Imagine your board pressing for clear answers on AI, while competitors announce new initiatives week after week and your inbox overflows with pitch decks promising nothing less than total transformation. Amid all this noise, one critical bottleneck is rarely named: it’s not a technology problem – it’s a translation problem.

Executives today navigate between two worlds whose languages rarely meet cleanly. On one side, technical feasibility – probabilistic models, latency, tokens. On the other, the language of business – risk, accountability, EBITDA, market dynamics. In the gap between these worlds, budgets disappear, teams are restructured without changing value creation, and reputations suffer.

Most organizations speak their core business fluently and have excellent technologists. But AI fundamentally changes the rules – not because every manager suddenly needs to understand neural networks, but because the nature of decision-making shifts. Dave Snowden, creator of the Cynefin Framework, sharply distinguishes between complicated and complex systems: traditional IT was complicated, where cause and effect could be analyzed by experts; generative AI makes organizations complex – outcomes emerge, are not always predictable, and change dynamically through interaction (Snowden & Boone, 2007). Trying to steer this with yesterday’s management tools risks losing control.

Translation as the Missing Core Competency

The best AI decisions are rarely made by those who know the code best. They are made by those who can translate between worlds without diluting the discipline of either side. It’s not about turning strategy into marketing slogans or technical limits into sales promises. It’s about the hard work of turning raw technological capability into decisions that survive organizational reality.

At Singularity Inc., we know this from years of hands-on advisory work across industries from financial services to manufacturing. We’ve seen how poor translation leads to expensive failures. This translation can be structured into five precise steps, which we visualize as a strategic funnel to make the process immediately clear.

The 5 Dimensions of AI Translation

Rather than a simple list, here’s the process as a funnel model – each phase building on the previous one and addressing a strategic question inspired by Roger Martin’s approach in Playing to Win.

This funnel helps keep the big picture in view and shows at a glance how abstract ideas flow into concrete execution.

Step 1: Separate Signal from Noise

The starting point is to ask what is technically feasible right now, what is likely in six months, and what is pure vendor prose. It’s not about being dazzled by demos; it’s about understanding the “frontier capabilities” that matter for your business model. Take IBM’s Watson Health as an example: despite massive investment, it failed in part because technical promises – such as accurate cancer diagnosis – didn’t align with real-world data quality and integration challenges, leading to disappointed expectations.

For a law firm, a model’s hallucination rate can be an existential risk; for an advertising agency, the same inaccuracy might be a creative feature. Without this technical understanding, strategic management is impossible, as a recent HBR article puts it: “A company’s success will not rest on AI per se; it rests on what companies do with it” (Chui et al., 2025).

Step 2: Turn Capabilities into a Decision Portfolio

Henry Mintzberg warned early against confusing strategy with planning – strategy is a pattern in a stream of decisions (Mintzberg, 1987). Applied to AI, this means you don’t need a long list of 50 use cases; you need a portfolio with clear owners. Where does AI change your cost structure? Where does it increase speed? Where does it create new risks? Roger Martin captures it perfectly: “Strategy is choice. It is not a long planning document; it is a set of interrelated and powerful choices” (Lafley & Martin, 2013).

A real-world example is General Electric’s Predix platform: billions invested in AI for industrial IoT failed because the portfolio wasn’t prioritized – technical feasibility was pursued without strategic relevance, leading to massive losses. The decisive translation often lies in saying “no” to technically possible but strategically irrelevant projects.

Step 3: Adapt the Operating Model

Governance, data access, and model selection are the visible parts; the dangerous invisible part is how decisions are made when AI generates options faster than humans can validate them. If your current approval process takes weeks for a marketing text but AI delivers twenty variants in seconds, it becomes a bottleneck that kills productivity.

The operating model must shift from “control by permission” to “control by principles.” Consider McDonald’s AI-driven drive-thru: the technology didn’t fail because of the AI itself but because the operating model wasn’t adapted, leading to order errors and frustration when human processes couldn’t keep up. As Marco Iansiti and Karim Lakhani argue in Competing in the Age of AI: “AI changes the boundaries of the firm, requiring a rethinking of operating models to integrate digital and analog processes seamlessly” (Iansiti & Lakhani, 2020).

Step 4: Lead People Through the Identity Shift

When AI designs, suggests, and automates, employees don’t just learn new tools – they renegotiate their professional identity. A senior analyst whose value lay in structuring data in Excel faster than others experiences a devaluation. If you don’t shape this psychological process, AI adoption becomes theater: tools are used secretly or avoided, resistance is disguised as ethical concerns. Microsoft’s Copilot integration in Office showed how teams became more productive – but only when leaders addressed the change. Otherwise, gains stay stuck in pilots. A recent HBR piece on AI-first leadership emphasizes: “Leaders must reimagine human-AI collaboration and invest in AI-specific upskilling to avoid resistance” (Choudhury, 2025).

Step 5: Anchor Translation Competency Internally

This must not be a one-off program or eternal dependence on external experts. The goal is stable internal capacity that continuously translates new AI possibilities into safe, value-creating change. As Mintzberg says, strategy is also a learning process through “emergent strategy” (Mintzberg, 1987). Your organization must treat AI as a permanent component. Amazon’s success with AWS illustrates how internal translation – from tech to business – drives scalable growth, in contrast to firms that rely on external consultants and fail.

Where Most Consulting Conversations Fail

Many consultants stay in one language: tech discourse remains with models and proofs of concept; business discourse stays abstract with visions and generic roadmaps. The hard part is the interface – where you decide what not to do, what to measure, and which structures to dismantle because they’re too rigid for the AI era. That’s why so many programs drift: not from lack of intelligence, but because the translation layer is missing and old habits fill the gap – committees, pilots without scaling paths, tool rollouts without accountability.

Our Approach at Singularity: Building AI-Intelligence

We don’t sell AI – we build AI-Intelligence into organizations. What we defined by the term "AI-Intelligence" is the capacity to transform technological capabilities into decisions, working systems, and new ways of working – without losing control over risk and coherence.

In other words, to us AI-Intelligence represents the organizational capability to effectively understand, manage, and integrate AI, enabling sound decision-making and reliable operational execution. On this basis, our work happens close to the decision surface, where incentives collide and execution breaks. Typically, we focus on three outcomes:

A translation map of your reality: Where is AI already present (e.g., shadow IT)? Where does it create unmanaged risk or untapped opportunity?

A decision portfolio with backbone: A small number of high-impact levers with clear owners, timelines, and necessary operating model changes.

An enablement path without hero myths: How leadership and governance must evolve so AI becomes a controlled production capability, not the hobby of a few enthusiasts.

This work is uncomfortable because it forces clarity – clarity about what is true, what is not yet true, and what you only pretend is true because the organization prefers comfort over precision.

If you’re serious about AI, you need less PowerPoint and more translation, fewer vendor promises and more decision discipline. We strongly believe that it is actually much less talk about the future than design of the present that is required. Because the present has already changed substantially.

Dr Alexander Fruehmann is a founding partner at Singularity.Inc. For further discussions or if you are interested in translating your own operational challenges into a clear strategy, contact us anytime.

References

• Snowden, D. J., & Boone, M. E. (2007). "A Leader’s Framework for Decision Making." Harvard Business Review.

• Mintzberg, H. (1987). "The Strategy Concept I: Five Ps for Strategy." California Management Review.

• Lafley, A. G., & Martin, R. L. (2013). Playing to Win: How Strategy Really Works. Harvard Business Review Press.

• Chui, M., et al. (2025). "Make Sure Your AI Strategy Actually Creates Value." Harvard Business Review.

• Iansiti, M., & Lakhani, K. R. (2020). Competing in the Age of AI: Strategy and Leadership When Algorithms and Networks Run the World. Harvard Business Review Press.

• Choudhury, P. (2025). "AI-First Leadership: Embracing the Future of Work." Harvard Business Publishing.