By Roshan Farhan, YC Founder, Founder – MyriadAI
For most companies, the first question about AI is still a technology question. Which model should we use? Where can we deploy a copilot? Which processes can we automate? Should we build internally or buy from a vendor?
Having spent over a decade as a YC and VC-backed tech founder and as a management consultant advising Fortune 100 enterprises across the US, India and APAC, I have come to think that the harder question sits further downstream:
What happens to the business when AI becomes capable of doing meaningful parts of the work?
An AI model can perform a task remarkably well and still create very little value for the company. Once that capability enters a real business process, it encounters legacy systems, fragmented data, approvals, compliance, human judgement and the economics of the workflow itself.
This is why the gap between an impressive AI pilot and a production system can be so large. A pilot can demonstrate that the technology works. The business then has to work out how the way work gets done should change around it.
In my experience, the pilot-to-production gap is often less about model capability than about everything surrounding the model: workflow design, systems integration, governance, human oversight and economics. The useful unit of analysis is the relationship between AI capability, the way work gets done, the organisation itself and the economics that follow. This can be understood through five layers:
1. AI capability
The starting point is still the technology. Frontier models can reason across increasingly complex problems, work across different modalities, write and execute code, use tools and increasingly perform multi-step tasks. The practical question is what those capabilities make possible inside an enterprise.
A customer-service organisation, for example, may begin by asking an AI model to draft responses. The larger opportunity comes when the system can understand a customer’s problem, retrieve information from internal systems, determine the appropriate action, execute what it is authorised to do and bring a human in when judgement is required. The first adds AI to an existing workflow. The second changes the workflow around what AI can do.
2. Work transformation: The workflow is where value starts to move
Many enterprises are discovering that the difficult part of AI deployment isn’t getting a model to produce a good answer. It is getting that capability into the right workflow, with the right data, systems, permissions and human oversight.
A production AI system has to work within the company’s existing reality. It needs to connect to the systems around it, handle exceptions, have clear ownership and perform reliably.
This is why the workflow is becoming the fundamental unit of enterprise AI transformation, rather than the model or individual use case. The key question becomes: how should the workflow be redesigned around what AI can now do, and does that redesign create a better business outcome?
This is also where AI adoption becomes AI commercialisation: the point at which technical capability starts translating into measurable business value.
3. Organisational redesign
When AI changes enough workflows, the organisation around those workflows has to evolve. This is where AI strategy moves beyond technology and becomes an operating-model question.
A sales organisation may need fewer manual handoffs. A software team may change how engineering work is divided between people and AI. A customer-service organisation may shift people towards complex cases as AI takes on more routine interactions.
Decision rights, management processes, incentives and governance may also need to evolve as AI moves from assisting people to executing parts of a workflow. The technology can work perfectly well and still fail to create value if the organisation is not designed to absorb it.
4. Economic Value: Eventually, the economics matter
Eventually, AI has to show up in the economics of the business. That might mean higher conversion, lower operating costs, faster product development, better customer retention, improved quality or greater output from the same resources. These are the outcomes that determine whether an AI initiative is creating meaningful value.
The number of pilots launched or employees given access to AI says little about the economic impact being created. The more useful question is: what changed in the economics of the business because of AI?
For an AI-enabled workflow, that means understanding the cost of the system, the level of human intervention required, the output and the business outcome attached to that output. As AI systems become more autonomous, the ability to automate a process is valuable only when the economics make sense.
5. From economic value to competitive advantage
If increasingly capable AI becomes broadly available, where does lasting competitive advantage come from?
Access to the technology itself is unlikely to be enough. The differentiation will come from what they build around those capabilities: their workflows, proprietary data, customer relationships, distribution, operating models and ability to learn and adapt. A company that redesigns a critical workflow around AI may gain an advantage in speed, cost, quality or scale. If it continuously improves that workflow and reinvests the resulting value, that advantage can compound.
This is the AI Value Transformation Framework. I used to think about enterprise AI strategy:
AI Capability → Work Transformation → Organisational Redesign → Economic Value → Competitive Advantage
The framework follows the path from what AI makes possible to how work changes, how organisations adapt and what those changes mean for the economics of the business. Over time, organisations that do this well can build strategic competitive advantages.
The AI strategy question is evolving
AI strategy now touches the entire business. That brings us back to the central idea of this article: AI strategy needs to move beyond technology adoption and into the design of the business itself.
The strategic opportunity for global enterprises lies in identifying where AI creates a fundamentally different way of working, then redesigning the workflows, organisation and economics around those capabilities.