There’s a lot of hype around artificial intelligence (AI) — and for good reason. With the right algorithms, businesses can quickly pinpoint and act on patterns, clusters, trends and anomalies in customer behavior, supply chains, internal operations and market dynamics. But as the core methodologies of data science become more widely adopted, AI algorithms become less of a competitive differentiator.
In fact, the availability of machine learning, natural language processing and other algorithmic methods on-demand via the cloud and off-the-shelf/plug-and-play models — many of which are open source or free — is likely to accelerate the commodification of AI in the near future, essentially making it table stakes.
So, the question facing business strategists will soon shift from, “How can we quickly adopt AI to beat the competition?” to “Given the fact that our competitors are also using AI, how do we use AI most effectively as part of a broader strategy to build a business with a sustainable moat around it?”
AI And The Enterprise-Specific Digital Value Chain
To answer this increasingly important question, we first have to recognize that while AI is a great catalyst for value, it does not itself deliver value. Instead, it is a powerful potential enhancement to an organization’s unique end-to-end digital value chain.
The potential value of AI is thus only realized to the extent that it fulfills its role in that particular value chain. And that role can be defined in three dimensions:
• Technical architecture. From a technical perspective, AI and other algorithmic resources are “systems of intelligence” that sit between systems of record and systems of engagement. Systems of record include internal databases and transaction processing, as well as a growing number of external systems, such as third-party databases, social media feeds and telemetry from customers’ or partners’ internet of things (IoT) devices. Getting the right data is critical, as data can be biased and may create violation of privacy laws if not captured properly.
Capturing and analyzing data from these diverse back-end sources, AI systems of intelligence deliver actionable outputs to systems of engagement. These are typically customer-facing mobile apps, IoT devices and cloud-based services. However, we might also call these “systems of interaction,” since they also sometimes face supply chains or internal constituencies, rather than customers.
• Business architecture. AI initiatives also have a specific footprint within businesses. We tend to think about AI narrowly in terms of data science, but there are multiple stakeholders in these initiatives, including primary stakeholders (whether that’s sales, marketing, product management, etc.), executives with profit and loss responsibility, corporate legal and compliance leaders and managers of the IT systems AI touches.
Every business structures collaboration by these stakeholders differently. And these collaborative structures tend to evolve as AI moves from early pilots to initial deployments in production to broader implementation across the enterprise.
• Operational management. All AI is not created equal. Different classes of algorithms are appropriate for different types of cluster, trend and anomaly detection. The speed with which results are returned can be affected by many factors, including compute, data input/output and data types. User acceptance and uptake can be dramatically affected by presentation UIs (user interfaces).
These operational parameters are non-trivial and must be attentively managed to ensure optimal business outcomes. This attentiveness is especially important because, as I noted in an earlier article, systems of intelligence running self-learning algorithms are non-deterministic and, therefore, exhibit far less predictable behaviors than traditional deterministic applications.
Sustainable Advantages In AI Execution
The three dimensions of AI execution above give us one clue as to how we can achieve sustainable advantages in a market where algorithms have become commodities. Consistently superior execution leads to consistently better results, just as in every other business discipline.
Every business sells. But some have developed an empirically superior sales culture and process that empowers them to consistently win accounts, especially in target markets. Every business tries to develop products that customers will love. But some have embraced research and design processes that enable them to more consistently produce winners.
The same is true of AI. If you build a technical architecture engineered to ensure data quality, safeguard privacy and eliminate bias in data, your systems of intelligence will drive more value than competitors who don’t. Similarly, if you build a business architecture that keeps your business SMEs (subject matter experts) closely involved with your data scientists, your systems of intelligence will drive more value than competitors who don’t.
And if you build operational management processes that keep the delivery of AI outcomes fast, reliable and easy to consume — even at scale — your systems of intelligence will drive more value than competitors who don’t.
Execution, process, architecture and culture all matter. Get them right, and your company can out-perform its competitors even in a world of commodified AI. But there are still other ways you can leverage AI to create a classic moat around your business — I’ll address those in part two of this series on AI differentiation.