FIELD NOTE / D04
Top-Down vs Bottom-Up AI Adoption for Hong Kong SMEs
Who starts AI transformation is a very real question for Hong Kong SMEs: should the boss learn first, or the team? Push top-down, or let momentum build from below? The answer is not either-or. The boss and the team each have roles that cannot be outsourced. Failed transformations almost always have one side missing.
The most workable model for AI transformation is the boss setting direction and boundaries, with the team owning the processes and execution. Pure top-down becomes directive procurement. Pure bottom-up stays at the level of scattered individual experiments. Only when both sides move in step does it become organisational capability. DKH is a Hong Kong AI-first digital consultancy with a training-before-tools approach.
TL;DR
- Owner-led push: fast but shallow, and it easily becomes an "owner's toy".
- Pure team push: solid but scattered, no boundaries, no resources, cannot scale.
- Most workable: the boss does three things (direction, boundaries, resources), the team does three things (process, experimentation, review).
- The first step is to start with one real workflow and review the results together.
What are the benefits and pitfalls of the boss leading the charge?
The biggest benefit of the boss leading from the front is speed and signal. When the boss personally uses AI to tidy up meeting notes, the team immediately understands that this is company direction. It does not read as a passing fad.
The trap is that the boss's usage may not reflect the team's actual needs. The boss may find AI useful for drafting emails. The bottleneck for marketing colleagues, however, may be asset production and report organisation. If transformation becomes "the whole company learning the boss's way", it will quickly lose relevance.
A second trap is command-style rollout. This means ordering everyone to use AI with no training and no workflow design. As our other article on why transformation fails makes clear, most initiatives collapse precisely because they put tools before people.
What happens when the team takes the initiative?
Many companies already have colleagues quietly using AI. This spontaneous adoption is a valuable starting point, because the approaches they discover stay close to real work.
Pure bottom-up, however, runs into three ceilings.
The first is no boundaries. Colleagues do not know whether company data can go into AI tools. So they either avoid using them or use them carelessly, which is a data security risk.
The second is no resources. Useful tools cost money, and without a budget the team makes do on free versions.
The third is no mechanism for spread. One person's efficient practice has no channel to become the whole team's standard workflow.
So how should the work be divided?
The division of labour that works best in our consultancy work is this:
The boss handles three things:
The team handles three things:
The essence of this division is that the boss does not micromanage the tools or how they are used. The team does not have to guess the company's position.
SMEs have small teams. Can the boss double up as part of the team?
Many Hong Kong SMEs have only a handful of people, and the owner is often the most active executor. Combining roles is a practical reality in that situation. Keep at least one separation, though. Wear the owner hat when setting boundaries, wear the executor hat when testing workflows, and do not conflate "I am testing this myself" with "the company already has rules".
DKH's own brand Skin Nods operates this way as a small team. AI workflows support daily content and operations, and people oversee brand direction and judgement. A small team achieves what a much larger one would. This is a model we validated in our own business before offering it to clients.
What does step one actually look like?
You do not need an all-hands meeting to announce a transformation.
The lightest possible start: the boss and the team pick one workflow everyone finds frustrating. Monthly reports, first drafts of social content, or client follow-up messages are good examples. Agree the boundaries, let the team redo it with AI for two weeks, then review the results together.
This single loop will naturally surface what your company needs to address. That might be training, tools, or workflow design.
Start from evidence, not from a position.
Next step: hand it to us. Free 15-minute consultation, leave your details on the contact page and we will WhatsApp you within one working day.
FAQ
FAQ
- Can a boss who knows nothing about AI still drive transformation?
- Yes. The owner's role is to set direction, boundaries and resources, and all three depend on business judgement rather than technical knowledge. That said, the owner should have personally used one or two basic applications, otherwise it is difficult to make a reasonable assessment of the team's output.
- Some people on the team are strong, others are completely lost. How do you balance it?
- Make the fast movers champions, responsible for trying new approaches and documenting the method. Slower movers start with proven workflows so they do not have to figure things out from scratch. Never measure the whole team with the same ruler. Wide variation in the early stage is normal.
- What is the role of middle managers in transformation?
- Middle management is the most critical and most easily overlooked layer. They hold decision-making authority over daily processes, and the process reorganisation that transformation requires is in practice approved by them. We recommend involving middle management from the outset in selecting processes and setting the review rhythm, rather than simply notifying them after decisions are made.
- When the boss and the team disagree, who takes precedence?
- Direction and boundaries are for the boss to decide. Process and tools are for the team. If a disagreement sits in the middle ground, replace debate with a small-scale trial: run it for two weeks, then bring the data back to the table.
KEEP READING
Keep reading
NEXT STEP