From Experimentation to Intelligent Enterprise Transformation
Author: Arun Sharma, Chief AI Business Officer, SpeedUP, Vietnam
Date: 14-Aug-26
As AI moves from experimentation into the core of how organizations work, compete, and create value, leaders must rethink infrastructure, workflow, governance, productivity, and the role of human intelligence.
| The next competitive advantage will not come from using AI. It will come from redesigning the businesses around it and for Vietnamese leadership, that window is open right now! where their Dreams turn into Reality. |
For the past few years, most organizations have been asking a narrow set of questions about AI: where could we use it, which model should we choose, how soon can we get a pilot running? Those were reasonable questions when generative AI first arrived in the enterprise. But technology has moved faster than the questions, and a lot of businesses are still asking them long after they stopped being useful.
The more useful question today is different: how should an organization redesign itself once intelligence is embedded in its infrastructure, its software, its workflows, and the everyday work of its people? That is a bigger question than “which AI tool should we buy,” and it is really what this article is about.
“The World Economic Forum has spent the past two years framing AI less as a technology story and more as a leadership one — something that reshapes the workplace, competitiveness, and even national economic capability, not just IT budgets.¹ Its recent work on AI infrastructure adds a related point: infrastructure itself has become a strategic question, shaped by compute, connectivity, data, resilience, and sovereignty, not simply a procurement decision handed down to the IT department.²”
This shift arrives at a moment when the regulatory floor is finally being built. Comprehensive national AI laws are now taking effect across multiple markets — from the European Union’s AI Act to a fast-growing list of national frameworks across Asia, the Middle East and the Americas — giving businesses their first real legal scaffolding for how AI should be developed, deployed and used, built around human oversight, safety, innovation and accountability.¹⁵ So the question for most businesses was never really whether they would enter the AI era. It is how deliberately they choose to do it.
The transformation before this one was mostly about digitizing what already existed. Paper became software, physical channels became digital ones, and disconnected systems slowly turned into connected platforms. It was a significant shift, but the underlying work stayed roughly the same — people using better tools to do what they had always done.
AI changes that equation. Digital systems help people execute a process. AI increasingly understands the process, recommends what to do next, creates the output, coordinates across steps, and in some cases acts on its own. That is a different kind of change, which is why this next transformation does not belong in the “technology upgrade” category. An AI-native organization is not one with a chatbot bolted onto its website. It is one where AI shows up in how the business serves customers, manages knowledge, develops products, communicates internally, handles legal obligations, supports employees and makes decisions.
“McKinsey’s 2025 global AI research captures where most companies actually are on that journey. Nearly nine out of ten organizations surveyed use AI regularly in at least one part of the business — but almost two-thirds have not started scaling it beyond that, and only 39% report AI showing up in enterprise-level profit.⁵ The gap between those numbers is really the whole story: AI experimentation has become common. AI transformation is still rare.”
This is why the next transformation should not be considered simply another technology upgrade. It is an evolution:
Digital → Data-Driven → AI-Enabled → AI-Native
| AI experimentation has become common. AI transformation is still rare — and that gap is where the next five years of competitive advantage will be won. |
Over the next five years, that gap is where the real competitive advantage gets created — by the organizations that manage to turn one-off experiments into something repeatable, governed, and measurable.
Software engineering is one of the first business functions being reshaped from the inside out. AI can already help with requirements analysis, writing and reviewing code, testing, documentation, and pulling up the right piece of institutional knowledge at the right moment. It is tempting to read that as “developers can type faster,” but the more interesting change sits upstream of that.
The traditional software lifecycle ran in a fairly fixed sequence — gather requirements, design, build, test, deploy, then maintained largely by hand and in sequence. That sequence is loosening. Requirements get explored through AI-assisted conversation instead of static documents. Prototypes get built and discarded in hours instead of weeks. Testing runs continuously instead of only at the end. Documentation updates itself instead of quietly going stale. And once something is in production, AI-assisted engineering helps keep it maintained, rather than waiting for the next scheduled overhaul.
The traditional lifecycle:
Requirement → Design → Development → Testing → Deployment → Maintenance
is evolving toward:
Business intent → AI-assisted design → AI engineering → Continuous validation → Intelligent deployment → Continuous optimization
That changes the economics of building a product. A team can move from ideas to a working prototype faster, test more alternatives before committing, and keep documentation and internal knowledge current instead of chronically behind.
“PwC’s 2026 AI Jobs Barometer puts a number on the broader pattern: productivity growth is running about 40% higher at companies most exposed to AI than at the least-exposed ones, and the skills required in highly AI-exposed jobs are changing substantially faster too.⁶”
None of that makes speed the only objective, though. If anything, it makes governance, security, and quality more important, not less — an AI-assisted team can amplify a mistake exactly as efficiently as it amplifies good work. The future of AI-enabled software development must hold four things together at once: agility, security, governance, and human judgment. Drop any one of them, and the other three stop mattering as much as they should.
One of the clearest lessons showing up globally right now is the gap between running a pilot and capturing enterprise value.
“McKinsey’s research finds that organizations are starting to redesign workflows, stand up AI governance, and create new AI-specific roles, but most still have not captured meaningful financial impact at the enterprise level.⁷”
“The World Business Forum’s recent programming on AI business transformation lands on a similar point from a different angle: the useful question has quietly shifted from ‘should we adopt AI’ to ‘how do we make AI produce a measurable business outcome,’ and plenty of companies have not caught up to that shift yet.⁸”
There is a practical maturity curve underneath this, and most organizations sit somewhere along it: curiosity and scattered experimentation first, then a cluster of disconnected pilots, then a smaller number of pilots that get properly resourced and governed, and only after that, enterprise-wide transformation. Very few companies skip straight to the last stage, and none of them get there by accident.
Experiment → Pilot → Integrate → Scale → AI-Native → Transform
This is roughly where the Chief AI Business Officer role earns its place on the leadership team. The job is not to ask, “Where can we put an AI model?” It is to ask a harder set of questions: where would intelligence create measurable economic value? Which processes are worth redesigning, and which are not? What should stay firmly in human hands? Which decisions can AI meaningfully augment, and where must a person remain the final check? What data and infrastructure does this genuinely require? How is the risk governed? And, critically, how will we know if it worked?
That list is the real difference between AI implementation and AI business transformation — one produces a demo, the other produces a result.
For years, the default infrastructure question was simple: cloud or on-premise? AI has made that question more layered. It is no longer just about where to host a server. It is about where your data actually lives, where AI processing happens, and who ultimately controls the intelligence your business now runs on.
“The World Economic Forum’s 2026 work on AI sovereignty makes a point worth sitting with: sovereignty does not have to mean building everything domestically from scratch.⁹” The more realistic path is balancing domestic capability with trusted international partnerships, while keeping enough resilience and control that a business — or a country — is not caught exposed if something changes upstream. In practice, that produces a more layered enterprise architecture than the old cloud versus on-premise debate: public cloud, private cloud, on-premise AI, and sovereign infrastructure, often all four running side by side, matched to what each workload actually needs.
Public cloud still makes sense for elastic, lower-sensitivity workloads — traffic spikes, non-sensitive experimentation, anything where scale matters more than control. Private or on-premise infrastructure earns its cost where an organization needs tighter control over sensitive data, latency, security or cost predictability, or is answering to a regulator who cares where the data physically sits. That describes several of the most regulated, highest-stakes sectors, such as financial services, legal services, government, healthcare, manufacturing, and critical infrastructure.
“McKinsey’s recent work on infrastructure evolution traces a similar arc at the technical level — from traditional on-premise, through cloud, and now toward hybrid, GPU-accelerated, AI-oriented infrastructure that does not force a single either/or choice.¹⁰”
| Use the cloud where it creates an advantage. Use sovereign infrastructure where control creates the advantage. Build the architecture around the business — not around a single technology ideology. |
This is the thinking behind why we treat sovereign AI infrastructure at SpeedUp as a business capability, not a data-center product bought once and forgotten. Our Sovereign Immersion-Cooled AI Infrastructure exists to give organizations an environment where sensitive AI workloads can run with real control over compute, data, and operational boundaries — not as an ideological stance against the cloud, but as one deliberate piece of a larger architecture. The underlying principle is simple to state, even if it takes real engineering to deliver: use the cloud where it creates an advantage, use sovereign infrastructure where control creates the advantage, and build the architecture around the business rather than around a single technology’s marketing.
AI creates a genuine paradox for leadership teams: the more autonomous a system becomes, the more governance it requires — right now, governance starts to feel like it is slowing things down. AI agents can now retrieve information, act inside other systems, generate real outputs, and carry out multi-step workflows without a person in every loop.
“The World Economic Forum’s framing is useful here: agents should not be thought of purely as productivity tools. They change processes and organizational behavior in ways a simple software upgrade never did.¹¹”
“EY’s 2025 Responsible AI Pulse research found a real gap between how fast organizations are adopting AI and how mature their governance actually is — adoption is outrunning oversight at most of the companies it studied.¹² EY frames the core principles worth building toward as accountability, data protection, reliability, security, transparency, explainability, fairness, compliance and sustainability. It is a long list, but the underlying point is short: governance cannot be the department that shows up after the innovation is already live. It has to be designed in from the start.”
In practice, that means building real capability — not just policy documents — around a consistent set of basics:
Regulators are moving in this direction at the policy level, not leaving it purely to individual companies to work out. If we take an example, the AI Law of Vietnam also lays the legal foundation, Decree 142/2026/ND-CP fills in the practical implementation details, and the Ministry of Science and Technology’s national AI ethics framework, issued under Circular 05/2026/TT-BKHCN, sets out the safety and human-oversight principles organizations are expected to build around.¹³ For enterprises elsewhere, none of this is only a compliance checkbox. Trust is becoming a genuine competitive advantage — customers, partners, and regulators are all starting to notice which companies build it in.
Some of the most visible AI transformation is going to happen in the most ordinary place: where people actually work every day. Enterprise communication has moved through a fairly predictable sequence over the past two decades — email, then messaging, then collaboration platforms. The next step in that sequence is what we would call an intelligent workspace.
Imagine an employee who can ask their workspace to summarize today’s important customer issues, prepare the management briefing, find the relevant clause in a contract, draft a reply to a customer, list what is still unresolved from last week’s meeting, compare two proposals and recommend one, or put together a first draft of a project plan. In that kind of environment, AI stops being something you occasionally query for an answer. It becomes the interface to the organization’s knowledge and workflows.
That is the thinking behind AI Connect, SpeedUp’s enterprise communication and productivity platform: one workspace where people work directly with organizational knowledge, their existing applications, and AI agents, instead of switching between five disconnected tools to get anything done. The goal was never to replace employees with it. It is to build a genuine human-AI partnership, where people bring judgment, accountability, relationships, and creativity, and AI brings speed, memory, pattern recognition, and execution. The organizations that learn to combine those two well are going to pull ahead of the ones treating AI as a bolt-on feature.
Legal work is a useful case study for a broader principle: not every AI problem in the enterprise should be solved with a generic chatbot. Legal teams work with proprietary documents, dense regulation, live contracts, precedent, and genuinely sensitive information — exactly the kind of material where a general-purpose model without the right guardrails and grounding can do real damage.
A properly built enterprise legal AI environment helps professionals search legal knowledge quickly, analyze and compare contract clauses, identify obligations buried in long agreements, summarize case material, track regulatory changes as they happen, prepare first drafts, organize evidence, and support deeper legal research, considerably faster than doing it by hand. What it does not do is replace professional judgment, and it should not try to. The realistic future is not AI replacing lawyers. It is lawyers working with legal intelligence they can actually trust and verify.
The same principle holds well beyond legal work. In finance, HR, procurement, supply chain, and customer service, the pattern repeats: the value comes from combining solution consulting, enterprise AI, and purpose-built specialized agents, rather than dropping a generic AI tool into a specialized workflow and hoping it fits. That is the approach we have built SpeedUp’s own Legal AI and agent portfolio around.
Across the industries, it helps to think of AI transformation across five practical stages where each one building on the last, at whatever pace an organization and the technology actually allow. This clear sequence of stages is:
Stage 1 — Build the foundation
Stage 2 — Integrate
Stage 3 — Redesign the workflow, not just the interface
Stage 4 — Scale intelligent operations
Stage 5 — Become AI-native
At this stage, AI is no longer a program with a name and a budget line. It is simply part of how the company runs — continuously redesigning its processes, products and customer experiences around the intelligence available to it.
There is a real difference between automation and transformation, and it is worth being precise about it. Automation asks how to get today’s work done faster. Transformation asks a different question: what could this business become if intelligence were available everywhere in it?
That distinction has real consequences. If AI only trims administrative work, the organization captures efficiency and stops there. If AI frees people to spend more time on customers, innovation, strategy and the decisions that genuinely need a human, the organization creates capacity for growth instead.
“McKinsey’s research backs this directly: the companies capturing the most value from AI increasingly pair efficiency goals with growth and innovation goals, rather than chasing efficiency alone.¹⁴”
That should be the ambition of every business too — not just doing the same work more cheaply but building the capacity to do genuinely new things.
At SpeedUp, we think about the AI transformation journey as one connected ecosystem, not four separate product lines. Sovereign AI infrastructure gives sensitive workloads a trusted foundation. Solution consulting connects business priorities to a realistic transformation roadmap instead of a generic one. AI Connect gives employees an intelligent communication and productivity environment to actually work in. Legal AI and our specialized enterprise agents bring real intelligence into high-value professional workflows that were previously too specialized for generic tools. Together, the goal is straightforward: help organizations move from AI experimentation to AI transformation that is secure, measurable and built to scale.
We do not think the opportunity is to predict every technology that will show up over the next five years — nobody can do that honestly. It is to build organizations capable of adapting to whatever shows up, deliberately and quickly.
AI is moving too fast for any five-year technology plan to survive unchanged, but business leaders can still hold onto a few durable principles: build intelligence into the business itself, not around its edges. Protect what genuinely needs to stay sovereign. Govern AI by design, not after an incident forces the question. Measure value, not activity. Redesign workflows, not just interfaces. Augment your people instead of simply automating around them. Build AI capability broadly across the organization, not inside one team alone. And above everything else, stop asking what AI can do, and start deciding what your business should become because of it.
For businesses in Vietnam, this is particularly a perfect moment because the country’s Law on Artificial Intelligence took effect on 1 March 2026, providing a comprehensive legal framework for how AI is developed, deployed, and used — one built around human oversight, safety, innovation, and national sovereignty.³ A follow-up decree and a national AI ethics framework have since filled in much of the practical detail.⁴ So the question for the businesses was never really whether they would enter the AI era. It is how deliberately they choose to do it.
The AI era will not be won by whoever deploys the most models. It will be won by the organizations that learn fastest, adapt responsibly, and turn intelligence into business value that shows up on the balance sheet. For us, that is more than a technology opportunity — it is a chance to build a more productive, innovative, secure, and genuinely competitive economy. And for every enterprise reading this, the question has already changed. It was never about whether AI would transform the business. It is whether the business will transform itself fast enough to lead.
1. World Economic Forum, Business Transformation in the Artificial Intelligence Era — leadership, workforce transformation and continuous organizational change, weforum.org.
2. World Economic Forum & Bain & Company, AI Infrastructure in the Age of Sovereignty — compute, data, connectivity and resilience, weforum.org.
3. Government of Vietnam, Law on Artificial Intelligence No. 134/2025/QH15, effective 1 March 2026.
4. Government of Vietnam, Decree No. 142/2026/ND-CP implementing the Law on Artificial Intelligence, effective 1 May 2026; Ministry of Science and Technology, National AI Ethics Framework, Circular No. 05/2026/TT-BKHCN, effective 10 March 2026.
5. McKinsey & Company, The State of AI in 2025: Agents, Innovation and Transformation.
6. PwC, 2026 AI Global Jobs Barometer — productivity, skills and workforce transformation.
7. McKinsey & Company, The State of AI: How Organizations Are Rewiring to Capture Value — workflow redesign, governance, and measurable value.
8. WOBI / World Business Forum, What You Missed at WOBI on AI & Business Transformation, wbf.wobi.com.
9. World Economic Forum & Bain & Company, Rethinking AI Sovereignty — strategic investment and trusted interdependence, weforum.org.
10. McKinsey & Company, infrastructure evolution analysis — from on-premise to cloud to hybrid, GPU-accelerated, AI-oriented infrastructure.
11. World Economic Forum, AI and Business: Agents, AI+, Data, Ethics and More — AI agents, workforce and responsible adoption, weforum.org.
12. EY, Responsible AI Pulse, 2025 — responsible AI, governance and business outcomes; EY, How Boards Can Confidently Steer an AI-Enabled Future.
13. Ministry of Science and Technology, National AI Ethics Framework, Circular No. 05/2026/TT-BKHCN.
14. McKinsey & Company, The State of AI in 2025: Agents, Innovation and Transformation.
15. European Union, Artificial Intelligence Act (Regulation (EU) 2024/1689), entered into force 1 August 2024.
16. OECD, OECD AI Principles (updated 2024); national AI legislation and implementing frameworks now in force across multiple jurisdictions in Asia, the Middle East and the Americas.
17. Illustrative examples include Singapore’s Model AI Governance Framework, the UK’s pro-innovation AI regulatory approach, and national AI ethics frameworks issued by science and technology ministries across Asia-Pacific.
