunity that agentic AI presents to solve this dilemma. Whilst individual organisations may collectively lose knowledge over time relating to technical assets, the sum of information available through the internet to train AI models has now become a viable means to resolve code migration challenges. QuantumBlack used an agentic AI approach to create a factory of digital workers spanning developers, testers, and documenters. Whilst PAC considers there to be a lot of hyperbole regarding IT departments turning into HR departments for AI agents, the real-world client use case example provided by QuantumBlack took the approach of using agents, spread across a range of agent teams, to perform roles typically performed by humans to deconstruct mainframe code, reconstruct in the desired new form, test the new form, and fully document it. They then scaled up the required number of agents within each agent team based on the client’s timeframe for delivery.

For PAC, this example is a groundbreaking approach to resolving a task that most organisations struggle with addressing. For example, one of the promises of cloud compute was the ability for easy code portability across cloud platforms provided by competing firms. However, for many organisations, the reality of cloud is much like prior IT compute mechanisms before it in that there is a degree of vendor lock-in that makes moving between IT services more costly than it should be. PAC, for example, considers that the approach taken by QuantumBlack for the mainframe is equally as transformative for organisations migrating to SAP S/4 Hanna, migrating components/capabilities from SAP to ServiceNow (using their low-code NOW platform), and moving between hyperscaler compute providers like AWS and Microsoft Azure. What previously would have required labour arbitrage, often achieved through an outsourcing-type model, is now in a position to be disrupted at scale and speed through highly orchestrated agentic AI factories. Lowering the number of people required to achieve the same task, whilst ensuring that the search for skilled resources to perform migrations for older technologies like mainframes is not the burden it once was.

Leveraging AI agents to challenge human-led operating models

Over the coming years, though, PAC would like to see more companies like QuantumBlack supportively challenge organisations in how they use agentic AI. This is because agentic AI can provide operational approaches that are not limited by workflows and operating models defined for human behaviour. Under relevant circumstances, it can also offer a means to challenge how human-led business operating models are currently configured. It is clear to PAC that the QuantumBlack team are already leading engagements with this approach in mind whilst balancing the need to take clients on a journey where challenging how processes behave can require complex discussions to drive the full value of agentic AI. Seeing the broader industry evolve regarding the delivery of agentic AI engagements, and in particular from QuantumBlack, is something that PAC looks forward to seeing evolve as more use cases and common patterns are adopted as the technology matures. Whilst orchestration and governance of agents across all forms of software and services is critical, as it has been for prior technologies, PAC believes that the true disruptive and innovative opportunity to be derived from agentic AI is in organisations challenging themselves to rethink how they can better operate by using it rather than just overlaying atop existing business processes designed around the needs of human behaviour and interactions.