Key Takeaways
- The cost and complexity of software development are falling, shifting the main constraint from building software to clearly specifying what it should do.
- AI does not create clarity on its own; it reveals ambiguities, gaps, and hidden tribal knowledge in existing processes.
- Precise, reusable specifications of work are becoming the core enterprise asset, while the software that expresses them becomes commoditized.
- Operational experts who understand how work actually happens must move closer to strategic decision‑making to author these specifications.
- Effective governance now hinges on turning institutional knowledge into reusable assets, not merely counting AI use cases or productivity gains.
- Companies that master upstream specification—designing work around true value creation—will outperform those that merely generate more software faster.
The Shifting Bottleneck
For most of my career, enterprise technology transformation was hampered by the high cost, scarcity, and complexity of software development. Large organizations needed product teams, engineers, architects, testers, program managers, funding cycles, release governance, and change management just to get meaningful systems into production. Even after overcoming those hurdles, the delivered solution often only partially matched the business’s original intent. AI is beginning to alter this dynamic: as the marginal cost of writing code drops toward zero, the limiting factor is no longer whether a firm can afford to build software, but whether it can precisely define what the software should accomplish and how work should change.
AI and Operating Ambiguity
Many enterprise processes were never cleanly designed; they evolved organically through accumulated policies, controls, regulatory steps, system limitations, and workarounds, often cemented by long‑tenured employees who became the de‑facto system of record. When AI is introduced into such an environment, it does not magically create clarity; instead, it exposes the absence of it. Poorly understood workflows accelerate confusion, tribal knowledge surfaces as gaps, undocumented exception paths cause AI to halt early or act overconfidently, unclear control points generate governance problems, and fragmented operating models become more fragmented—only faster. This explains why many AI pilots look impressive in demos yet falter in production: the underlying specification of decisions, rules, data dependencies, accountability, and outcomes is missing or vague.
Specification as Enterprise Asset
In the next phase of enterprise AI, most large companies will have access to similarly powerful models, cloud platforms, copilots, agent frameworks, and systems of record. The technology stack will still matter, but it will no longer be the primary source of competitive advantage. As software becomes easier to generate, value shifts from the software artifact itself to the specification that defines what the software is meant to achieve. The critical question becomes: who understands the work well enough to write that specification? Who knows where value is created, where delay and risk creep in, where customer experience breaks, where human judgment improves outcomes, and where human involvement merely reflects legacy process design? Those who can translate deep operating knowledge into clear, reusable specifications will be able to redesign work—not just automate tasks—building systems that mirror how value is actually created rather than forcing the business to conform to generic assumptions embedded in vendor products. Over time, the specification becomes the enduring asset; software is merely its expression.
The Central Role of Operational Talent
This shift raises an uncomfortable question for leadership: where does the talent that truly understands the work sit today? In many firms, the people with the deepest operational knowledge reside in operations, shared services, outsourced units, or process‑excellence teams. They know how work really gets done but are often treated as participants in transformation rather than as architects of the future operating model. If the next wave of enterprise value derives from specifying work with precision, these experts must be moved much closer to the strategic core. They need to sit alongside technology, data, risk, product, and business leaders to decide what should be automated, augmented, redesigned, owned, or retired. Elevating operational insight to a design role is essential for turning tribal knowledge into a scalable, reusable asset.
Rethinking Governance for AI
Boards and executive teams must therefore adjust the questions they ask. It is no longer sufficient to inquire merely about the number of AI use cases, how many employees are using copilots, or reported productivity lifts. The more pertinent inquiries are: Which workflows have we specified well enough to redesign? Which parts of our operating model should no longer be constrained by standardized software assumptions? Are we converting institutional knowledge into reusable enterprise assets? Do we have the right people positioned close enough to the core to define what good looks like? Traditional concerns—code quality, architecture, governance—remain relevant, but they are now downstream activities. The decisive advantage lies upstream, in the ability to articulate, continuously improve, and govern the specification of work itself.
Conclusion: Upstream Advantage
The bottleneck in enterprise transformation has moved from the costly, scarce act of building software to the upstream challenge of specifying work with clarity and precision. AI lowers the cost of software creation, but it simultaneously amplifies any fuzziness in processes, making precise specification not just helpful but essential. Organizations that harness their operational expertise to craft clear, reusable specifications will gain a durable edge: they can redesign work around true value creation, rather than forcing the business to fit legacy software molds. In this new paradigm, the specification is the asset, software is its expression, and the winners will be those who master the upstream act of defining what the technology should serve.