Minh-Hai Nguyen
COO, Founding Partner
Rework
AI is making it faster to turn ideas into solutions. But faster execution does not tell us what ideas are worth building, whether the underlying work should change first, whether the organisation can absorb the change, or whether anything improves.
As execution accelerates, the constraint moves elsewhere: to prioritisation, decision-making, work redesign, adoption and value realisation. At Rework, we use the term AI Value Operations (AI Value Ops) to describe the operating layer. An AI software factory can accelerate delivery, but not repair broken Value Operations. It will expose them.
The Real Constraint Is Outside Execution
A decade ago, I worked with mobile garages and mobile factories. The promise was start-up speed with enterprise scale: a small core team close to the client, reusable baselines, clear guardrails and offshore capacity ready to build once direction was clear.
The model worked when the team had knowledge, trust and autonomy. Reference architectures, reusable components, delivery practices and agreed boundaries created a baseline within which people could make sensible decisions.
But development capacity was rarely the real bottleneck. I remember one project where the development team was ready to continue, but business wasn’t able to agree on the open design questions. Nothing was technically blocked.
Stand-ups and decision logs were not status rituals. At their best, they removed decision obstacles and forced open questions to closure and kept the pace. Without that discipline, the factory had capacity but the work did not flow. That lesson feels current again.
Then, the capacity was in offshore teams and global delivery factories. Now it is increasingly in agentic development and agents themselves. Faster coding, testing and production are real. But weak Value Operations means a faster factory simply reaches the bottleneck sooner.
AI Moves the Hard Work to the Edges
Traditional delivery relied on people to carry work through the system. That flow was slow and often hand-off heavy, but people also absorbed ambiguity along the way: missing details, architectural fit, edge cases and unresolved assumptions.
In increasingly agentic flow, agents can take structured intent, decompose it, generate implementation, create tests and prepare something close to finished. As production accelerates, the human role shifts.
People move increasingly to the edges: at the front, deciding what deserves to enter; at the back, judging what comes out; and around the flow, defining guardrails, ownership and value measures.
AI is not eliminating constraints. It is relocating them from engineering to management: from production capacity toward choosing the right change, closing decisions, redesigning work, driving adoption, and proving value.
AI Value Operations Is Not a Traditional PMO with an AI label
Value Operations is the operating system around execution. It connects intent, prioritisation, decision rights, work redesign, quality, adoption, maintenance, feedback and value measurement.
Some of this is familiar from programme management, transformation offices and portfolio governance. AI does not make those disciplines obsolete. But AI Value Operations should not become a traditional PMO with an AI label.
Value Operations should not create another control or approval layer. Its role is not to collect status updates or create another layer of approvals. It should help teams make better decisions while the work is happening. It should identify overlapping initiatives, challenge assumptions, and compare whether the right answer is to build, buy, adapt an existing solution, or solve the problem in another way.
AI should also support the Value Operations work itself. Benchmarking, initial analysis, requirements definition and process performance assessments should not always require another manual round of presentations and spreadsheets.
Five Disciplines That Keep Fast Execution Connected to Value
1. What is the value intent?
What outcome should improve, for whom, and who owns it? A weak idea does not become stronger because it can be implemented quickly.
2. How will decisions flow?
Who can approve, stop or scale the initiative? If decisions remain trapped between functions and steering groups, faster production will only bring us to the next unresolved question sooner.
3. How should the work change?
AI should not simply automate the current process. Which steps could disappear? Which handovers are unnecessary? Where is the actual bottleneck? Process data may show that the real problem sits between teams rather than in the task everyone first wanted to automate.
4. How will adoption happen?
Deployment is not adoption. Release notes are not change management. Value appears when behaviour changes, a process runs differently, a decision improves, a cost disappears or a risk is reduced.
5. Who will own this and how will value be proven?
Faster production creates more assets, dependencies, controls and things to maintain. Who will monitor, improve, secure, fund and eventually retire the solution? Did the bottleneck disappear or move elsewhere?
The Same Constraint Applies Beyond Software
Software factories make this shift particularly visible. But the same constraint is emerging anywhere AI increases the number of technically possible ways to change how work gets done.
Finance, procurement, customer service and enterprise-system renewal all offer multiple ways to improve work: standardise the ERP, automate a rule, use an existing capability, deploy an agent, redesign the process or combine several interventions. The hard part is selecting what addresses the real constraint and proving what changed.
That is why the unit of management should not be the AI use case alone. It should be the business change around it: the decision, process, behaviour or outcome that needs to improve.
Can We Improve as Fast as We Can Build?
We are gaining the ability to create AI solutions faster than we are improving our ability to choose, absorb and evaluate them.
Advantage comes from making better choices, stopping initiatives that do not bring value, redesigning work and turning delivery speed into measurable improvement. That is the role of Value Operations. The real competition is no longer who can build fastest, but who can improve fastest.