Driving discovery and reuse of AI
Brix from QuantumBlack Labs, AI by McKinsey
Organizations face significant challenges when scaling their AI efforts beyond experiments and proof-of-concept models. QuantumBlack Labs developed Brix for our growing library of reusable use cases. Brix has five key offerings to facilitate asset discovery and reuse: reusable code discovery, telemetry, siloed code consolidation, easy project integration, and clear versioning.
QuantumBlack Labs is the R&D and software development hub within QuantumBlack, AI by McKinsey. QuantumBlack Labs has more than 250 technologists dedicated to driving AI innovation and supporting and accelerating the work of its more than 1400 data scientists across over 100 locations.
We use our colleagues’ collective experience to develop suites of tools and assets that ensure AI/ML models reach production and achieve sustained impact. In the coming months, we will publish a series about the technology challenges behind digital and AI transformations, and the solutions required.
Projects using artificial intelligence and machine learning (AI/ML) are a key driver of business value, and companies have shifted from exploring what the technology can do to exploiting it at scale to gain market share. But the complexity involved in building, deploying, and managing AI/ML at scale should not be underestimated.
One reason that organizations find it hard to scale AI/ML is a tendency among teams to reinvent the wheel rather than reuse existing solutions.
Assetization promotes scale
As our recent book, Rewired: The McKinsey Guide to Outcompeting in the Age of Digital and AI (Wiley, June 2023) describes, scaling effectively relies on being able to reuse code, insights, and resources as much as possible, to streamline operations and enhance productivity. The benefits of packaging reusable solutions as a set of modules, or assets (hence assetization) are as follows:
- Greater speed: it is faster to deploy solutions that are already 60 percent ready to go, focusing on solving business problems rather than writing redundant code.
- Lower cost: most use cases need to be deployed across multiple business units to unlock their true potential. Reinventing the solution every time fundamentally undermines the realized value.
- Reduced risk: reusing solutions that have already been tried and tested reduces the chances of introducing bugs or errors into code and limits the potential risk of failure.
- Greater consistency: reuse of modular assets across an organization aligns the teams.
- Easier evolution: By adding constant improvement loops, every version of the solution creates a strong foundation for future scalability and adaptability.
So why is a culture of assetization difficult to establish? Some of the reasons include:
- Solutions are not designed from the ground up for reuse by other teams.
- Reusable solutions are hard to find due to a lack of a single source of truth for all AI assets.
- Individuals are reluctant to reuse code, preferring to create a new component because of “not invented here” syndrome.
Brix by QuantumBlack Labs
At QuantumBlack, AI by McKinsey, we faced the challenge of data science component reuse a few years ago, as we grew in numbers. There was significant time pressure on our data practitioners, who found they were tackling problems solved elsewhere in the company.
Discovering assets in wikis and disparate repositories wasn’t satisfactory, so QuantumBlack Labs set out to build a product to facilitate assetization. The result was Brix, a code sharing and discoverability platform that plugs into existing code repositories across the organization.
To make it easier to find the entries with Brix, they are organized in a taxonomy that covers use cases, industries and industry sectors, and the type of asset. Key features of Brix include:
Reusable code discovery — Search for code snippets in Python, R, SQL, Java, C++ and more, utility functions, internal plugins, pipelines (e.g. Kedro) by topic, author, or keyword, within the QuantumBlack taxonomy.
Telemetry — Insights about the usage of contributions through analytics dashboards. Asset owners track the impact of their shared resources to understand how code contributions are used. Telemetry data also enables organizations to measure reuse across teams and determine assets that offer the best return on investment.
Siloed code consolidation — Unified access to siloed code content by integrating with code repositories (e.g., GitHub, Azure DevOps).
Easy project integration — One-click downloads of code, content, and documentation to accelerate development.
Clear versioning — Dynamically updated change history and versioned distribution to ensure teams are using the latest and greatest.
The business impact through faster time-to-value for AI is two-fold:
- Increased developer productivity: The impact internally has been huge, with up to 40 percent acceleration towards first insights on client projects.
- Fostering collaboration: Brix builds an internal community and a dynamic content ecosystem for contributing and reusing analytics assets. In QuantumBlack, over 1,600 code assets have been shared to date.
Many organizations are unlocking similar benefits by focusing on assetization and reuse. Freeport-McMoRan (a mining business) developed a machine learning model to predict how much copper could be recovered under any set of conditions. To make the model reusable, Freeport refactored and repackaged it so that it could be adapted effectively across different sites. The core code was organized so that 60 percent could be repurposed easily; 40 percent was then customized for individual plants. The company also invested in a central code base that site-specific modules could use, making it easier and more cost-efficient to maintain and improve code.
Does my team need Brix?
Organizations often ask us, “When is the right time to think about assetization and reuse?”. The short answer is before you attempt to scale. But the following are all clear signals:
- A critical mass of engineers building AI/ML use cases that would benefit from reusing each other’s code (50+).
- No central repository that can enable teams to discover reusable assets.
- The organization has an early hypothesis on assets that could be shared, and teams are incentivized to collaborate.
When establishing those early hypotheses, the organization should keep this equation in mind when identifying candidate code for reuse:
Value Add = Cost of Creation (Cost of developing an asset from scratch) — Cost of Reuse (Cost of discovery, understanding & integration)
Summary
At McKinsey, we have seen first-hand that moving AI solutions from idea to implementation can be challenging. We recognized that scaling an AI/ML project relies on being able to reuse as much of the solution as possible to decrease the time-to-value and reduce development costs.
Designing for reuse from the outset is a strategic move towards sustainable and successful AI/ML initiatives:
- Individuals reduce the number of repetitive coding tasks
- Teams shift their focus towards more strategic activities such as collaboration, innovation, and refining business strategies
- Projects are leaner and more resource-efficient, allowing for better allocation of human and computational resources
- Organizations benefit from the culture of creativity and continuous improvement. They can modify, expand, or even repurpose projects to suit different environments, be it across various plants, geographic markets, customer segments, or organizational groups
QuantumBlack Horizon is a family of enterprise AI products, including Brix, Kedro, and Alloy, that provides the foundations for organization-level AI adoption by addressing pain points like scaling. It’s a first-of-its-kind product suite that helps McKinsey clients discover, assemble, tailor, and orchestrate AI projects.
To learn more about what QuantumBlack Horizon can do for you, please email Yetunde Dada.
Thanks to all who contributed to this article: Mohammed ElNabawy, Tanaka Kungwengwe, James Mulligan, Jo Stichbury, Joanna Sych, Sarah Mulligan & Matt Fitzpatrick.
